AI+D Lab

Teaching AI Intuition: Iterative Development of a Foundational Course Format

Mon, 30 Mar 2026 00:00:00 GMT

Iterative Development of a Foundational Course Format for AI Education in Design

How do you teach designers to work meaningfully with AI — without training them to become developers, and without reducing the technology to a set of new tools in the creative toolkit? This question has driven the development of a foundational course in AI and Design, running since the summer semester of 2023 within the fourth semester of the Interaction Design programme at HfG Schwäbisch Gmünd.

The course makes the technical complexity of AI accessible to design students through a deliberate combination of theoretical input and hands-on experimentation. Its overarching goal is the development of AI Intuition [1], a grounded understanding of how AI systems work, what they can and cannot do, and how to engage with them critically and purposefully as a design material. This intuition equips students to develop a precise vocabulary for AI concepts, make informed decisions about when and how to use these technologies, and reflect critically on their societal, ethical, and ecological implications.

Over three years and six iterations, the course has been continuously reshaped in response to student feedback, technological change, and our own evolving understanding of what design education at the intersection of AI actually requires.

Course Modules

The course is built around a set of foundational modules, each addressing a different dimension of AI that is relevant to design practice. Together, they form a deliberate progression in terms of both content and complexity: from understanding the underlying functional principles of neural networks, to prototyping, to designing interfaces and interactions with and for AI, to working with generative AI technologies that increasingly shape creative and communicative processes.

Module 1: AI Fundamentals for Designers

A broad understanding of how AI technologies work is a prerequisite for using them thoughtfully as tool and as design material. The first module provides an accessible entry point into the landscape of AI technologies relevant to design practice, covering image classification, object detection, and segmentation, as well as text and image generation, speech synthesis, and sensor data analysis for interactive applications.

This overview is followed by an introduction to the principles of neural networks. To make these abstract concepts tangible, students work with SandwichNet, an interactive tool developed at the AI+Design Lab that visualises how neural networks learn by training a model to distinguish between edible and inedible sandwiches. By manipulating parameters and observing the model's behaviour directly, students develop an intuitive feel for concepts that would otherwise remain purely theoretical.

The module also lays the foundation for an overarching mindset that we aim to promote throughout the course: AI is reshaping design practice in fundamental ways, changing not only the tools designers use, but the nature of the systems they design, the processes they follow, and the questions they need to ask. As technology shifts from passive tool to active counterpart, entirely new design challenges emerge around collaboration, autonomy, trust and ethics. This makes it all the more important that designers take an active and critical role in shaping future human-AI systems, moving well beyond the mere application of tools.

Module 2: Prototyping AI-Based Interactions

The second module makes the abstract principles introduced in Module 1 physically tangible through what we call Physical AI [2]: students work with low-cost microcontrollers (Arduino Nano 33 BLE Sense) and sensor data to build their own machine learning models, training them to recognise gestures and movement patterns in real time.

Students work through the complete ML development cycle, from collecting and labelling their own sensor data, to training a model, to deploying it on hardware that interacts with the physical world. For the technical implementation, they use Edge Impulse, an online platform that streamlines the entire workflow without requiring extensive programming knowledge. The use of affordable, accessible hardware and self-generated datasets keeps the models small and their behaviour legible, making the underlying mechanics of machine learning more directly observable.

Beyond the technical fundamentals, the module opens up a design space: by working with sensors that detect subtle movements and complex motion patterns, students begin to explore interaction forms that go well beyond conventional input methods and to ask what new possibilities these might create for human-machine communication.

Module 3: Designing Interfaces and Interactions with and for AI

AI-based systems challenge many of the assumptions that have long underpinned interaction design. Unlike traditional software, they do not follow fixed, predictable rules. Their outputs vary, their behaviour is difficult to predict, and their capabilities are harder to communicate to users. This raises urgent questions around trust, transparency, and expectation management. Students analyse Google's PAIR Guidelines alongside Nielsen's and Norman's ten usability heuristics, asking which classical UX principles still hold, where AI introduces entirely new challenges, and how established methods need to be adapted for intelligent systems.

At the same time, AI opens up new design possibilities for personalisation, adaptive interfaces, and entirely new forms of human-machine interaction. One particularly rich and complex dimension concerns the relationships that form between users and AI systems, whether intentionally designed or not. This raises fundamental questions around animism, anthropomorphism, and the ethical implications of designing systems that evoke emotional responses. Students explore both the potential and the risks of these approaches, including the concept of "Otherware" by Hassenzahl and colleagues [3], which proposes alternative design paradigms beyond anthropomorphic patterns.

Module 4: How to Talk to Transformers — An Introduction to Language Models

Transformer-based language models have become central to a growing range of creative, communicative, and design processes. Yet for most designers, they remain opaque: powerful tools whose outputs can be steered through prompting, but whose underlying logic stays hidden. This module builds a precise enough understanding of how these models work to use them deliberately and critically, grounded in technical concepts, but oriented towards design practice.

Students learn how language is broken down into tokens, how these are organised as vectors in an embedding space, and how meaning emerges through self-attention and feed-forward networks. Building on this foundation, students develop a strategic prompting competency: learning to navigate the embedding space through precise word choices, to use examples rather than abstract descriptions, and to control parameters such as temperature to balance creativity and precision.

The module combines analytical frameworks, hands-on prompting experiments, and embodied exercises such as spatially mapping semantic relationships, to make the relational, context-dependent nature of transformer-generated meaning tangible. As a concluding exercise, students design their own transformer persona: a coherent AI character defined through a system prompt, tone of voice, example interactions, and temperature settings. This task makes immediately visible how technical parameters, linguistic patterns, and design decisions combine to shape AI behaviour.

Module 5: Image Generation

Generative image tools have rapidly become part of everyday creative practice, making it essential for designers to move beyond the belief that image generation is just about finding the perfect prompt. The module combines structured theoretical input with hands-on exercises, including prompt challenges in small groups, to build both technical literacy and practical confidence.

The theoretical foundation focuses on diffusion models, the architecture behind most widely used image generation systems today, including proprietary tools such as DALL-E, Midjourney, and ChatGPT's image generation, as well as openly available models like Stable Diffusion and Flux. Students learn how these models are trained, how they generate images from a learned representation of a training dataset, and key concepts such as denoising and CLIP embeddings. Equally important is understanding the differences between proprietary and open models in terms of accessibility, transparency, and customisability, as these have direct implications for reflective creative practice. The course consistently encourages the use of open models for their flexibility and adaptability.

The practical core of the module is a series of Prompt Challenges, in which students are asked to reproduce a given target image using Stable Diffusion. Through the process of trying to match a specific visual outcome, students encounter the gap between intention and result and incrementally learn different image generation techniques, including text-to-image and image-to-image generation, prompt design, and more advanced methods such as using ControlNet, reference images, and inpainting. Given the pace of development in this field, the module evolved considerably across semesters, reflected in our use of different open-source interfaces for Stable Diffusion, from Automatic1111 and ComfyUI to Invoke AI.

The module closes with a critical reflection on AI-generated imagery: students are encouraged to read generated images not as neutral outputs, but as invocations of training data and the human decisions that shaped it, engaging with questions of bias, characteristic aesthetics, and the cultural implications of these systems. Drawing on the Better Images of AI initiative, they are challenged to produce work that moves beyond the visual stereotypes that AI image generation tends to reproduce.

As the field evolved and our own understanding of what designers need deepened, two further modules were added to the course.

Module 6: Introduction to Agentic Design

Agentic AI — systems that can plan and execute tasks autonomously — represents one of the most significant recent developments in the field. Designing autonomous systems requires a fundamentally different mindset: no longer just shaping interfaces, but defining behaviour, boundaries, and decision-making processes — and asking when autonomy is actually the right choice.

Students learn to distinguish between defined workflows and autonomous agents: while workflows orchestrate processes through fixed sequences, agents dynamically steer their own processes and make independent decisions about tool use and execution strategy. To understand the underlying architecture, students explore the four core components of an agent: the large language model as central intelligence, the system prompt as its operating instruction, the memory component for short and long-term context, and tools for interaction with external systems such as APIs or databases.

For practical implementation, students work with n8n, a visual automation platform that enables rapid prototyping as well as more complex implementations. As a hands-on exercise, they build an agent configured as a personal creative sparring partner, developing an understanding of when agents are genuinely useful and when the reliability of a defined workflow is the better choice.

Module 7: Playful Ideation with Generative AI

Unlike the other modules, Playful Ideation is less concerned with technical understanding and more with creative practice and attitude. It responds to two problems that emerged from working with generative AI in design contexts. First, creative work with generative AI tends to be highly cognitive and outcome-focused, driven primarily by text-based prompting. Yet creative processes thrive on messiness, iteration, and creative friction. The module offers students concrete tactics to reintroduce these qualities into their work with AI, making co-creation less cognitive and more intuitive and action-based. Second, AI-generated content tends to look flat, glossy, and increasingly similar — a uniformity that reduces diversity and, with it, the conditions for creative thinking, ideation and innovation. Ambivalence and surprise, however, are crucial for fostering creativity and imagination. This module explores strategies to promote surprise, unexpected results and to encourage "otherness".

The technical setup of the module builds on Transferscope, a tool that combines image generation with ControlNet to enable a deceptively simple but powerful interaction: with a single button, users can capture the visual style of any object or concept and transfer it onto another object or scene. This one-button interface deliberately removes the complexity of text-based prompting, encouraging a more playful, intuitive, and exploratory engagement with generative AI.

Connecting the Modules: A Hands-On Assignment

One recurring challenge in earlier iterations of the course was that the modules felt isolated from one another — students engaged with each topic individually, but the connections between them were not always visible. In response, the two most recent iterations introduced a hands-on assignment designed to tie the modules together by applying them into a final project.

Real-world AI systems rarely rely on a single technology or modality, but on a combination of more of them. Language models are combined with image generation, physical sensors feed into agentic workflows, and multimodal interactions draw on several of these layers at once. The assignment is designed to make this visible, giving students the experience of combining the tools and concepts from across the week into a coherent, functioning system. Beyond the technical integration, the assignment also marks a first step towards conceptual engagement and use case development, inviting students to explore interesting scenarios and applications for AI systems in a playful, low-stakes way.

Two different assignment formats were developed and tested across iterations. The first uses Input/Output Cards as a starting point to explore combinations of input and output modalities across different AI types.

<cite>Example of combinations of the Input/Output Cards for concept generation</cite>

Students are given different combinations of input/output cards at random and are encouraged to ideate concepts that make sense of them. The use of these constraints aims to encourage students to think beyond familiar concepts and outside the box, building an understanding of how different input/output modalities can be orchestrated into coherent interaction concepts in a playful, low-threshold way.

<swiper-container navigation="true" cssMode="true" grabCursor="true" loop="true"> <swiper-slide> <cite>A concept created with the Input/Output Cards: Gloomy, a lamp which based on facial expressions (input) provides visual and auditory feedback (output), by Rebeka Tot and Vivien Cai.</cite> </swiper-slide> <swiper-slide> <cite>A concept created with the Input/Output Cards: Gloomy, a lamp which based on facial expressions (input) provides visual and auditory feedback (output), by Rebeka Tot and Vivien Cai.</cite> </swiper-slide> </swiper-container>

The second assignment is a more technically integrated pipeline that connects sensor-based machine learning, language models, and image generation into a single agentic workflow. A physical sensor captures events from the environment and sends them to an n8n workflow, where a storytelling agent generates a narrative. At key moments in the story, the agent produces images to accompany the text, maintaining visual consistency through image-to-image generation. Students configure the entire pipeline themselves, making design decisions at every level: from the sensor interaction and the agent's narrative style, to the visual aesthetic of the generated images.

<swiper-container navigation="true" cssMode="true" grabCursor="true" loop="true"> <swiper-slide> <cite>The project by Trang-Anh Nguyen and Emily Ulrich tells stories of everyday life in a café from the perspective of a plant. It perceives its surroundings through the vibrations of the machine and interprets the events happening around it.</cite> </swiper-slide> <swiper-slide> <cite>The project by Trang-Anh Nguyen and Emily Ulrich tells stories of everyday life in a café from the perspective of a plant. It perceives its surroundings through the vibrations of the machine and interprets the events happening around it.</cite> </swiper-slide> </swiper-container>

Cross-Cutting Elements

Alongside the individual modules, some recurring elements run through the entire course and were introduced in response to specific challenges that emerged across iterations.

The first concerns vocabulary. We noticed early on that key concepts introduced in one module were not being carried forward into the next, and the shared language needed for that was not sticking. In response, we introduced a daily practice of collecting the most important terms from each module on a large poster, displayed throughout the week. This made concepts visually accessible and created a growing reference point that could be explicitly linked to each new day's content.

The second concerns critical reflection. The constant emergence of new AI tools and capabilities tends to generate enthusiasm among students, which is valuable, but needs to be balanced by a critical perspective. After each module, we introduced a structured discussion asking what ethical, societal, or ecological implications the day's content might carry, and how designers can engage with these responsibly. This was not treated as a separate topic, but as an integral part of each module.

In early iterations, we developed a self-study course — the AI Primer — designed to be completed before the block week. It covered foundational concepts, societal implications, and introduced key vocabulary through videos and reading materials. In practice, however, we found that a front-loaded, self-study format was not the most effective way to build this foundation: the content was difficult to discuss and contextualise in isolation, and technical explanations were harder to absorb without hands-on experience to anchor them. Over time, we moved these elements into the course itself — distributing vocabulary, conceptual framing, and critical reflection across the modules, where they could be directly connected to the content of each day. The vocabulary posters and the daily reflection discussions are both expressions of this shift.

Reflections and Learnings

Course Format

Intensive block format proved effective. The course was initially offered in a distributed format, with sessions spread across several weeks. Moving to an intensive one-week block allowed students to devote their full attention to the complex and demanding subject area. Full days rather than isolated few-hour slots allowed concepts to accumulate and connect, creating the conditions for a deeper and more sustained engagement with the material.

Earlier curriculum placement would be preferable. The course has been offered in the fourth semester, and while this has worked well, we would advocate for an earlier placement. There is a genuine tension here: introducing AI tools too early risks shortcutting foundational skills. Evaluating the quality of AI output requires knowing how to do things yourself. At the same time, students are using these technologies regardless, often without a critical or reflective framework. Teaching that framework early, in parallel with foundational skills, seems increasingly important.

Content and Development

Foundational modules are robust, but require ongoing updates. The core structure has remained stable across iterations, and the modules build well on each other. Nonetheless, the specific content needs to be revisited regularly to keep pace with developments in the field.

New topics must be scanned and selectively integrated. The field moves fast, and new themes regularly emerge that are relevant for design education. Integrating them requires making deliberate choices about shifting focus and removing content — the scope of the course is finite.

Teach the principles, not the tools. The tool landscape in AI changes rapidly, and what is standard practice today may be obsolete within a semester. Rather than building the course around specific tools, we focused on teaching the underlying principles and functional logic that transfers across them. Tools were used as vehicles for learning, not as ends in themselves. The goal was to equip students with the conceptual foundation to independently adopt and critically evaluate new tools as they emerge — as illustrated by our own repeated shifts in the image generation module, from Automatic1111 to ComfyUI to Invoke AI.

Pedagogical Approach

Teach co-creation, not automation. Working with AI in creative practice raises fundamental questions about the role of designers and the nature of the design process itself. We consistently encouraged students to engage with AI as a creative counterpart rather than a shortcut — exploring different modes of collaboration, questioning when automation is appropriate, and developing a conscious, reflective stance towards the tools they use. Developing a reflective design practice with AI is, we believe, as important as technical literacy.

The framing of AI must connect to students' own practice. Abstract introductions to AI rarely resonate in a design context. What makes the subject tangible is connecting it directly to students' existing design practice and motivations — a framing that will inevitably vary depending on the institutional culture and programme.

Hands-on exploration should connect across modules — and towards application. From the beginning, hands-on exploration was central to the course. Over time, however, we recognised that keeping it within individual module silos was not enough — students needed the experience of combining technologies and concepts across modules. The connecting assignment bridges this gap, and in doing so also marks a first step towards conceptual engagement and the development of meaningful AI use cases.

Critical reflection is not a separate topic — it is part of every topic. Ethical, societal, and ecological questions do not sit outside the technical content of the course; they arise directly from it. By embedding reflection into each module, we aimed to make it an instinctive part of how students engage with AI — not something they do afterwards, but something they do throughout.

Conclusion

The course Foundations of AI + Design has evolved considerably since its first iteration in 2023, through ongoing reflection, experimentation, and response to a field that continues to change at a remarkable pace. What has remained stable throughout is the underlying goal: to develop AI Intuition in design students, a broad understanding of how AI technologies work, what they can and cannot do, and how to engage with them critically and purposefully as a design material.

The modular structure, the intensive block format, and the cross-cutting elements of vocabulary-building and critical reflection have proven to be a robust foundation. At the same time, the course has had to remain open to change, adding new modules, refining approaches, and continuously asking what designers actually need to engage meaningfully with AI.

We share it in the spirit of the course itself: as a work in progress, open to critique, iteration, and further development.

Acknowledgements

The following current and former members of the AI+Design Lab were involved in the development and implementation of the course format and the modules described: Rahel Flechtner, Jordi Tost, Felix Sewing, Christopher Pietsch, Ron Mandic, Moritz Hartstang, Aeneas Stankowski, and Alexa Steinbrück.

References:

[1] Flechtner, R., & Stankowski, A. (2023). AI Is Not a Wildcard: Challenges for Integrating AI into the Design Curriculum. Proceedings of the 5th Annual Symposium on HCI Education, 72–77. https://doi.org/10.1145/3587399.3587410 [2] Rahel Flechtner, Jakob Kilian, Ivan Iovine. 2025. Physical AI: Sensor-Based AI in Art and Design. In: Florian Jenett, Rahel Flechtner, and Simon Maris (Eds.). 2025. un/learn ai : navigating AI in aesthetic practices. Hochschule Mainz. https://doi.org/10.25358/openscience-11811 [3] Marc Hassenzahl, Jan Borchers, Susanne Boll, Astrid Rosenthal-von der Pütten, and Volker Wulf. 2021. Otherware: how to best interact with autonomous systems. Interactions 28, 1: 54–57. https://doi.org/10.1145/3436942

Steering Generative Models for Accessibility: EasyRead Image Generation using OpenMoji

Wed, 25 Mar 2026 00:00:00 GMT

{frontmatter.title}

<div style="opacity: .65; margin-bottom: 6rem;"> This is a guest post by Nicolas Dickenmann and co-authors from ETH Zurich / UNICEF. </div>

Artificial intelligence is rapidly transforming how visual content is created. Generative models can now produce stunning, intricate images in a matter of seconds. But while these highly detailed outputs are visually impressive, more detail isn't always better.

EasyRead is a format designed to make information universally understandable, pairing simple text with clear, unambiguous pictograms. It is an essential communication tool for people with intellectual disabilities, learning difficulties, or limited language proficiency. Unfortunately, modern image generation systems are not designed with cognitive accessibility in mind. They default to producing visually complex, abstract, or cluttered outputs that directly contradict EasyRead guidelines.

This gap is the motivation behind our recent paper, Steering Generative Models for Accessibility: EasyRead Image Generation, which was recently accepted at the upcoming CHI26 conference. In it, we tackle a critical question:

Can we teach state-of-the-art generative models to prioritize cognitive accessibility and visual simplicity over complex aesthetics?

Method: A Custom Pipeline for Cognitive Accessibility

To solve this, we couldn't just rely on standard prompt engineering, as base models still tend to inject unwanted details. Instead, we needed to systematically steer the model’s visual representations by building a dedicated EasyRead pipeline.

Our pipeline consists of a few key steps:

Curating and Enhancing the Data: This is where the incredible foundation built at HfG Schwäbisch Gmünd by the OpenMoji project became essential to our research. To teach our model what a consistent, clear pictogram looks like, we used OpenMoji's extensive, open-source library as a core pillar of our training data, combining it with complementary datasets from ARASAAC and LDS. Because standard emoji labels are naturally brief, but a bit too sparse for training a complex diffusion model, we used the BLIP image captioning model to generate rich, natural-language descriptions.
Strategic Augmentation: To enable targeted customization at inference, we leveraged the ARASAAC API to systematically augment the training images with varying background colors, skin tones, and hair colors.
Parameter-Efficient Fine-Tuning: We fine-tuned Stable Diffusion v1.5 using Low-Rank Adaptation (LoRA) on the attention layers. By keeping the core model weights frozen and introducing a unique instance token (sks), we successfully taught the model the strict "EasyRead" style without losing its broad semantic knowledge.

<figure> <figcaption><strong>Figure 1:</strong> Examples of the three datasets in our corpus: ARASAAC pictograms, OpenMoji icons, and LDS symbolic illustrations.</figcaption> </figure>

<figure> <figcaption><strong>Figure 2:</strong> Our pipeline at training and inference time</figcaption> </figure>

Measuring Success: Introducing the EasyRead Score (ERS)

A major bottleneck in accessibility research is evaluation. To bridge this gap, we introduced the EasyRead Score (ERS), the first quantifiable metric that formalizes accessibility-oriented design principles. The ERS is an aggregate of six distinct sub-metrics that evaluate:

Palette Complexity (Low visual clutter)
Edge Density (Simple geometry)
Saliency Concentration (Clear visual focus)
Foreground-Background Contrast (Strong separation)
Centering Error (Stable layout)
Relative Stroke Thickness (Consistent outlines)

<figure> <figcaption><strong>Figure 3:</strong> The EasyRead Score (ERS) consists of sub-metrics to capture the properties of EasyRead pictograms.</figcaption> </figure>

Results: Measurable Clarity and Precision

<figure> <figcaption><strong>Figure 4:</strong> EasyRead pictograms generated with our finetuned model</figcaption> </figure>

We evaluated our fine-tuned model quantitatively against the baseline Stable Diffusion v1.5 across 275 generated images. The data showed that our lightweight adaptation substantially improved both style and semantic accuracy. Additionally, we qualitatively evaluated our model against closed-source systems, and noticed substantial advantages on stylistic consistency and instruction following.

Looking ahead

Generative AI shouldn't just be a tool for creating complex digital art; it needs to be a practical utility for making the world more accessible. Moving forward we see exciting real-world applications, reducing cost of accessible communication while increasing availability. By designing generative systems with accessibility in mind, we can ensure the AI revolution leaves no one behind. Check out our full paper here, and explore our open-source code and checkpoints on GitHub!

<div style="opacity: .65; margin-top: 6rem;"> Nicolas Dickenmann is affiliated with a research team at ETH Zurich / UNICEF. All rights to the images remain with their respective copyright holders. </div>

KITeGG publication 'Un/learn AI'

Fri, 28 Nov 2025 00:00:00 GMT

Finally out! We are happy to announce that all three KITeGG publications have been released and are available online and as PDF downloads. The three-part publication series “Un/learn AI” with the titles “Approaching,” “Navigating,” and “Integrating AI in Aesthetic Practices” offers insights into the KITeGG research project. The three books contain articles by KITeGG researchers and guest lecturers, student projects, and contributions from the summer and winter schools.

On the website, you can also find more information about the publication's innovative approach in terms of its experimental AI-powered editorial design, where semantic chunking creates thematic links for alternative reading paths.

Photo by Julia-Jasmin Bold and Maika Dieterich.

Predicting Comprehensibility in Scientific Text Based on Word Facilitation

Mon, 03 Nov 2025 00:00:00 GMT

PhD Progress Update:

Towards making every scientific text comprehensible for anyone.

Motivation:

Despite the growth of open and accessible science [Klebel et al., 2025], levels of scientific literacy are declining [OECD, 2023]. Closing this gap requires writing that communicates science in ways that remain comprehensible across disciplines and to broader audiences. A key step toward this goal is the ability to predict scientific text comprehension.

In this work, we derive a novel metric for predicting scientific text comprehension based on the attention mechanism in transformer architectures (Figure 1). Attention scores quantify how strongly each word facilitates the processing of other words in context [Vaswani et al., 2017]. Here, we evaluate word facilitation based on attention scores as a metric of text comprehensibility.

Figure 1: Abstract transformer architecture of a large language model.

Methods:

We compared the explanatory power of word facilitation (normalized and computed using inverse entropy and directional weighting) against that of traditional linguistic metrics for text comprehensibility, such as length, frequency (SUBTLEX-US corpus [Brysbaert et al., 2012]), surprisal (negative log probability from GPT2), and predictability (human rating on a 1–5 scale) using the dataset from [de Varda et al., 2024]. This dataset contains reading times and neural activations reflecting text processing, recorded for each word in 205 English sentences.

Figure 2: Pearson correlations (r) between predictors and both reading times and neural activations reflecting text processing. Word facilitation shows the strongest overall correlation (mean |r| across measures).

Results:

Word facilitation demonstrates the strongest overall correlation with text processing measures, outperforming traditional predictors (length, frequency, surprisal, and predictability) across both reading times and neural activations (Figure 2).

Word facilitation especially explains unique variance for long, less frequent, and less predictable words. It even explained the most unique variance for first fixation reading times in quantiles with the longest words. A measure that reflects initial lexical access [Demberg and Keller, 2008] and is normally dominated by length and frequency (Figure 3).

Figure 3: The unique variance (∆R²) explained by a predictor for reading times (first fixation duration and go-past reading time) and neural activations (N400 and P600 amplitude) reflecting integration difficulty. Unique variance is shown across word property quantiles.

Conclusion:

We introduced an attention-based metric of text comprehensibility, based on how strongly each word facilitates the processing of other words in context within a transformer model.

Word facilitation complements the traditional linguistic metrics used to predict text comprehensibility by explaining late reading times and neural activations reflecting the integration of a word into its broader sentence context. It especially explained unique variance for words that the existing linguistic metrics misclassify as difficult, highlighting its sensitivity to context and its suitability for scientific texts rich in long, infrequent, and less predictable yet contextually comprehensible words.

Ressources

M. Brysbaert, B. New, and E. Keuleers. Adding part-of-speech information to the subtlex-us word frequencies. Behavior research methods, 44(4):991–997, 2012.

A. G. de Varda, M. Marelli, and S. Amenta. Cloze probability, predictability ratings, and computational estimates for 205 english sentences, aligned with existing eeg and reading time data. Behavior Research Methods, 56(5):5190–5213, 2024.

V. Demberg and F. Keller. Data from eye-tracking corpora as evidence for theories of syntactic processing complexity. Cognition, 109(2):193–210, 2008.

T. Klebel, V. Traag, I. Grypari, L. Stoy, and T. Ross-Hellauer. The academic impact of open science: a scoping review. Royal Society Open Science, 12(3): 241248, 2025.

OECD. PISA 2022 Results (Volume I) The State of Learning and Equity in Education: The State of Learning and Equity in Education. OECD Publishing, 2023. ISBN 9789264351288.

A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, Ł. Kaiser, and I. Polosukhin. Attention is all you need. Advances in neural information processing systems, 30, 2017.S. Wiegreffe and Y. Pinter. Attention is not not explanation. arXiv preprint arXiv:1908.04626, 2019.

The corresponding poster was presented at the AI in Science Summit 2025.

Building AI Intuition: Four Pillars for Teaching AI in Design Education

Wed, 01 Oct 2025 00:00:00 GMT

How do we prepare design students to work meaningfully with AI technologies? Over the past three years, we've developed and tested a comprehensive approach grounded in four educational pillars.

This article summarizes "Building AI Intuition – Four Educational Pillars for Teaching AI in Design at AI+D Lab" [1], published in the third volume of the KITeGG un/learn AI series. The full text offers detailed descriptions of teaching formats, tools, and insights from three years of practice. Read the complete publication at unlearn.gestaltung.ai or access our article directly via the “Read the paper”-button.

The Challenge

As AI increasingly becomes part of the design landscape, a critical question emerges: how do we prepare students to engage with AI thoughtfully, critically, and creatively? The challenge goes beyond technical training. Students need to understand when to leverage AI and when to rely on their human intuition. They need to assess AI's capabilities and limitations, develop precise language to communicate about these systems, and critically evaluate their societal, ethical, and environmental impacts.

In 2023, we outlined our ambition to provide students with what we call AI Intuition—a broad understanding of AI's working principles and constraints that enables meaningful engagement at multiple levels [2]. This article synthesizes three years of experimentation, iteration, and learning at the AI+Design Lab, presenting our approach grounded in four complementary pillars.

Four Pillars for AI Intuition

Our approach rests on four interconnected pillars, each contributing to developing AI Intuition:

1. Improving Technical Literacy

Understanding the foundational principles of AI is essential for practical engagement at the intersection of AI and design. However, technical literacy doesn't require turning designers into machine learning engineers. Instead, we focus on developing an intuition for what AI can and cannot do, covering the foundational principles of neural networks, the specifics of sensor-based AI, image generation models, and large language model architectures.

To make these foundational principles tangible, we've developed playful teaching tools. SandwichNet visualizes neural networks through training a sandwich classifier, while Acting out AI Systems has students physically role-play AI components to understand system functionality and limitations from the inside out.

<swiper-container navigation="true" cssMode="true" grabCursor="true" loop="true"> <swiper-slide> <cite>A screenshot of the SandwichNet application.</cite> </swiper-slide> <swiper-slide> <cite>Acting out AI Systems: Students acting as human sensors and recording a data stream of a thermal camera on paper.</cite> </swiper-slide> </swiper-container>

2. Fostering Hands-on Exploration

The complexity and non-deterministic nature of AI makes it challenging to approach as design material. Traditional creative processes rely on what Donald Schön called "reflection-in-action"—ideation that unfolds through direct engagement with materials. To enable this with AI, we prioritize hands-on engagement and a learning-through-making mindset.

One focus area is Physical AI—machine learning on microcontrollers that interact through sensors and actuators. Their limited computing power makes models small and functionality easier to grasp. Using Edge Impulse and Arduino boards, students experience the full pipeline from data collection to deployment. Reusable code elements for sensing and acting allow students to quickly turn concepts into interactive prototypes.

In the field of generative AI, we provide students with accessible tools by hosting a Stable Diffusion service and open source language models on our project server infrastructure. Custom-built tools like Transferscope—both a physical Raspberry Pi device and web platform—encourage playful engagement with image generation. The tool lets users capture the style of one object and transfer it to another with a one-button interface.

<swiper-container navigation="true" cssMode="true" grabCursor="true" loop="true"> <swiper-slide> <cite>Physical AI Prototype: Sound-based snake game controller by Ron Eros Mandic and Lukas Speidel.</cite> </swiper-slide> <swiper-slide> <cite>Transferscope. Physical device (left) and image generation process (right). Project by Christopher Pietsch.</cite> </swiper-slide> </swiper-container>

3. Enabling Conceptual Engagement

Developing meaningful AI applications poses a particular challenge for designers. To support ideation, we've created methods that foster creative friction and encourage unconventional thinking.

Our imagination cards—inspired by movements like Fluxus and Oulipo—provide random constraints that encourage creative exploration. The Input/Output cards, for example, combine sensor technologies with output media, prompting students to consider both technical and conceptual possibilities.

We explore multiple approaches to AI interaction design. Metaphors serve as design resources, as in the "Caring for Machines" workshop that reimagines human-AI relationships through care. Experimental methods like thing ethnography and object personas encourage students to adopt technological perspectives, revealing unique affordances without deep technical knowledge.

The Futures Lenses course format uses generative AI as a co-designer for creating speculative scenarios and fictional design objects. By deliberately applying constraints, ambiguity, or exaggeration in prompts, the course provokes errors and unexpected results that become sources for both critical and creative thinking—helping students defamiliarize everyday objects, expose biases, and gain inspiration through iterative exploration.

<swiper-container navigation="true" cssMode="true" grabCursor="true" loop="true"> <swiper-slide> <cite>Caring for Machines: Social Spark detects awkward silences in conversations and fills them with random facts or jokes. Project by Philipp Maginot and Sebastian Weinert.</cite> </swiper-slide> <swiper-slide> <cite>Experimental methods: Students role-playing a fitness tracker (left) and a digital companion (right).</cite> </swiper-slide> <swiper-slide> <cite>Futures Lenses: A speculative concept (left) and a catalog of fictional things (right), showing a lamp for promoting well-being through symbiotic, reciprocal interactions. Project by Anja Gutmann, Lea Haferbier and Carolin Kaltwasser</cite> </swiper-slide> </swiper-container>

4. Encouraging Critical Reflection

Critical examination of AI's societal, ethical, and ecological implications forms the fourth pillar. This happens through both theoretical discussion and hands-on engagement that sparks reflection on emerging phenomena—from biases in training datasets to environmental costs of computation, from the evolving role of designers to questions of responsibility in shaping these technologies.

Formats like Fake or Real (co-developed by AI+Design Lab and Lisa Kern) challenge students to create and then identify convincing but fabricated social media content, revealing how easily misinformation can be produced and what indicators might suggest inauthenticity. The unStable Mirror installation makes AI biases tangible by highlighting how generative systems reflect and amplify existing societal patterns.

<swiper-container navigation="true" cssMode="true" grabCursor="true" loop="true"> <swiper-slide> <cite>Fake or Real: Fictional social media content. Students had to work out whether the content was fake or real.</cite> </swiper-slide> <swiper-slide> <cite>unStable Mirror exhibited at KIKK Festival in Namur. Photo by Quentin Chevrier. Project by Christopher Pietsch.</cite> </swiper-slide> </swiper-container>

Moreover, exploring technological futures with AI offers another strategy for critical reflection. The Futures Lenses format doesn't just support the exploration of novel concepts; it also exposes stereotypical future visions embedded in and reproduced by generative AI systems. Recognizing and questioning these mainstream narratives becomes a key skill as AI becomes more prevalent in design ideation.

Lessons from Three Years of Practice

Implementing these pillars across various formats—from compact workshops to extended project courses—has yielded valuable insights.

Sequential formats matter. Balancing technical depth and conceptual engagement in teaching proved challenging. This points to the need for consecutive formats that start with foundational technical knowledge, deepen it through hands-on engagement, then move to advanced conceptual exploration and critical reflection.

Different stages need different tools. Simple tools support early exploration and unexpected discoveries, while complex tools enable refinement and precise control as projects mature.

Co-creation requires openness and letting go. Teaching AI intuition requires shifting students' mindset: they must learn to embrace AI as a co-creator and agent of surprise rather than pursuing the "perfect prompt."

AI-generated concepts enable critical distance. While students often replicate stereotypical AI narratives in preliminary ideas, seeing them echoed in AI-generated outputs provides critical distance—AI's outputs become a mirror for examining mainstream metaphors and future narratives.

Experimental methods unlock new perspectives. Object-centered and performative approaches enabled students to break away from solution-oriented thinking, encouraging in-depth discussion of ethical dimensions and designers' responsibility in shaping these technologies.

Looking Forward

As AI technologies evolve, our teaching approaches must evolve with them. What remains constant is our commitment to AI Intuition—a holistic understanding that enables designers to engage meaningfully with AI, make informed decisions, and critically evaluate implications.

The tools, methods, and insights described represent three years of collaborative development by the AI+Design Lab team. We offer them as starting points for others navigating similar territory in design education, and we remain committed to sharing our approaches and learning from the community as the field continues to develop.

References:

[1] Flechtner, R., & Tost, J. (2025). Building AI intuition – Four educational pillars for teaching AI in design at AI+D Lab. In M. Grund, J. T. Wallenborn, L. Scherffig, & F. Jenett (Eds.), un/learn ai: Integrating AI in aesthetic practices (Vol. 3, pp. [30-42]). Hochschule Mainz. https://doi.org/10.25358/openscience-13021

[2] Flechtner, R., & Stankowski, A. (2023). AI Is Not a Wildcard: Challenges for Integrating AI into the Design Curriculum. Proceedings of the 5th Annual Symposium on HCI Education, 72–77. https://doi.org/10.1145/3587399.3587410

Transform 2025 @ HS Trier

Wed, 24 Sep 2025 00:00:00 GMT

Transform 2025: Conference on AI, Art, Design and Society

"Against the backdrop of evolving AI capabilities and mounting social and ecological challenges, we find ourselves at a critical juncture. Art and design practices, education, and society are undergoing profound transformations through artificial intelligence — yet the direction, impact, and meaning of these changes remain contested and uncertain.

Join us at TRANSFORM25 — where critical thinking meets active creation. This conference serves as the culmination of KITeGG, a four-year collaborative research project across five German universities exploring AI integration in art and design education.

We're bringing together diverse communities — design educators, creative professionals, researchers, policy-makers, developers, and advocates — to explore complex intersections of AI, art, design and society with nuance and critical awareness. Our aim is to foster connections among practitioners and craft practical solutions that address pressing ethical and practical questions in AI-driven design."

📅 September 24-27, 2025

📍 KINDLAB, Campus Gestaltung • Hochschule Trier

Teaching AI Intuition Talk @ HAWAII Online Conference

Mon, 22 Sep 2025 00:00:00 GMT

We are excited to announce that Rahel Flechtner and Jordi Tost will be presenting their talk, "Teaching AI Intuition - Four pillars of AI teaching in design at the AI+Design Lab", at the HAWAII Conference!

About the Talk:

After three years of teaching AI and design at the HfG Schwäbisch Gmünd, Rahel and Jordi will share insights into the teaching approaches developed at the AI+Design Lab to foster "AI intuition" — a foundational understanding of AI’s mechanisms, opportunities, and limitations in creative fields.

About the HAWAII Conference:

The HAWAII Conference highlights practical examples of AI in university teaching, showcasing innovative applications and the transformative potential of AI in higher education. The event is open to university lecturers, students, and all interested parties — both within and beyond the HAW in BW.

Don’t miss this opportunity to explore the future of AI in education and design! Sign up now!

White-boxing AI text comprehension to augment human text comprehension

Sun, 03 Aug 2025 00:00:00 GMT

PhD Progress Update:

Based on LLMs' internal metrics, we can calculate and visualize the reciprocal contextualizing and contextual integration of words, potentially increasing text comprehensibility.

Introduction:

To predict the word “you” after the text “Hello how are”, transformer-based Large Language Models (LLMs) process the text in several steps (Figure 1).

The text gets split into words or sub-words, and their meaning is represented through the position in a vector space (Tokenization and Embedding)
The meaning of the previous words gets added to the vector representation of the word “are” to contextualize it (Self-Attention-Mechanism)
The vector representation of the word “are” is further transformed (FFNN)
The word whose vector representation is most aligned with the contextualized “are” vector representation is chosen (in this case, “you”) [1]

Even though these LLMs are called “Black Boxes”, we can retrieve and interpret the Attention Scores of words to calculate and visualize the reciprocal contextualizing and the contextual integration of words.

Methods:

The Attention Scores of the words in a text determine how important words are contextually for the meaning of a specific word in the text. In other words, how much their meaning should be added to the specific word to contextualize it.

The Attention Heatmap (Figure 2) shows us how much the column word attends to the row word, so how much of the column word’s meaning is added to the row word’s meaning to contextualize it.

If we sum up the row attention scores, we get the total received meaning
When summing up the column attention, we get the total provided meaning

This enables us to calculate the normalized sum of the total provided and total received meaning, which represents the contextual integration of a word.

Visualization:

By visualizing the contextual integration of words through font (high = bold, middle = regular, and low = italic) and color (from high = blue to low = white), we enable the reader to directly see the core of the text, as well as words that might need further contextualization if they are not known to them.

By visualizing whether a word receives most of its contextual meaning from the left or right (white to blue gradient), contextual units emerge that could help the reader anticipate and understand the context (Figure 3).

To see where exactly a word receives its contextual meaning from, the reader can select it and understand its contextual integration (Figure 4)
The same is possible to see where exactly a word provides its meaning to (Figure 5), enabling the reader to understand the reciprocal contextualizing

Outlook:

The visualization of the reciprocal contextualizing and contextual integration of words in a text could give readers a visualized contextual understanding of the text that was hidden before in our very own “Black Box”, the human brain. This work explores how “white-boxing” AI can not only support its explainability and potential to simulate human text comprehension, but also be used to augment and empower humans.

The corresponding poster was presented at the Minds in the Making Workshop Workshop at the CogSci Conference 2025

Reclaiming the Process – Thoughts on the Role of Designers in the Age of AI

Fri, 11 Jul 2025 00:00:00 GMT

Reclaiming the Process – Thoughts on the Role of Designers in the Age of AI

This article shares some thoughts in preparation for the re shape 2025 conference. Over the past two years, the AI+Design Lab has brought together artists, designers, and researchers through two symposia—re shape 2023 and re shape 2025, exploring the evolving role of AI in creative practice.

At reshape 2023, we discussed how artificial intelligence would expand the practice of art and design, how it would change creativity, and what role designers should play in the development of AI and the responsibilities that come with it.

Since then, the AI landscape has evolved significantly. We are still riding the wave of hype around generative AI, but on another level of quality, speed, and efficiency. AI-based tools are spreading rapidly in areas such as interaction, product, and graphic design, art, literature, coding, and multiple research fields, optimizing processes and promising quick and easy results without requiring prior knowledge or skills.

This brings us back to a question that has haunted us in the last two years now:

What gets lost when design becomes automated?

If we continue certain developments uncritically, we risk losing the core of our competencies. If we delegate design decisions to tools and artificial agents and allow AI to take over our responsibilities as designers, we might forget our design standards — and even unlearn them over time. And what's more: Automated design decisions will reinforce stereotypes and may reduce design to the same old forms and functions. If we lose sight of our visions, what’s left to do, left to craft, left to imagine for us designers?

We need to rethink our responsibilities as designers in the development of artificial intelligence, including but not limited to power structures, environmental impacts, and the uncritical use of generated outputs. We need to reflect on our relationship with A systems, rethink our position as humans in a world that contains a lot more than humans. We need to rethink our role as designers in shaping a future with AI.

In the end, it is the creative process that defines us as designers. And to this end, we need to shift our focus back to this process:

The process of writing is thinking.

Experimenting, exploring, and playing are thinking.

Making, as a process, is also thinking.

We need this thinking as a mechanism for imagining and designing the new, for opening up alternative, desirable futures. And to think, we need the process. We should be asking how AI can enhance human intuition and creativity rather than constraining and diminishing it. We must research new paradigms for AI: new ways of knowing, of seeing, of thinking. New ways of designing.

What if we use AI not for providing efficiency and making everything smoother, but rather for creating creative friction? How can we use AI to support creative and critical thinking, to provoke, to surprise, to even create nonsense? Which other roles, which other metaphors can we use to open new paradigms for designing through and with AI?

Here is where research as well as education play an important role — to address, understand, develop, and embrace other perspectives towards different futures — not only at the intersection of design and AI — but for design as a discipline in general. We need to reshape design, and we need to reshape our role as designers. This is the challenge that we addressed at the reshape symposium in Schwäbisch Gmünd. A challenge that remains open and that needs further work, further research — and further process in the future.

The AI+Design Lab and the reshape symposia are part of the KITeGG project.

Beyond Tools: Exploring Human-Machine Relationships through the metaphor of 'Care'

Mon, 26 May 2025 00:00:00 GMT

How might the metaphor of 'caring for' rather than 'using' technology expand the design space for interactive systems? This question drove our recent workshop "Caring for Machines" at HfG Schwäbisch Gmünd, where we explored new ways of engaging with technological systems during our lab week.

Advances in sensor technology and artificial intelligence are gradually reshaping how we interact with technological systems. As these systems evolve from functional tools to more responsive counterparts, our relationship with technology moves beyond the traditional "technology as servant" metaphor. Contemporary technology design typically ranges between assisting technology that serves us as obedient tools, and demanding technology that competes for our attention through gamification and addictive interfaces. But what possibilities exist in the space between these poles? How does our emotional attachment to technologies change how we relate to them—and to ourselves? The workshop employed "caring for a machine" as a conceptual lens for reimagining human-machine interactions. We proposed the metaphor of "care," characterized by reciprocal relationships of giving and taking, as a means of exploring the possibilities that arise in the space between assistance and demand. Through a combination of conceptual exploration and hands-on engagement, we provided students with insight into sensor-based machine learning and encouraged them to experiment with creative and unconventional ways of interacting with technological systems. Working with simple and accessible tools, students trained and implemented their own machine learning models, turning abstract ideas into tangible, interactive prototypes. The four resulting projects embody diverse aspects of care, interaction, and technological companionship.

The following sections showcase the students’ final projects and outline the workshop methods and process that guided their development.

Final Projects

Social Spark

Philipp Maginot, Sebastian Weinert

Social Spark tackles one of the most human challenges: awkward silences in conversation. This interactive table gadget detects conversational lulls and responds with well-timed jokes or interesting facts designed to catalyze social engagement. This prototype explores the potential role of technology in fostering social dynamics through ambient interventions. <swiper-container navigation="true" cssMode="true" grabCursor="true" loop="true"> <swiper-slide> </swiper-slide> <swiper-slide> <cite></cite> </swiper-slide> </swiper-container>

Nature Pulse

Yingxun Li

Nature Pulse is an ambient companion that "feeds" on natural sounds throughout the day. The system processes natural audio signatures—bird calls, water sounds, wind patterns—translating these into chromatic responses that accumulate throughout the day. By evening, the accumulated light patterns serve as both functional illumination and reflective documentation of daily environmental encounters. The prototype explores how technology can create awareness of our connections with nature.

DanceBox

Carolin Schröpel, Taha Celik

DanceBox creates a dynamic dialogue between movement and music. Starting with a steady beat, this Arduino-powered device uses machine learning to recognize three distinct dance movements, each adding layers to the rhythm. The result is an evolving musical collaboration where the box becomes a responsive dance partner. This prototype promotes embodied interaction by exploring how machine learning enables the fluid and improvised collaboration of humans and machines.

Security Night Guard

Arthur Tomasi, Felix Schönherr

Security Night Guard investigates care through ritualized interaction with technology. Each evening, you hand over your keys to this vigilant companion, which watches over them through the night. In the morning, the "tired" guard releases the keys, creating a gentle wake-up call that signals the changing of the guard from machine to human. The prototype demonstrates a relationship with technology based on trust, responsibility, and mutual dependence rather than on mere utility.

The workshop illustrated the potential of "care" for designing alternative human-machine interactions. By combining conceptual exploration with technical implementation, participants were able to transform abstract concepts into functional, experiential prototypes. These prototypes point toward technological futures in which systems participate in care networks rather than merely serving human needs. They suggest a shift from instrumental technology relationships to more nuanced, emotionally engaged human-machine collaborations, fostering profound reflections on how we design and live with technology. The developed prototypes reveal interesting patterns that suggest alternative directions for understanding human-machine relationships. These patterns should be explored further in future courses or research. Reciprocal Agency emerges as a central theme. Each prototype establishes a bidirectional relationship in which humans and machines both meaningfully contribute to the interactive experience. Rather than responding to simple commands, these systems engage in an evolving, ongoing relational dynamic. Temporal Accumulation appears across multiple prototypes. Through sustained engagement, interactions with objects become more significant. This suggests that meaningful interactions are more likely to result from long-term engagement than from isolated transactions or single applications. Most prototypes are characterized by Environmental Integration, in which subtle, context-sensitive responses are favored over explicit interface interactions. Accordingly, care-oriented technologies may benefit from environmental integration rather than requiring targeted attention. Several prototypes integrate into everyday routines and practices. This type of Ritual Embedding suggests that care interactions may evolve from habitual rather than exceptional interactions and become interwoven into daily life.

Workshop Methods and Schedule

The workshop combined conceptual exploration and hands-on experimentation in a three-day format. On the first day, students were asked to bring an object of personal significance. Discussing these objects established a tangible foundation for understanding care relationships beyond human contexts. For further conceptual engagement, we used custom ideation cards that generated prompts in the format of "Caring for [an object] to care for [a purpose]". Combinations like "caring for a plant to care for your mental health" deliberately created an irritating yet inspiring starting point for short warm-up concept sprints. Students further engaged with foundational AI concepts through hands-on experimentation using Edge Impulse, training custom machine learning models to recognize gestures and movements. A second card-based tool paired sensors with actuators, facilitating systematic exploration of input-output modalities while maintaining connection to care-based concepts.

On the second day of the workshop, students worked on their final concepts based on care interactions. The implementation was initiated using the online machine learning service Edge Impulse. They trained a custom model to classify specific movements, gestures, or sensor patterns based on microphone or accelerometer data. The model was deployed on an Arduino Nano BLE Sense microcontroller, and the classifications were sent to the computer via emulated keyboard inputs. To support the technical experiments, the students were provided with boilerplate code for input and output modalities. As output, p5.js code was provided to either create custom interactions and visual outputs or to use a large language model (LLM) or text-to-speech model to generate text or speech outputs.

On the third day, the prototypes (presented above) were finalized and prepared for an exhibition.

re shape forum @ HfG Schwäbisch Gmünd

Fri, 16 May 2025 00:00:00 GMT

re shape Forum 2025 Recap: Rethinking AI in Art, Design, and Beyond

On May 15–16, 2025, the re shape Forum brought together a dynamic community of artists, designers, technologists, and researchers to explore AI’s evolving role in creativity, interaction, and society. Hosted by HfG Schwäbisch Gmünd, the forum questioned dominant AI narratives, critiqued extractive systems, and envisioned regenerative futures grounded in more-than-human perspectives. Through critical keynotes, boundary-pushing experiments, and cross-disciplinary dialogue, re shape 2025 carved out a vibrant space for reimagining the intersection of AI, art, and design.

Highlights From Every Speaker:

Elisa Giaccardi – Design After Extraction: AI and the More-than-Human Turn

A powerful opening keynote from the Full Professor of Design at Politecnico di Milano. Elisa challenged extractive design paradigms, urging a shift toward more-than-human futures. Her talk introduced regenerative AI practices rooted in multispecies collaboration, engaging plants, fungi, animals, and bacteria, to explore AI’s potential beyond control and exploitation.

Tobias Revell – Design and the Social Construction of Artificial Intelligence

Tobias, Design Futures Lead at Arup, dissected the narratives shaping AI through speculative design. His critical lens exposed how design reinforces power structures around AI, and how alternative imaginaries can subvert tech determinism, broadening futures beyond speed and consumption.

Maria Luce Lupetti – Drawing Windows Into Better Timelines

From Politecnico di Torino, Maria Luce brought a deeply reflective critique on our current sociotechnical crises. She questioned the designer’s role amid systemic failures, environmental collapse, and media saturation, calling for practices that resist dystopia and aim for better collective futures.

Niklas Muhs – How Do We Want to Interface With AI?

Niklas, the first designer at Mistral AI and former MIT researcher, explored the mechanics of AI interfaces. By moving away from agent-centric paradigms, he proposed novel modes of human-AI interaction that are transparent, co-creative, and cognitively aligned with user needs.

Eryk Salvaggio – Feel the Noise: Notes from an Adversarial AI Artist

Eryk, an artist-hacker-researcher hybrid from Harvard metaLAB and RIT, took an adversarial approach to AI. His provocative talk questioned collaboration as a default mode, presenting artworks that subvert, disrupt, and “remove the imagination” of what AI is, revealing its hidden logics and epistemologies.

Niko Klein – Lighting a Fire With Damp Wood

As a prototyping lead at Figma, Niko offered a grounded, hands-on perspective on design tooling in an AI age. He spoke about the changing role of prototyping when working with AI, emphasizing flexibility, process-over-output, and the importance of “designing with uncertainty.”

Sures Kumar – Reframing Creative Practice

From Google DeepMind, Sures showcased a speculative design engine fully powered by AI. Through this case study, he introduced frameworks for embedding AI meaningfully in creative workflows, stressing the importance of intentionality, context, and criticality when integrating generative tools into practice.

Iohanna Nicenboim – Designing-with AI

As postdoctoral researcher at TU Delft, Iohanna asked the audience to shift focus from human-centered to more-than-human design. Her talk spotlighted inclusive, ecological AI systems developed through participatory experimentation, challenging designers to listen to nonhuman voices in shaping future technologies.

Andreas Refsgaard – Beyond Prompting – Silly Interactive AI

Copenhagen-based artist and creative coder Andreas delivered a joyful and subversive talk on playful interaction design with AI. From meme machines to musical eye-trackers, he demonstrated how absurdity and humor can be powerful tools for critically engaging with digital systems, highlighting both the delight and danger of generative technologies.

A Forum That Lives Beyond the Event

re shape Forum 2025 was more than a series of talks, it was a collective act of imagination. It asked urgent questions: What stories are we telling about AI? Who gets to design it? What kinds of futures do we want to co-create?

<swiper-container navigation="true" cssMode="true" grabCursor="true" > <swiper-slide> <cite></cite> </swiper-slide> <swiper-slide> <cite></cite> </swiper-slide> <swiper-slide> <cite></cite> </swiper-slide> <swiper-slide> <cite></cite> </swiper-slide> <swiper-slide> <cite></cite> </swiper-slide> <swiper-slide> <cite></cite> </swiper-slide> </swiper-container>

The conversations sparked here will continue to resonate across design studios, labs, classrooms, and creative communities worldwide.

More info and speaker details: reshapeforum.hfg-gmuend.de

Welcome Address @ re shape forum 2025

Thu, 15 May 2025 00:00:00 GMT

A weird member of the team

Welcome Address to the re shape Forum by Maren Schmohl, Rector of Hochschule für Gestaltung Schwäbisch Gmünd

In her thought-provoking welcome address, Maren Schmohl set the stage for the re shape Forum 2025 by reflecting on the curious and sometimes unsettling relationship we have with AI today. From navigating the post-hype reality of generative tools to questioning human exceptionalism, she invites us to see AI not as a perfect machine or looming threat, but as a “weird member of the team,” brilliant, unpredictable, and increasingly present at our creative tables.

Schmohl draws parallels between AI and other humbling historical shifts, from Copernicus to Freud, reminding us that losing control may not be a failure, but a moment of intellectual honesty. In an age of accelerating complexity, she urges design schools and universities to embrace not answers, but uncertainty—to teach not just skills, but the humility and imagination needed to ask better questions.

This opening set the tone for two days of critical, creative, and courageous dialogue on how we reshape AI’s role in art, design, and society.

"Good Morning and welcome intelligent entitites:

Students, colleagues, guests and clever machines - welcome to the AI Reshape Forum. The second forum of the KITeGG project at HfG Schwäbisch Gmünd, and the 7th iteration of this format overall.

As most of you know, KITeGG is a 5 year long research project in which several German schools of design collaborate to explore ways of integrating AI in art and design education and curricula.

The forum brings together the KITeGG research team as well as esteemed international guests to explore AI’s evolving role in creativity, interaction, and society. The forum’s aim is to question dominant AI narratives, rethink design beyond human-centered approaches, and imagine regenerative futures where technology coexists with a more-than-human world. So we have an ambitious and exciting agenda and a lot to look forward to today and tomorrow.

My name is Maren Schmohl, I am the rector of HfG and I am very happy that I have been given the opportunity to greet you and share some of my thoughts while you finish your coffee and mentally gear up for the expert talks that will start in a couple of minutes.

I feel that we are at a curious moment right now concerning our experience with AI.

It seems that the first stunned than feverish excitement surrounding the release of ChatGPT in late 2022 may have subsided a little bit and the hype-cyle of public imagination may even have surpassed (a first) peak.

AI as a novelty has lost a little of its otherworldly glow and sparkle: we don’t perk up in excitement (or fear) every time a new development is announced. And we don’t quite expect AI to solve all of humanities problems within the next hours (or bring about the end of the world) - depending on whether you are more on the apocalyptic or the integrated end of the spectrum to use Umberto Eco’s famous dichotomy.

And so while we have gotten a little used to and habituated with AI I for one still have the feeling that we are standing in the middle of an 8-lane autobahn with objects zipping in all directions with 200 mph while we are trying to steer a bicycle: We haven’t been hit by a big truck yet, but it may happen any minute. A feeling of ominous uncertainty is still there.

Many have said that AI presents a further “insult” to humanity in that it throws into question some dearly held beliefs about ourselves. I agree. The three big insults that are referred to in this context are of course:

A cosmological one when Copernicus and Gallileo recognized that the earth is not the center of the universe but instead only one of a handful of planets that orbit the sun – not the other way around. We experienced a biological demotion when Darwin discovered that humans are not designed by God as the crown of creation but are descended from animals like all other species. And thirdly a psychological insult when Freud theorized that our rational mind, our free will, is in battle with or even determined by the subconscious and other psychic forces that we don’t have much control over.

We as modern people shrug this off, we laugh a little at the silly people of prior ages who were shocked by what to us seem like banalities. But now we are faced with an identity crisis of our own: a technological one, that once again throws the idea of what humans are for a loop and deals a serious blow to the notion – still deep seated – that we are different, special, better than other species or forces with which we share the earth. What is AI, what are we vis à vis AI, what is AI vis à vis to us: these questions face us as students, teachers/researchers but also as humans.

For a long time we have found it easy to set ourselves apart and in opposition to ever more sophisticated machines and technology. We defined areas of exclusive expertise or natural traits that surely only we as humans had: machines are dumb, we said, we are “intelligent”. Machines are cold, we are empathetic. Machines are just crunching numbers or auto complete sentences, we are creative. And finally: if machines are perfect – well then we are fallible, imperfect thus loveable, thus human. We have moved the goalposts to make sure we’re always winning.

If we play the game like this we’ve run out of space. These little islands of exceptionalism dissolve underneath our feet like sand and I don’t believe in any of them anymore. Machines, certainly AI, can beat us in all these categories. Is that a problem? Maybe not. But it is certainly profoundly unsettling.

New analogies arise. Dealing with AI is like being the driver of a self driving car. We simply let AI do the tedious work of navigating traffic while we have better things to do. Yet we still must learn how to drive, so that can take over should it be required.

Or: when using AI tools in our process we are like the conductor of an orchestra – someone who combines and ‘orchestrates’ the specific properties and talents of a host of tools and devices and decides what to use when and how.

In his typically understated way my colleague Prof. Hartmut Bohnacker recently said that he sees the challenge of working with AI as “finding interesting jobs for it”. This deceptively simple statement really sums up a lot of experience with AI. In this image AI has become what I would call “a weird member of the team”. Brilliant in one second, dumb as a brick in the next, but a team member non the less – it sits at the table with us, it’s not just a machine in the corner.

All of these images grapple with the idea of control. Who is in charge, who decides, who is the boss – or at least the author. They circle the question of what kind of skills are needed, what qualities do we need to develop in order to successfully work with AI. This of course is the perennial question for us as universities. We are talking about skills of application, skills of curation, certainly, but also good old fashioned expertise, standards, aesthetic and even moral values by which to judge whether an outcome is helpful, innovative and fulfils our demands.

A conductor produces wonderful music not because she can play every instrument herself but because she knows how the music should sound like. She knows exactly what she wants to achieve and bases her decisions of when to silence the violin, when to speed up the drums, on her expertise and vision.

This is a nice image – even if I don’t fully believe it.

I don’t believe that we are fully in control concerning AI. And I admit that I do feel a bit shocked and insulted by this notion when I let it hit me. But then: loss of control and supremacy does seem to be the defining headline of our age: We cannot control nature. Or the outcomes of capitalism. Or the predicted march of humankind towards progress and enlightenment. We are learning to be humble vis à vis forces that we thought we had conquered and well under control. AI maybe just a weird, imperfect member of the team – but so are we.

To be humble, to have been humbled, is not a comfortable place to be in, but a fitting one for us as teachers, students, explorers and experimenters, because it’s a good place to ask questions from.

So despite the feeling that during the last years we have gained firmer footing in handling and working with AI, and despite the enormous pressure that is put upon us to understand, control, put to use, make money with AI - let’s not presume that our questions have been answered or will be answered or that this is even the point. Let`s not set up another rigged competition. As always I say our job as schools and universities is not to produce certainty but to embrace uncertainty. Not to perpetuate the illusion of control but develop the skills to deal with question marks, openness and standing firm on an ever shifting ground of sand.

I want to thank Benedict Groß, Jordi Tost, Felix Sewing, Moritz Hartstang, Rahel Flechtner, Christopher Pietsch and the whole team here at HfG and our partners for organizing this wonderful event and I wish you all a great forum.

Thank you!"

Picture: Stefan Eigner

Tutorial-Serie: Bildgenerierung mit Invoke AI und SDXL

Wed, 14 May 2025 00:00:00 GMT

Tutorial-Serie: Bildgenerierung mit Invoke AI und SDXL

Diese Video-Serie des AI+D Lab an der HfG Schwäbisch Gmünd führt in die praktische Nutzung von Stable Diffusion XL (SDXL) in Invoke AI ein. Sie deckt alle grundlegenden Funktionen und Workflows ab – von der Installation und ersten Bildgenerierung bis zu Referenzbildern, Inpainting, Depth Maps und Guidance-Techniken.

Invoke AI ist eine Open-Source-Oberfläche für die Arbeit mit Diffusions-basierten Bildgenerierungsmodellen. In Kombination mit SDXL ermöglicht es detaillierte und kontrollierbare Bildausgaben aus Text-Prompts.

Die Tutorials richten sich an alle, die verstehen möchten, wie die Tools funktionieren und wie sie kreativ oder experimentell eingesetzt werden können.

Playlist

YouTube-Playlist öffnen

Einzelne Videos

Introduction – Welcome!

Video 2

Video 3

Video 4

Video 5

Video 6

Video 7

Video 8

Video 9

Video 10

Video 11

Video 12

Video 13

Video 14

Video 15

Video 16

Video 17

Autor: Robert Kremer

re shape forum coming up!

Wed, 14 May 2025 00:00:00 GMT

Join us on May 15–16 for the re shape Forum for Artificial Intelligence in Art and Design, where we’ll explore AI’s evolving role in creativity, interaction, and society. From playful experiments that challenge conventional AI use to critical discussions on its ethical and ecological impact, this event brings together visionary designers, artists, and researchers. We’ll question dominant AI narratives, rethink design beyond human-centered approaches, and imagine regenerative futures where technology coexists with the more-than-human world. Be part of the conversation as we reshape AI’s role in art, design, and beyond!

🌐 Find the full programme at: reshapeforum.hfg-gmuend.de

✅ The event is free and open to everybody. Thank you for registering online here

Edge Impulse CLI in Docker: A Cross-Platform Guide

Wed, 12 Mar 2025 00:00:00 GMT

Edge Impulse offers powerful tools for building machine learning models optimized for edge devices. Running the Edge Impulse CLI inside a Docker container can simplify setup and keep your host system clean. This guide explains how to configure and run the CLI on Windows and, if applicable, continue directly on Linux or Mac in Step 3.

Prerequisites

Ensure you have the following installed on your Windows machine:

Docker Desktop - Install Docker from the official site: Docker Desktop.
Windows Subsystem for Linux (WSL 2) - Required for Docker's Linux containers. Follow the Microsoft guide to install WSL 2.
Edge Impulse account - Sign up at Edge Impulse.

Step 1: Enable WSL 2 Backend in Docker

Open Docker Desktop.
Navigate to Settings > General.
Check Use the WSL 2 based engine.
Restart Docker Desktop for the change to take effect.

Step 2: Configure the COM Port for Docker in WSL (Windows Only)

To allow Docker in WSL to communicate with USB devices, install the USBIPD-WIN tool. Follow the official Microsoft guide here: Connecting USB devices in WSL.

Installing USBIPD-WIN

Download and run the latest usbipd-win .msi installer from the usbipd-win releases page.

Or install via Windows Package Manager:

winget install --interactive --exact dorssel.usbipd-win

Sharing and Attaching Your USB Device

Open PowerShell as an administrator and list connected USB devices:
```
usbipd list
```
Find your device’s busid and bind it:
```
usbipd bind --busid <busid>
```
In your WSL distribution, attach the device:
```
usbip attach --busid=<busid>
```
Confirm the device is attached:
```
lsusb
```

Step 3: Run Docker and Install the Edge Impulse CLI

Using the Microsoft devcontainer, launching Docker is straightforward. Execute the following command on any platform:

docker run -t -i --device=/dev/ttyACM0 mcr.microsoft.com/devcontainers/javascript-node /bin/bash

Inside the container, update packages and install required utilities:

apt-get update && apt-get install -y udev
npm install -g edge-impulse-cli

Step 4: Use the Data Forwarder with Your Edge Impulse Account

Now, you can run the edge-impulse-data-forwarder to collect sensor data and upload it to your Edge Impulse account.

Following these steps, you can successfully run Edge Impulse CLI inside a Docker container while ensuring access to USB devices. Happy coding!

Presenting unStable Mirror @ IPAI LIVING ROOM

Sat, 01 Feb 2025 00:00:00 GMT

Presenting the unStable Mirror in the IPAI LIVING ROOM. See you there!

The IPAI LIVING ROOM is a public visitor center on 1,000 square meters that offers the opportunity to get in touch with artificial intelligence (AI) and learn more about the Innovation Park for Artificial Intelligence (IPAI).

"The IPAI is set to become a flagship project for Baden-Württemberg, Germany and Europe, with sufficient appeal to compete on the world market and in global competition. This ambitious project is being supported with a funding volume of up to 50 million euros.

The aim of the IPAI is to create a world-class platform for applied artificial intelligence. The creation of several points of contact plays a central role in promoting transparency and understanding of artificial intelligence in society. A central element is the public visitor center, which provides the general public with access to artificial intelligence. In addition, exhibitions, events and innovative formats of information and knowledge transfer are promoted in order to make the diverse possibilities and applications of artificial intelligence tangible."

Progress Report: Learnings and experiences from the last two years

Mon, 16 Dec 2024 00:00:00 GMT

We look back on two years of teaching AI at HfG Schwäbisch Gmünd. Two years in which we worked together with the KITeGG project consortium on the goal of integrating AI into design education, and two years after we formulated our first thoughts and goals on this topic [1]. In spring 2023, we published an article on the subject. What has happened since then, what experiences and lessons have we learned, and how has our view of things at Schwäbisch Gmünd changed?

Looking back: Our goals for the integration of AI into design teaching at the HfG Schwäbisch Gmünd

The high technical and conceptual complexity and rapid development in AI technologies make it difficult for designers, especially design teachers, to keep track of what AI could offer their respective fields. It is even more challenging to keep pace with the technical implementation of these possibilities.

One question that has, therefore, accompanied us since the beginning of the project is to what extent we, as designers, need to immerse ourselves in the technical level of AI and how we can still focus on the core competencies of our discipline. Conceptual engagement with AI technologies without technical background knowledge leads to unrealistic fictional designs, and designers have found it challenging to develop interesting AI-based use cases. Therefore, We have thought about what we think design students need to know about AI to create engaging, meaningful, and feasible designs. Our goal for teaching was to provide students with an "AI intuition" - a technical intuition for the effects of the technology and its limitations. In this way, we want to offer students just as much technical discussion and basic knowledge as necessary to enable them to engage with the technology appropriately.

The AI+Design Lab

An overarching goal of the KITeGG project is integrating AI teaching into the design curriculum. One challenge we faced at HfG Schwäbisch Gmünd is the already full curriculum of the individual degree programs. This leaves hardly any room for adding new modules without shortening or replacing existing ones. At this point, we see a general problem that many universities will face when deciding what space AI should occupy in design teaching and the curriculum. Our approach to dealing with this problem is the idea of the AI+Design Lab as a cross-university location for AI expertise.

The AI+Design Lab is a contact point for students, who can find targeted advice and technical support for their projects here. It offers introductory courses for the development of technical "AI intuition." The intuition imparted enables students to make informed decisions about using AI technologies. They learn to assess the possibilities and limitations as well as the effort and costs involved in developing and implementing AI systems. In addition, they will be able to use existing technologies in creative and unconventional ways and use frameworks and tools to prototype and test the AI-based interactions they have designed. Students are sensitized to the possible consequences of AI technologies and learn to mitigate or avoid potential side effects through design decisions. In addition, precise language on AI topics is taught to prepare students for communication within interdisciplinary project teams and with stakeholders.

In addition to advising students, the lab provides a contact point for lecturers and integrates into existing project-based teaching. This allows for incorporating current and specialized AI knowledge on specific topics into teaching across the university.

Learnings and experiences from the last two years

We are taking stock after six months of setting up and one and a half years of operating the AI+Design Lab at HfG Schwäbisch Gmünd. What have we achieved so far? What has worked well? Where do we still see the potential for expansion?

Courses offered by the AI+Design Lab and how they are anchored in the curriculum

At the HfG in Schwäbisch Gmünd, we were able to firmly anchor three basic course formats in the curriculum to develop students' technical intuition concerning AI technologies. One course takes place in the Interaction Design Bachelor's program and two in the Strategic Design Master's program. In addition, we established further course formats during the workshop week, which takes place once a semester. These are one-week, more practical-experimental workshop formats in which students deal conceptually with AI technologies. The module is a compulsory elective module that can be chosen across all Bachelor's degree programs.

Due to the structural hurdles, we could not integrate any other course formats into the curriculum besides the individual project supervision offered. However, we still see great potential here regarding the technical and conceptual level of teaching and projects relating to AI technologies. For example, we believe it is essential to offer a compulsory AI basics course in all degree programs to be able to draw on basic knowledge in further courses or project supervision. In addition, we believe it is essential to offer elective project courses with a specific reference to AI to enable interested students to develop further in this area and to explore and help shape current technological developments in teaching. The results of these courses can, in turn, serve as "best-practice examples" that convey the current possibilities of the technology and inspire both students and lecturers, and drive forward the conceptual development and implementation of exciting projects.

To date, we have implemented our own projects in the lab for this purpose, which both keep us up to date with the latest developments and communicate the possibilities of the technology. For example, we created the Unstable Mirror project. This mixed-media installation draws attention to the options but also, in particular, to the limitations and inherent prejudices of generative AI. In addition, the project Transferscope was created, a tool with which the possibilities of image-generative diffusion models in connection with ControlNets (models that offer significantly better control over the generated content via edge or line detection, for example) can be explored in an explorative and low-threshold manner. The implemented tools also always have an educational level and can be used in teaching or the (university) public sector, for example, to make ethical implications visible and address them with the Unstable Mirror.

In addition, we at the AI+Design Lab are developing specific tools for teaching learning content. SandwichNet, for example, is a teaching tool that offers an interactive and playful approach to teaching the basics of machine learning, with a focus on neural networks. We use the tool to accompany primary teaching in various course formats, and it is freely available to students and other lecturers.

Connection to existing teaching formats and project courses

In addition to offering its courses, the AI+Design Lab is also keen to build on existing teaching formats. Colleagues take advantage of our offer to give short primer formats on AI in their courses. These offerings allow them to incorporate current and specialized AI knowledge on their subject area into their teaching, even if they do not (yet) have the necessary expertise.

Our experience over the last year and a half has shown that more AI input for lecturers would be beneficial. This would anchor knowledge sustainably at the university even after the project ends and prevent the unconscious promotion of technically unrealistic concepts. We are, therefore, planning special input formats on AI for lecturers in the winter semester of 2024/25.

The lab's support services were able to integrate with teaching across all degree programs and have already become well-established at the university. For example, students from other lecturers' project courses come to us for advice when dealing with an AI-relevant topic or are sent to us specifically by the lecturers.

Our experience has shown that it is crucial for students to find out about our offer promptly at the start of the project and come to us for advice. Technically unrealistic ideas and perceptions of the possibilities of AI technologies are often already anchored at an advanced stage of the concept phase, and students are frequently frustrated when they have to discard or heavily adapt their concept. This sometimes leads to the rejection of our offer and the insistence on fictitious concepts.

The university has also established a system for supervising Bachelor's and Master's theses related to AI by those involved in the project.

Visibility and accessibility of the lab

Despite the good influx of requests for supervision and support and the integration into the curriculum, we have noticed in recent years that the Lab's cross-university "satellite role" still makes it difficult to generate attention and awareness of the existence of the AI+Design Lab among students across all degree programs, especially among new students. For this reason, the Lab regularly offers event formats that draw attention to the Lab and involve students. For example, a regular "Prompt Battle" format was established, which is now organized and further developed by students.

The Lab also provides information about all the latest news, events, and achievements on this website.

In 2023, we started setting up a digital lab to improve accessibility to its resources. This enables providing resources such as link collections on specific topics, tutorials, best-practice results, or documentation that students and teachers can access quickly and asynchronously. The digital lab is also intended to help establish AI knowledge at the university in the long term.

Looking back on the last two years, we see great potential in the lab's cross-course position as a place of specialization and will continue to expand and test this concept in the coming year.

Illustration by Mark Jonathan Freyer

Ressources

[1] Rahel Flechtner and Aeneas Stankowski. 2023. AI Is Not a Wildcard: Challenges for Integrating AI into the Design Curriculum. In EduCHI 2023: 5th Annual Symposium on HCI Education (EduCHI ’23), April 28, 2023, Hamburg, Germany. ACM, New York, NY, USA. DOI: 10.1145/3587399.3587410.

KI & Zukünfte: Imaginaries, artificial irritations and machine manifestos @ Correlations Forum, HfG Offenbach

Thu, 12 Dec 2024 00:00:00 GMT

How Does AI Envision the Not-Yet-Known?

While established narratives often present the future as singular, futures—plural—are diverse, shaped by different perspectives and worldviews. Together with Johanna T. Wallenborn, I will examine dominant visions of AI and discuss how AI can serve as a partner in imagining the unknown.

In this workshop, we will experiment with strategies from Critical Design and methods from Critical Futures Studies to deconstruct and reconstruct AI futures. Through unusual and absurd AI metaphors, we will decontextualize, question and redefine our vision of AI. The focus will then shift to co-writing fictional manifestos using language models.

The workshop combines critical reflection and creative practice, opening new perspectives on the relationship between humans, machines, and desirable futures.

The talk and workshop is part of the regular event series „KI Abend“ and will take place on Thursday 12th December 2024, 19h, at the AI Lab of the University of Arts and Design Offenbach in the context of the Correlations — Forum for AI in Art and Design.

Program and further information:

correlationsforum.de

hfg-offenbach.de

KITeGG — gestaltung.ai

Correlations Forum 2024 @ HfG Offenbach

Wed, 11 Dec 2024 00:00:00 GMT

From 11. to 13. December 2024, we will present Transferscope and give a talk and workshop at the Correlations Forum in Offenbach. See you there!

"Correlations Forum aims to explore the integration of AI and robotics into artistic practices and discussions, providing a platform for innovative artworks and stimulating exchanges of ideas."

Transferscope Setup @ Correlations Exhibition

Inaugural Lecture Visiting Professorship Creative AI — Jordi Tost

Wed, 20 Nov 2024 00:00:00 GMT

I, Jordi Tost, am thrilled to start my position as a Visiting Professor for Creative AI at the University of Applied Sciences Schwäbisch Gmünd this October, where I will be sharing the role with Rahel Flechtner. As part of the AI+D Lab research group and the KITeGG project, I’m eager to explore the intersection of AI and design, particularly design futuring. In my inaugural lecture on Wednesday, 20th November 2024 at 18:00, I will introduce myself, my ideas, and the research and teaching topics I plan to cover over the next year.

Futures Lenses. Co-speculating Things with AI

Fri, 15 Nov 2024 00:00:00 GMT

The way we imagine possible futures are diverse and shaped by different perspectives and worldviews. Each offers its own possibilities and influences how we design things. The workshop Futures Lenses — Co-speculating Things with AI, developed and held by Jordi Tost, aimed to explore the potential of looking into different speculative futures through the lens of AI as a design method, using generative AI as a co-designer and critical counterpart for ideation.

The workshop was held over five days at HfG Schwäbisch Gmünd as part of the Lab Week in autumn 2024. It was part of the joint research project KITeGG and involved students from HfG Schwäbisch Gmünd and HS Mainz.

The workshop took a deliberately hands-on, playful approach to support imagination through the engagement with speculative scenarios and the co-creation of fictional designs in collaboration with large language models and image generators. Instead of creating broad, abstract future images, participants focused on redefining everyday objects as entry points into imagining alternative worlds and unconventional design possibilities. Yet, accessing perspectives beyond our own habitual ways of thinking remains a challenge. What if the very "failures" of AI—its hallucinations and misunderstandings—could be deliberately provoked to foster imagination?

Creative Friction through AI Hallucination

The core methodological approach was counterintuitive: rather than seeking controlled AI outputs from previously defined concepts, the main idea was to deliberately apply constraints, ambiguity, or exaggeration through prompts, as a strategy to provoke misunderstandings, glitches, hallucinations, and unexpected results in the outcoming AI-generated images—what we might call "creative accidents."

This iterative process combined strategies from critical design practice—anti-solutionism, irony, and absurdity—with low-fidelity prototyping, and simple, no-code AI tools. These tools included simple text-to-image generators, Transferscope, and the Futures Lens—a custom tool developed at the AI+D Lab for the workshop format based on Stable Diffusion. For example, the Futures Lens transforms input images based on text prompts.

<swiper-container navigation="true" cssMode="true" grabCursor="true" loop="true"> <swiper-slide> <cite>Futures Lenses Workflow</cite> </swiper-slide> </swiper-container>

In this process, the model hallucinates details, changing or removing elements of the input image, or adding new ones. By using different prompts, one can reinterpret and defamiliarize the same image or object in multiple ways, opening diverse possibilities for ideation and imagination.

<swiper-container navigation="true" cssMode="true" grabCursor="true" loop="true"> <swiper-slide> <cite>Futures Lenses concept: an input image of a tea pot (left) gets transformed based on the prompts “Cyberpunk” and “Toy land” (right)</cite> </swiper-slide> </swiper-container>

Futures as Constraints

Students, who worked in groups, were assigned an unconventional future as a main constraint—such as "calm," "inefficient," "selfish," or "spooky"—and were challenged to make sense of these futures, no matter how uncanny, by reflecting on everyday objects and their interactions. Following this strategy in a “quantity-over-quality” approach, students transformed input images and analyzed the outcomes, interpreting strange elements and visual glitches as sources of critical and creative thinking. Through multiple iterations, students defamiliarized everyday objects, exposed paradoxes and biases embedded in AI systems, and discovered new design principles and possibilities.

Catalogues of Fantastic Things

As a result of the workshop, each student group produced a "Catalogue of Fantastic Things"—a design fiction artifact that serves as both documentation and speculation. Each catalogue visualizes diverse fictional concepts and reflects on alternative futures through their everyday objects. They do not aim to function as product proposals but as provocations, inviting us to question our assumptions about efficiency, innovation, solutionism and the purpose of design itself.

Calm Future

Anja Gutmann, Lea Haferbier, Carolin Kaltwasser

Welcome to the calm future. In this world, it is important to embrace presence and nurture well-being. Of course, we have to deal with the paradox of technology: how to create tools that promote slowness in a fast-paced world. Luckily, we discovered the principle of intentional slowness—designing objects that encourage mindfulness rather than speed.

This project explores a world where homes become living, breathing entities that rely on our care and attention, and where products serve not only functional needs but also promote well-being and happiness. Technology in this future moves beyond efficiency and productivity to cultivate emotional intelligence and empathy.

The catalogue "Calm Design: Future Objects for a Quiet World" showcases a curated selection of objects that reimagine relationships with time, daily routines, and self-care. The Mindful Alarm, a clock that only rings if you tend to its plant—water it, give it light, and it will gently wake you each morning. The Calmness Backpack serves as a minimalist carrier with soft, natural textures and hidden pockets for mindfulness tools, symbolizing tranquility you can keep close. Objects that remind us to take care of ourselves, inviting us to move in calm and stay in quiet, where time unfolds in a softer way through steady steps in a peaceful place.

<swiper-container navigation="true" cssMode="true" grabCursor="true" loop="true"> <swiper-slide> <cite>Calm Design — Future Objects for a Quiet World</cite> </swiper-slide> <swiper-slide> <cite>Calm Design — Future Objects for a Quiet World</cite> </swiper-slide> <swiper-slide> <cite>Calm Design — Future Objects for a Quiet World</cite> </swiper-slide> <swiper-slide> <cite>Calm Design — Future Objects for a Quiet World</cite> </swiper-slide> <swiper-slide> <cite>Calm Design — Future Objects for a Quiet World</cite> </swiper-slide> </swiper-container>

Inefficient Future

Cosima Rieger, Jonas Wienberg, Tim Bluthardt

Welcome to the inefficient future. In this world, it is important to embrace imperfection and take time for things that truly matter. Of course, we have to deal with the paradox of efficiency and instant gratification that modern society pushes on us. Luckily, we discovered a design principle based on purposeful slowness and mindful imperfection.

Presented as an "Instruction Manual for an Inefficient Future," the project takes the form of a LEGO guide created in fictional cooperation with LEGO and Dieter Rams. It features gadgets that celebrate the art of waiting, the beauty of error, and the charm of manual labor—devices that deliberately take longer to process simple tasks.

The work steps outside efficiency-driven norms of everyday life, consciously practicing inefficiency as a design choice to challenge the meritocratic push for streamlined results. Like building with LEGO, it invites us to assemble ideas for a future where slowing down, playing, and questioning purpose become essential building blocks, bringing richness to design through complexity.

<swiper-container navigation="true" cssMode="true" grabCursor="true" loop="true"> <swiper-slide> <cite>Instruction Manual for an Inefficient Future</cite> </swiper-slide> <swiper-slide> <cite>Instruction Manual for an Inefficient Future</cite> </swiper-slide> <swiper-slide> <cite>Instruction Manual for an Inefficient Future</cite> </swiper-slide> <swiper-slide> <cite>Instruction Manual for an Inefficient Future</cite> </swiper-slide> <swiper-slide> <cite>Instruction Manual for an Inefficient Future</cite> </swiper-slide> </swiper-container>

Selfish Future

Georgia Ioannidou, María José Merino Zapata, Lucía Navarro Bloise

Welcome to the selfish future. In this world, it is important to prioritize individualism and personal satisfaction above the needs of others. Of course, we have to deal with the necessity of sharing spaces and resources, but we strive to avoid this as much as possible to prevent tolerating others and internal conflicts. Luckily, we discovered the principle of personalized separation, allowing each person to maintain their autonomy even in shared environments.

The project manifests as commercial brochures featuring products designed for power individualism: a multi-compartment pot ideal for shared flats where no one has to cook together, a dual-spout teapot to host guests without compromising your tastes, and a single-seating table for dining solo. This darkly satirical work exposes contemporary trends toward hyper-personalization while raising provocative questions about isolation, community, and social connection in an age of extreme individualism.

<swiper-container navigation="true" cssMode="true" grabCursor="true" loop="true"> <swiper-slide> <cite>Selfish Future" product brochures</cite> </swiper-slide> <swiper-slide> <cite>Selfish Future" product brochures</cite> </swiper-slide> <swiper-slide> <cite>Selfish Future" product brochures</cite> </swiper-slide> <swiper-slide> <cite>Selfish Future" product brochures</cite> </swiper-slide> </swiper-container>

Spooky Future

Joshua Quere, Yannick Schwab

Welcome to the spooky future. In a world driven by hyper-efficiency, we risk losing the sense of presence and mystery that once made life meaningful. Our surroundings have become flaw- less and automated, but this perfection comes at the expense of something deeper—our connection to the world around us. To reclaim this, we embrace design that intentionally introduces subtle discomfort, slo- wing down our experiences and prompting reflec- tion. By incorporating purposeful flaws, our spaces and objects encourage us to pause and engage more meaningfully with each moment. The catalogue "Modern Living" features devices that subtly disrupt instant gratification, reminding us to reconnect with our surroundings and ourselves. Embrace the quiet unease—it is within the unknown that we truly rediscover life.

<swiper-container navigation="true" cssMode="true" grabCursor="true" loop="true"> <swiper-slide> <cite>"Modern Living" catalogue</cite> </swiper-slide> <swiper-slide> <cite>"Modern Living" catalogue</cite> </swiper-slide> <swiper-slide> <cite>"Modern Living" catalogue</cite> </swiper-slide>

</swiper-container>

Towards playful ways of engaging with AI

Rather than offering theoretical deep dives or technical training, the workshop focused on supporting practical, conceptual and critical engagement with generative AI, providing design students with hands-on, playful ways of working with the technology for ideation, co-creation and co-speculation.

By focusing on critical prompting and embracing the unexpected, students discovered how AI hallucination, typically viewed as a technical problem to be solved, can be reframed as a creative and critical AI + Design strategy. Generative AI, when used speculatively and critically as a disruptive counterpart, serves a dual function: it can help us expose embedded biases and assumptions in both the technology and our own thinking, unravel stereotypes, defamiliarize the familiar and discover unexpected possibilities for design that might not emerge through conventional ideation.

The workshop suggests that engaging critically with AI does not have to be abstract—it can begin with the concrete, the playful, and the everyday: one object at a time.

Presenting Projects @ KIKK Festival

Thu, 24 Oct 2024 00:00:00 GMT

From 24. to 27. October 2024, we will present the Transferscope and the Unstable Mirror at the KIKK festival in Namur. See you there!

Photo by Quentin Chevrier

"Founded in Namur in 2011, the non-profit organization KIKK is committed to building bridges between the arts, sciences, culture and technology. KIKK transforms and densifies space. Free-spirited and daring, KIKK stands between technological innovation and creativity, refreshing the foundations of a reinvented economy.

Every year, KIKK organizes the eponymous international festival in Namur, which showcases digital practices in the broadest sense. For its 11th edition in 2022, the event attracted over 25,000 visitors (professionals and the curious alike) from nearly 40 nationalities."

KI und Kuchen

Thu, 24 Oct 2024 00:00:00 GMT

On October 30, 2024, it wasn't the graphics cards that were running hot, but the oven for a break. After about 2.5 years of research, it was time to present the new findings, projects, publications, and teaching tools to teachers and interested parties in a relaxed setting. After a short presentation on the current status of the project, the lab was transformed into an AI & cake bazaar with four themed islands: The first themed island presented import formats and ready-made course modules. These included teaching and learning tools that can be integrated into existing course formats. The second island focused on interactive knowledge exchange and transfer in the rapidly changing field of AI. A prototype learning environment (virtual AI+D Lab) was presented, which is intended to serve as a collection point for new tools and teaching materials. The third island presented the existing infrastructure: the KITeGG Cluster Server and the Teaching and Learning Platform, together with the services developed. The fourth island offered insights into current research papers and projects.

The concluding discussion round provided an opportunity for an intensive exchange of experiences and future developments. The feedback was particularly valuable for the further orientation of the research project. We were delighted with the great interest and lively participation, for which the AI+D Lab team would like to thanks!

<swiper-container navigation="true" cssMode="true" grabCursor="true" > <swiper-slide> </swiper-slide> <swiper-slide> </swiper-slide> <swiper-slide> </swiper-slide> <swiper-slide> </swiper-slide> </swiper-container>

Presenting Transferscope @ IEEE VISAP

Sun, 13 Oct 2024 00:00:00 GMT

From 13. to 18. October 2024, we will present the Transferscope online at the the VIS Arts Program (VISAP). See you there!

"The VIS Arts Program (VISAP) is a mini-conference and exhibition where visualization researchers, designers, and artists come together to showcase and discuss works at the intersection of data visualization, art and design.

VISAP'24 is an occasion for international and local audiences to share concepts, exchange ideas, and investigate a multitude of transdisciplinary connections. It is the biggest associated event in the IEEE VIS conference program, connecting media arts practitioners, data scientists, graphic designers, and theoreticians, to foster discussion and collaboration between fields of research and practice."

Presenting Projects @ MuC Conference

Sun, 01 Sep 2024 00:00:00 GMT

From 01. to 04. September 2024, we are presenting the Transferscope and Sandwichnet at the MuC conference in Karlsruhe. See you there!

Proceedings

Transferscope (10.18420/muc2024-mci-demo-248)
Sandwichnet (10.18420/muc2024-mci-demo-322)

"The “Mensch und Computer” (MuC) conference, launched in 2001, is the largest conference series on human-computer interaction in Europe. MuC offers participants from science and industry a platform for contributions and discussions on innovative forms of interaction between people and digital technology, human-centered development methods, interactive applications, and other topics in the field between users, teams, and communities on the one hand and information and communication technologies used on the other hand. The aim of the conference is to discuss innovative research results, promote the exchange of information between academia and practitioners, stimulate research activities and training, and sensitize science, practice, and the public to the relevance of human- and task-oriented technology design."

Amber | AI-Assisted Bark Beetle Detection

Fri, 12 Jul 2024 00:00:00 GMT

Bachelor Thesis by Emma Rahe and Levin Budihardjo (in BA Product Design)

Forests play an essential ecological role, but rising temperatures and droughts pose significant risks, including increased bark beetle activity. Amber is an AI-powered monitoring device that records natural sounds year-round and detects bark beetle infestations before they become externally visible. This gives foresters more time to act and prevent the spread. Amber uses seasonal indicators: in summer, it captures beetle feeding sounds in high-risk areas, while in winter, it monitors woodpecker feeding behavior to detect dormant beetles. The AI processes these recordings by converting them into spectrograms, applying image recognition techniques to analyze the visual representation of the audio data. A convolutional neural network scans the spectrogram for patterns characteristic of the target sound. Amber demonstrates how AI can bring significant benefits in areas with seemingly low risk. The AI does not make decisions but highlights natural indicators for foresters. As a flexible, futureproof concept, Amber’s algorithm can be adapted to new contexts, such as biodiversity monitoring or pest detection in fruit plantations.

Summer Semester 2024 Supervised by Prof. Matthias Held and Aeneas Stankowski Supported by FVA Forstliche Versuchs- und Forschungsanstalt Baden-Württemberg

Link

https://portfolio.hfg-gmuend.de/s-2424/projekte/amber

KITeGG Summer School @ Köln International School of Design

Wed, 12 Jun 2024 00:00:00 GMT

Join us at the 5th KITeGG summer school at the Köln International School of Design. Here is some preliminary information on our upcoming conference Hidden Layers. More information will be available in late May at hiddenlayers.de

First puplic appearance: you can try out Transferscope at the exhibition and experience first-hand how easy it is to create your own AI assisted artworks.

Making (Non-)Sense @ EduCHi24

Fri, 07 Jun 2024 00:00:00 GMT

{frontmatter.title}

We are delighted to have shared our insights and experiences on AI education for design students at the EduCHI symposium 2024 in New York City. Our paper, ‘Making (Non-)Sense—A Playful and Explorative Approach to Teaching AI Intuition for the Design of Sensor-Based Interactions,’ co-authored by Rahel Flechtner from HfG Schwäbisch Gmünd and Jakob Kilian from the Köln International School of Design (KISD), contributed to the ongoing challenge of integrating AI into the design curriculum effectively. This workshop, as part of the joint research project KITeGG, aimed to equip design students with the necessary intuition for AI technologies to develop sensor-based AI-based interactions. EduCHI 2024, held from June 5-7 at the Pratt Institute's School of Information in New York City, brought together an international community of scholars, practitioners, and researchers to discuss and shape the future of Human-Computer Interaction (HCI) education. Organized in cooperation with the ACM Special Interest Group on Computer-Human Interaction (SIGCHI), EduCHI provides a platform for sharing HCI curricula, teaching materials, and pedagogical experiences. The symposium is designed to foster the exchange of innovative pedagogical approaches and address the needs of HCI educators from around the world, reflecting the latest trends and innovations in the field. You can find the original paper here or download an author's version by clicking the yellow button on the top left of this article. Additionally, you can access the course materials, published as open source/hardware, here. A video documentation of the workshop is available on Vimeo. For a more detailed discussion on the workshop, please refer to this article.

Transferscope — Synthesized Reality: Sample anything. Transform everything

Tue, 28 May 2024 00:00:00 GMT

import { Image } from 'astro:assets';

import image1 from './images/transferscope-illustration2.png'; import image2 from './images/transferscope-steps2.jpg'; import image3 from './images/transferscope-inner.jpg';

Transferscope is a working prototype that merges human creativity with artificial intelligence, resulting in a handheld device that transforms our physical world through AI. The device allows users to capture an object or concept with a simple click and seamlessly blend it onto any scene, creating new and imaginative realities.

With its single-button interface and clear screen feedback, users can swiftly capture and transform images, allowing them to focus on their creativity and experimentation. This simple design ensures that the technology seamlessly supports the process without causing any distractions or complications. The device is lightweight and portable, making it easy to carry and use in various environments.

A Maker project

Transferscope's handheld device is powered by a Raspberry Pi Zero 2, encased in a custom 3D-printed housing. It features a 720x720 pixels resolution screen (HyperPixel) and integrated Li-Ion batteries (2x 18650). The device runs on Debian with a custom Python script for image capture and processing (PyGame, OpenCV, Asyncio, libcamera). It handles an edge detection algorithm directly on the device, which it sends with additional data to a server (RTX 4080, Intel i7-12700K). There the interpretation of the image, and generation of the output is processed.

The server runs a optimized Pipeline that links two central functionalities: Stable Diffusion for image generation and an image interpreter called Kosmos 2 by Microsoft. The image is first interpreted, then sent to an IPAdapter which synthesizes contextual and visual information into an entirely new image. The module used to steer the image generation towards the input geometry is a ControlNet. This setup allows Transferscope to transform captured images into new textures and patterns and apply them to a second scene almost instantly ({'<'}1s), creating a seamless experience for the user.

AI as a collaborator

Transferscope is more than just a technological demonstration. It offers a new way to view and interact with the world. By allowing users to overlay textures and patterns onto various objects, it encourages a rethinking of everyday environments and objects. This capability can be used for artistic expression, educational purposes, or simply to explore new perspectives. My vision for Transferscope is to capture the essence of today's generative image AI, utilizing open-source tools and models, marking the current state of the technology. It showcases the shift from genuine photography to synthetic image creation. Situating the project at the boundary between reality and AI generated content that is infiltrating our daily lives.

While the tech behind the device is complex, the user experience is simple and intuitive. This balance between advanced technology and user-friendly design is what makes it fun and engaging to use. It’s not just technology deciding the result: the user's guidance directs the transformation. This is a example of a collaboration between human and machine that opens up new possibilities for creativity and exploration.

Process

<swiper-container navigation="true" cssMode="true" grabCursor="true" > <swiper-slide> <cite>3D printed backplate with mesh structure for better heat dissipation</cite> </swiper-slide> <swiper-slide> <cite>The first iteration consisted of two devices: one for sampling and one for transfering</cite> </swiper-slide> <swiper-slide> <cite>Finding shape and structure for Transferscope</cite> </swiper-slide> <swiper-slide> <cite>Transferscope in shipping case and hardware components</cite> </swiper-slide> <swiper-slide> <cite>Christopher turning into a cup</cite> </swiper-slide> </swiper-container>

Creating transferscope was a iterative process, exploring the technology and form of the device. Multiple prototypes were 3D printed, tested and refined. The project is related to the unStable Mirror, a previous project that explored similar themes. It evolved from two devices to one, simplifying the user experience and enhancing the tool's character. While working on Transferscope new advancements in AI and image synthesis were made, making it a living project that evolves with the technology.

Credits

Concept, Idea and Production: Christopher Pietsch
Extended Team: Benedikt Groß, Aeneas Stankowski, Felix Sewing, Rahel Flechtner
Illustration: Stamatia Galanis
Generative Voice: ElevenLabs

Shaping Human-AI Relationships – Workshop

Fri, 17 May 2024 00:00:00 GMT

{frontmatter.title}

As we design human-machine interactions, we inevitably shape expectations and relationships with technical systems. This becomes especially important as intelligent, AI-based systems play an increasing role in everyday life. The workshop "Shaping Human-AI Relationships – Exploring the Design Space through Performative Methods" aimed to explore how these human-technology relationships can be designed in a conscious and reflective way. In this workshop, held over five days at HfG Schwäbisch Gmünd as part of the international seminar week in summer 2024, we explored the design space for human-AI relationships through a range of experimental, performative methods, including role-playing and improvisation. The aim was to examine the potential of these methods for designing human-AI interactions.

On the first day, students were introduced to the technical fundamentals of AI-technologies through the “Acting out AI Systems” roleplay method in which students embodied different components of a sensor-based AI system [1]. The second day, we shifted from a user-centered to an object-centered perspective. Using again roleplay but also methods such as Thing Ethnography [2], Object Persona [3], and Techno-Mimesis [4], students examined AI systems from the perspective of the technology itself, analyzing the characteristics and behaviors of systems such as a fitness tracker, a virtual companion, or a virtual pet.

<swiper-container navigation="true" cssMode="true" grabCursor="true" loop="true"> <swiper-slide>

![Alt Text.](./Thing-Centered_Virtual-Friend-2.jpg)
<cite>Students roleplaying a virtual companion system</cite>

</swiper-slide> <swiper-slide> <cite>Students roleplaying a fitness tracking system</cite> </swiper-slide> </swiper-container>

On day three, students explored the influence of system characteristics on human-machine relationships. A Slot Machine was used to generate randomized scenarios, mixing user needs with system behaviors, allowing students to role-play these combinations and explore the ethical implications of different interactions.

At the end of the workshop students applied the gained knowledge to consciously design the characteristics of an AI system. The freedom to choose between critical or solution-oriented design projects led to a variety of critical installations, which were presented in an exhibition on the final day.

The workshop took a hands-on approach with a focus on practical and conceptual engagement with the technology rather than theoretical deep dives. Especially courses with time constraints face the challenge of creating a low-threshold introduction to the informed design of AI technologies, without being held back by the technical complexity, while also not ignoring it. In this workshop, we therefore addressed the technical aspects in an abstract way using experimental and performative methods.

References:

[1] Rahel Flechtner and Jakob Kilian. 2024. Making (Non-)Sense—A Playful and Explorative Approach to Teaching AI Intuition for the Design of Sensor-Based Interactions. In Proceedings of the 6th Annual Symposium on HCI Education (EduCHI '24). Association for Computing Machinery, New York, NY, USA, Article 9, 1–9. https://doi.org/10.1145/3658619.3658643

[2] Elisa Giaccardi, Nazli Cila, Chris Speed, and Melissa Caldwell. 2016. Thing Ethnography: Doing Design Research with Non-Humans. In Proceedings of the 2016 ACM Conference on Designing Interactive Systems (DIS '16). Association for Computing Machinery, New York, NY, USA, 377–387. https://doi.org/10.1145/2901790.2901905

[3] Nazli Cila, Elisa Giaccardi, Fionn Tynan-O’Mahony, Chris Speed, & Melissa Caldwell. 2015. Thing-Centered Narratives: A study of object personas. In Proceedings of the 3rd seminar international research network for design anthropology (pp. 1-17)

[4] Judith Dörrenbächer, Diana Löffler, and Marc Hassenzahl. 2020. Becoming a Robot - Overcoming Anthropomorphism with Techno-Mimesis. In Proceedings of the 2020 CHI Conference on Human Factors in Computing Systems (CHI '20). Association for Computing Machinery, New York, NY, USA, 1–12. https://doi.org/10.1145/3313831.3376507

The “effect method” and representational models

Mon, 15 Apr 2024 00:00:00 GMT

Bachelor Thesis by Moritz Hartstang (in BA Product Design)

Learning practical skills in complex problem-solving, working with artificial intelligence, and human creativity is increasingly important. At the same time, these areas are increasingly important as practical parts of teaching. However, there are significant gaps in our theoretical knowledge of them. Complex problems, artificial intelligence, and human creativity are terming whose indefinability is often part of their current definition. The "lack of transparency" as part of complex problems, the "black box" as part of artificial intelligence, and the "illumination" as part of human creativity. So, how can we teach practical skills in these areas and use them in teaching, even though we lack theoretical knowledge on how complex problems, artificial intelligence, and human creativity function?

Building on different theoretical knowledge and testing, we can create representational models to describe practical skills in an area without the need for a complete theoretical understanding of it. As a result, these skills can already be learned and used for teaching, even though some parts of the areas are still seen as theoretically indefinable. In addition, representational models can be the starting point for growing theoretical knowledge by inferring theory from practice, thus counteracting indefinability. Furthermore, it enables students to learn about theory while engaging in practical skills.

The “effect method,” developed in the Bachelor Thesis by Moritz Hartstang, is a representational model that describes a practical skill of solving complex problems creatively. Students can identify clear problems and solution approaches within “intransparent” complex systems through structured research questions for search engines based on large language models. The method empowers students to identify and understand relevant problems and solution approaches by researching defined interrelations between subjects (terms considered as single entities) and systems (terms considered as multipart entities). The interrelations are defined by effects to be prevented (problem) or achieved (solution). Based on this, the effect method enables the development of system-compatible solution concepts. As a result, each student’s creativity, knowledge, and best abilities can contribute to the collaborative development of relevant solutions.

At the same time, they learn the necessary knowledge for the subsequent collaborative development of creative solution concepts as well as practical skills in working with artificial intelligence. The effect method shows how representational models can enable the necessary implementation of complex problem-solving, artificial intelligence, and creativity in education and thus fundamentally improve it.

The corresponding paper, which was enabled by generous support from the Institute of Applied Research (IAF), will be published this summer in the renowned MIT Press journal “Design Issues.”

Presenting the AI+Design Lab at MWK Networking Event

Mon, 11 Mar 2024 00:00:00 GMT

On 11. March 2024, we were invited to the networking event "KI in der Hochschullehre" to discuss artificial intelligence (AI) in university teaching, connect and exchange ideas with more than 150 representatives from different universities in Baden-Württemberg.

The event was organized by the Baden-Württemberg Ministry of Science, Research and the Arts (MWK) and took place in the impressive building of the DHBW Stuttgart. Seventeen selected teaching concepts were presented as part of the "Transfer impulses from good practice teaching concepts" program. The AI+Design Lab was one of the selected formats and we presented and discussed our concept in a poster presentation. You can find our poster by clicking the yellow button at the top left.

Digital AI+D lab

Tue, 13 Feb 2024 00:00:00 GMT

The “Digital Lab” is a digital learning space for students and lecturers. The plan is to implement virtual learning spaces, which will be accessible as a web app via a browser.

The learning environment should offer an asynchronous extension to the classic teaching presence format. In order to structure the classic sticky-note board, the analogy of a real space was used. It provides interactive tools, collections of documents or papers, videos and other online materials for self teaching, prepared in such a way that they can be easily located in the digital space.

Will AI shape the design discipline of tomorrow?

Wed, 20 Dec 2023 00:00:00 GMT

Rather than grappling with the changes brought about by a deterministic and speculative AI future, we should seek ways to support sustainable appropriation processes of AI technologies in the design process and other disciplines.

The debate surrounding the opportunities, risks, and consequences of developing for and using AI in the design industry has been in full swing since breakthroughs were announced with the release of ChatGPT and StableDiffusionmore than a year ago. Since then, myriads of models are been released on a weekly basis, leaving busy practitioners challenged with keeping up with the latest developments, and not much time to experiment, and investigate what these new technologies could mean for their individual practice. Whispers of profound upheaval in design professions resonate, whilst voices foreseeing modest shifts are few and far between. Whether one approaches these changes with concern or anticipation hinges on their individual appreciation of the essence and merit inherent in the design process, but junior and senior designers contemplate alike: How will the future of the design discipline look like, and who will shape it?

The instantaneous high-fidelity outputs generated by generative AI technologies not only raise doubts among designers about the continued relevance of painstakingly acquired and cherished skills but, of course, also unsettles other disciplines and professions, like lawyers, writers, and software developers - no one knows for certain whether their work activities will be fully or partially taken over by AI, or if their jobs, for better or worse, will be forever changed.

Gradual Transformation or a Dramatic Revolution?

Current generative AI models showcase remarkable aptitude in tackling varied challenges, yet envisioning the future job market transformation, parallels could also be drawn to the pocket calculator's steady evolution since inception. This notion, garnering considerable approval, implies an organic assimilation of AI, characterized by gradual refinement rather than seismic change, rendering resistance futile. Indeed, human-computer cooperation has evolved consistently over time through the emergence and subsequent enhancement of digital instruments. In a processes of humans shaping technology to being shaped by technology in return, technological developments have always expanded and limited what could be made and thought. Not through the invention, but through the adaptation of digital tools, a collaborative relationship between humans and machines emerged that is still unfolding and which several disciplines, including design, are concerned with.

Some of these interrelations are described anecdotally in Sherry Turkle's article "How computers change the way we think": Students working with early calculators were reported to lose their sense of scale, for example.

For what might we lose our sense when we generate ideas and images within seconds?

It is easy to imagine what began with calculators as the first personal computers in the literal sense having developed very differently. Many of the interaction patterns that have been established in the use of digital tools today are not necessarily the best solutions, but rather stable equilibria that are based on habits and compatibility ecosystems, are easy to learn, are highly useful or have practical benefits.

We know that it is not only technical innovations, but also the design and use of digital interfaces and technologies that significantly shape our current forms of digital interaction and communication and thus create key advantages and disadvantages for digital societies. Therefore, instead of engaging in conjecture based on a presumed deterministic AI future, we should ask ourselves how to incorporate the potential afforded by these technologies into the design process sustainably. It is equally crucial to recognize that designers play a significant role in shaping the adoption processes of AI within other disciplines as well.

Integrating AI into the design process

How do we want AI to change or not change the way we think? The widespread assumption that humans and AI can work and decide better and/or more responsibly collaboratively invites an exchange on this question. The collaborative imperative also asks designers about the future of the discipline: "How do we define the design discipline of tomorrow through our design and use of AI today?" Our early attempts to integrate generative AI technologies into design practice have been the first steps in this process, setting its future trajectory.

Collaborative scenario between humans and AI systems

The collaborative scenario of a human-AI relationship, in which AI systems complement and support users in processes that they dislike or find difficult, offers many starting points beyond flat out automation. Yet there are not yet many clear visions that strive for a stable balance between reducing labor costs and the sustainable development of co-creative work. Confronted with a supposedly extremely complex and closed technology, many designers limit themselves to efficiency-enhancing application of AI tools such as Midjourney or StableDiffusion.

Fuzziness semantics

What can collaboration between humans and AI models look like beyond the interaction paradigm embedded in current applications? The Text2X interaction paradigm, in which AI models generate images or texts based on text input from users, for example, is characterized by a lack of clarity for technical reasons, but also due to inherently individual semantics: The input length is limited to a certain number of characters and an image or concept says more than a thousand words, which neither users are willing to enter nor current models are able to accurately process. This fuzziness was identified early on by designers as a divergent feature to create variants or overcome a blank page.

Dependence on AI

An appropriation of such a work process is also located in questions of the independence from and dependencies of AI. Some tough lessons we learn early on provide the foundation of integral design skills. Who doesn't remember of being out of ideas as an undergraduate? Automating them may not only mean parting with these skills, but also losing the feeling for them and paving the way for new, better, but possibly also less rigorous ways of thinking and erosion of cherished aspects of our practices. Various studies on creative flexibility show for example, that as soon a visual is presented, it strongly limits what we can imagine and hampers our abilities to think in different terms. Of this effect, we are all painfully aware of in participative processes and client presentations. But perhaps we have already found ourselves in a similar situation when we were using ChatGPT or Midjourney? Unfortunately, chances we noticed are very slim.

"Engaging with the worldviews and values inherent in AI models is imperative."

The arbitrariness that co-creative work with AI currently entails is particularly evident here. What can pass as arbitrariness in individual cases is, in sum, a reproduction of the data and values implicitly engrained in AI models. This makes it essential to engage with the worldviews that are inevitably part of every language- and image-based model. While our current debates focuses on the elimination of bias, we will need to accept that an unbiased model might not be possible after all. Similar to our efforts to reconcile our individual meaning perspectives, a single monolithic model will unlikely be able to solve this problem.

This weighs in heavily on the question of how the discipline can influence a collaborative human-ai scenario and design participation in answering important questions about how AI should influence our ability to think and make judgments.

A Challenging task for design

Latent effects of design decisions only become apparent during use - consequences and effects only become visible through long-term adaptation and observation of subsequent decisions, even when we rigorously scrutinize our decisions. The high level of complexity involved in the design and development of AI applications, which ranges from issues of data generation, curation, resource consumption and interaction design to political issues of fair access, biases and discrimination, make AI a challenging, often overwhelming task for design. It would be too easy to separate the design from the technical development and fundamental research from it's later application, as has been often done, and came to haunt us shortly after. If we live in a world riddled by wicked problems, we should think again before leaving technical developments set the trajectory how we as designers engage with AI and how we chose to make accessible.

"Not everyone is comfortable with the definition of a heterogeneous design definition."

The immediate influence that design exerts on the development paths of models and their possible applications is currently rather limited. The integration of design into research processes in general, but especially in the field of AI, is too restrained to make a significant contribution to the sustainable development of models and AI technology. Despite the relevance of these paths, which are not deterministic but whose genesis can be traced back to a variety of factors, we primarily exert indirect influence on them. In this context, we may also need to question our understanding of technology in training and professional practice.

Another redefinition of design, again.

Despite many unforeseen twists and turns, design has prevailed in recent decades with designers always being willing to redefine the discipline itself, and create new sub-disciplines. This always meant that not everyone agreed with every definition of a pluralistic concept of design.

By appropriating AI software to generate images and ideas, restructuring the design process and outsourcing work steps, it is not AI, but designers themselves who shape the design practice of tomorrow. Whether our opportunities for active participation in these development processes will be limited to this part in the future depends on whether we want to take up the challenge of breaking down and redefining the boundary between design and technology.

Controller – Lab Week Workshop 2023

Tue, 19 Dec 2023 00:00:00 GMT

{frontmatter.title}

The integration of sensor technology with artificial intelligence opens up a world of intriguing new possibilities for interaction. The goal of this recent workshop was to explore creative and unconventional ways of engaging with devices and simple games, and bring these ideas to life in the form of functional prototypes.

Participants delved into the essentials of machine learning, examining the foundations of neural networks and their operational principles. The workshop addressed key questions about training a model with custom data, the steps necessary for data collection and preparation, crucial parameters in the training process, and strategies for model improvement. A crucial element of the workshop was to run the trained models on a microcontroller, enabling the creation of custom controllers.

In this workshop, we worked with Edge Impulse, a platform that enables code-free training of machine learning models and their deployment on edge devices.

Initially, students concentrated on understanding data, exploring the types of data that could be captured using the given sensors and how these could translate into distinct interaction patterns. The objective was to develop interesting and unconventional sensor-based interactions that could be consistently recognized by a machine learning model. After a phase of experimentation with sensor interactions, the students transformed their ideas into specific concepts for interacting with technology and implemented them in hands-on prototypes.

The concepts were brought to life using the Arduino Nano BLE Sense microcontroller. For those interested in the technical details and code from the workshop, information can be found here.

The week-long workshop concluded with a showcase of diverse interaction prototypes, demonstrating the creative and technical achievements of the participants. Highlights included Olman Arias Valverde's pressure-gesture-controlled lamp using barometer data (Fig.2), and Tamara Solís Matamoros's cooking-themed gesture game, where actions such as flipping a pancake were translated into game controls (Fig.3). Ron Eros Mandic and Lukas Speidel created a sound-based snake game controller (Fig.4).

The workshop aimed to encourage students to develop engaging and novel ways of interacting with the world around us. It demonstrated the potential and practical applications of machine learning and sensor technology to create innovative interfaces and user experiences.

Zoomaker - Friendly house keeping for your AI model zoo and related resources

Tue, 19 Dec 2023 00:00:00 GMT

Zoomaker is a command-line tool for maintaining project related "AI model zoos" in a friendly way.

👉 Zoomaker Github Repository

Motivation

Teaching AI is hard, not alone because of the complexity of the topic, but also because of the complexity of the tools. I'm running for instance a course on Designing Prompts, which uses Automatic1111 (think of an UI for Stable Diffusion) to make image generation accessible for the students. This is a great combination, but it's also a very complex combination. From installing the right version of Python, to installing the right version of Automatic1111, to installing the right version of the models, to installing the right version of the datasets of the Automatic1111 extension ... to how to actually start everything on the server in the terminal. It is simply too much to ask from the students, and is in addition no longer manually feasible.

However this dependency hell complexity is also know in other domains. For instance in the web development world, where tools like npm are used to manage the dependencies.

So my idea was really simple, what if something like npm install would exist for creative AI projects which depend heavily on manifold resources? It turned out to be relatively easy to build such a tool, and to be very useful for my students and courses.

Zoomaker Features

Zoomaker offers a range of features designed to manage and use models, embeddings, git repositories and downloads. Its main attributes include a centralized configuration through the zoo.yaml file, ensuring a single source of truth for all resources. It allows users to precisely track and freeze versions of the resources, ensuring consistency and reliability in their projects. Zoomaker also optimizes bandwidth usage by caching models locally, avoiding repeated downloads. This was a must, as a typical project with Stable Diffusion can easily require 100GB of models. Hence, it manages disk space efficiently by using symlinks for larger files, while duplicating smaller files under 5MB, streamlining the installation process and optimizing overall resource management.

Installation

Zoomaker is a Python command-line tool, which can be installed with pip install zoomaker.

zoo.yaml Example

Here is an example of a zoo.yaml file, which is used to configure all resources for a project. It is a simple YAML file, which can be edited by hand. It defines the name of the project, the version, the description, the author, and the resources. The resources are defined as a list of resources, which can be of type "git", "huggingface" or "download". Each resource has a name, a source, a type, a install path and an optional revision. The "scripts" section defines a list of scripts, which can be run with zoomaker run <script-name>.

name: my-automatic1111-model-zoo
version: 1.0
description: Lorem ipsum
author: your name

resources:
  image_generator:
    - name: automatic1111
      src: https://github.com/AUTOMATIC1111/stable-diffusion-webui.git
      type: git
      revision: 22bcc7be428c94e9408f589966c2040187245d81
      install_to: ./

  models:
    - name: v2-1_768-ema-pruned
      src: stabilityai/stable-diffusion-2-1/v2-1_768-ema-pruned.safetensors
      type: huggingface
      install_to: ./stable-diffusion-webui/models/Stable-diffusion/

scripts:
  start_webui: |
    conda activate automatic1111
    cd /home/$(whoami)/stable-diffusion-webui/
    ./webui.sh --theme dark --xformers --no-half

To install the resources, simply run zoomaker install in the same directory as the zoo.yaml file. This will install all resources, in this case it will install Automatic1111 and the model to the stable-diffusion-webui folder. To start the webui, simply run zoomaker run start_webui. This will run the commands defined in the "start_webui" script. Done :)

See the Zoomaker Readme on Github for more examples and full reference of the structure of the file.

Conclusion

In conclusion, Zoomaker proved for me to be an indispensable tool for streamlining the management of AI models and resources in creative and educational AI projects. Its project centric design, embodied in the zoo.yaml file, and the embracing the model zoo metaphor, offers a user-friendly approach to handling complex dependencies. By simplifying the installation process, optimizing resource management, and enhancing overall project consistency, zoomaker addresses the challenges faced in AI education and development. It empowers students and practitioners in the realm of AI project management, making the dependency hell of AI projects more accessible and manageable for a broader audience.

Transform @ HS Trier

Wed, 25 Oct 2023 00:00:00 GMT

Join us at the 4th KITeGG winter school at the HS Trier.

🌐 Find the full programme at: kind-lab.de/transform2023/schedule

unStable Mirror at Push UX 2023

Thu, 19 Oct 2023 00:00:00 GMT

At the PUSH UX Conference 2023, a platform for hands-on design professionals, product minded innovators and creative technologists, we are proud to have exhibited our mixed media installation unStable Mirror.

The installation combines technology and art to provide a hands-on experience in understanding generative AI algorithms. We are happy that many visitors had a lot of fun while exploring there own self-feedback loop to find out if the results from a generate AI are truly neutral and unbiased.

<cite>Images by PUSH UX 2023 Conference</cite>

This article was written with the support of ChatGPT

Interview AI in Design Education @ Page Magazine, Issue 11.2023

Sun, 01 Oct 2023 00:00:00 GMT

Interview, KITeGG Research Project and AI in Design Education, Page Magazine, Oktober Issue, DE

State Secretary Arne Braun visited the AI+Design Lab

Sun, 20 Aug 2023 00:00:00 GMT

On August 14, 2023, the State Secretary in the Ministry of Science, Art and Research, Arne Braun, visited the University of Design Schwäbisch Gmünd. We had the pleasure of welcoming him to the AI+Design Lab and discussing with him the implementation of and engagement with AI in design education. In a small exhibition, we presented projects that creatively and critically examine the use of AI technology in the design of products and media: Among others, the thesis "anticipate" by Niklas Muhs, that explores the impact of large language models (LLMs) on design practices, as well as the installation "Unstable Mirror" that allows participants to explore the underlying model of a generative AI process in an explorative and playful way, discovering potential biases of the model.

More information on the visit by State Secretary Arne Braun can be found here.

Photo by Maurice Rio.

Anticipate - Challenging hidden assumptions within the design process

Sun, 23 Jul 2023 00:00:00 GMT

Challenging hidden assumptions within the design process

Bachelor Thesis by Niklas Muhs (in BA Product Design)

The project Anticipate delves into the intersection of design and artificial intelligence (AI). It explores how AI influences design processes, especially during initial ambiguous and complex stages. The project acknowledges the limitations of AI, which often projects current societal thought patterns into the future, potentially reinforcing biases and existing mental models. This can lead to a cycle where AI-based designs further solidify current societal beliefs and structures.

The core idea behind "anticipate" is to utilize AI in a way that enhances the design process without overriding human decision-making. It aims to use AI to uncover hidden assumptions and biases in the design process, thereby encouraging more holistic thinking and innovation. This is achieved by using AI not to generate decisions or ideas, but to provide stimuli for reflection and to challenge existing thought structures. The project seeks to identify 'unknown unknowns' in design, helping designers to consider aspects they might not have recognized independently.

A prototype was developed to make the research hypothesis tangible and to explore the potential and applicability of the concept, including the uncovering of unknown unknowns through user tests. The user interface of the prototype allows users to input objectives and constraints, which helps in generating relevant presumptions and organizing the necessary validations. The interface also educates users on the AI model's operation, aligning expectations with outcomes and enabling users to define the preferability of futures.

The "anticipate" interface includes a collaboration feature, allowing teams to reveal and discuss individual beliefs and assumptions, which could improve stakeholder alignment. Users can highlight critical presumptions to emphasize their significance and convey priorities to the AI model, increasing the likelihood of receiving relevant results. Additionally, the interface allows the creation of additional views for prioritizing presumptions based on personal criteria.

Overall, "anticipate" aims to responsibly use AI in design processes, not to generate decisions or ideas, but to provide stimuli for reflection and challenge existing ways of thinking.

Further development of Anticipate was enabled by generous support by the Institute of Applied Research (IAF).
Publication by Niklas Muhs and Aeneas Stankowski currently under review.

unStable Mirror: A Journey into the Heart of AI Generativity and Bias

Sat, 15 Jul 2023 00:00:00 GMT

import Video from '@/components/Video.astro'

In an era where artificial intelligence (AI) has permeated various facets of our lives, it becomes imperative to understand not just its capabilities, but also its limitations and inherent biases. The unStable Mirror is a cutting-edge mixed media installation that serves this very purpose. Developed with a keen eye on AI’s generative processes, this installation delves into the intriguing world of AI models, particularly focusing on the generative algorithms like Stable Diffusion.

unStable Mirror is an inventive installation that combines technology and art to provide a hands-on experience in understanding AI. The setup includes a custom-built display frame equipped with an integrated camera and an information panel with a buzzer button. The core of this installation is an AI processing pipeline based on Stable Diffusion, a generative AI algorithm.

The installation is an interactive exploration tool. Participants are invited to interact with the installation, which captures their image like a mirror when they press a red button. This image then undergoes a transformation process.

Upon capturing an image, the installation first extracts its edges using canny edge detection. Simultaneously, an image classification model (CLIP: ViT-L-14/openai) generates a textual description of the captured image. This dual input – the edge map and the textual description – is then fed into the Stable Diffusion model to generate a new image.

This is where unStable Mirror begins to unfold. The process doesn’t stop at just one iteration. Instead, it recursively feeds the AI-generated image back into Stable Diffusion. With each cycle, the original image undergoes a metamorphosis – contours are reinterpreted, shapes are reshaped, and new variations emerge. This recursive process is central to the installation’s explorative journey.

One of the core objectives of this installation is to illuminate the potential biases inherent in AI models. As the AI reinterprets the initial image over multiple iterations, any biases in the algorithm become amplified and more visible. This recursive nature allows participants to witness firsthand how AI models can deviate from neutrality and objectivity.

The installation thus serves as a mirror – albeit an unStable one – reflecting not just the physical image of the participant but also the underlying biases of the AI model.

The project is more than just an art installation. It's an educational tool, a conversation starter, and a mirror reflecting the current state of AI. By enabling participants to interactively explore the generative process of AI, it raises awareness about the technology’s capabilities and its limitations and can play a crucial role in educating and informing the public.

unStable Mirror is an outcome of the AI+D Lab residency program at the HfG Schwäbisch Gmünd.

Credits

Resident Researcher/Artist: Christopher Pietsch
Programming and AI Pipeline: Christopher Pietsch
Hardware Product Design: Felix Sewing
Extended Team: Aeneas Stankowski, Alexa Steinbrück, Rahel Flechtner, Felix Sewing, Benedikt Groß
Video and Photo Documentation: Katharina Neugart

Prompt Battle at reshape forum

Wed, 10 May 2023 00:00:00 GMT

On Wednesday we opened our doors for the 3rd KITeGG summer school. After a warm welcome from the organizer team followed by a guided tour of the university we started our second Prompt Battle event at HfG Schwäbisch Gmünd.

This time to top the event even a bit more than the previous one, we invented spezial moderators: Bernadette Geiger and Ella Zickerick. They are part of the team from HTW Desden that developed the Prompt Battle event format.

After the selection of the participating from the audience, the match could start! The tournament was held in a 1v1 single elimination bracket. The goal: To convince the audience that your generated image fulfills the task best with their prompting skills or a bit of luck from the latenspace from Dalle2. But as always, the audience decided the winner of the rounds, so a little humor in the pictures couldn't hurt either.

The event was a lot of fun for all of us, not least because of the good moderation by Bernadette Geiger and Ella Zickerick and the great organization team in the background. Many thanks to everyone involved!

<cite>Photos and Videography by Stefan Eigner</cite>

Making (Non-)Sense Workshop

Mon, 08 May 2023 00:00:00 GMT

{frontmatter.title}

In a collaborative effort, Rahel Flechtner from HfG Schwäbisch Gmünd and Jakob Kilian from the Köln International School of Design (KISD) conducted a workshop titled “Making (Non-)Sense.” This workshop was part of the joint research project KITeGG and took place in 2023 at both HfG Schwäbisch Gmünd and KISD, involving students from both institutions. The workshop aimed to equip design students with a hands-on, intuitive understanding of AI technologies, focusing on the development of sensor-based interactions. It was structured in two engaging segments that combined playful learning with practical experimentation.

The first part of the workshop introduced students to the fundamentals of machine learning through a creative role-playing activity. In this exercise, students embodied the components of a sensor-based AI system. They took on various roles, such as human sensors and algorithms, to understand how sensor data is collected and processed. By acting out scenarios like an office setting, where sensor data was used to determine whether a person was working, students learned about data acquisition and model training in an accessible and engaging manner. This playful introduction aimed at demystifing the technical capabilities and limitations of AI, making the concepts more relatable.

Building on this foundation, the second part of the workshop was an exploratory, hands-on session. Students worked with ready-made hardware and software modules, allowing them to dive deeper into the technology. They were introduced to tools that facilitated the code-free training of machine learning models, which could then be deployed on microcontrollers. This segment emphasized active engagement and creative ideation, encouraging students to experiment with sensor-based AI technology to develop new interactions. By collecting and preparing data, training models, and running them on microcontrollers, students gained practical experience in the iterative process of designing AI-driven interactions.

<swiper-container navigation="true" cssMode="true" grabCursor="true" loop="true"> <swiper-slide> <cite>Technical setting of the workshop: ready-made sensing and acting modules and the online platform used for machine learning.</cite> </swiper-slide> <swiper-slide> <cite></cite> </swiper-slide> <swiper-slide> <cite></cite> </swiper-slide> </swiper-container>

As the briefing of this hands-on experience, students were tasked with creating a link in an AI-driven Rube Goldberg machine-like chain reaction, integrating sensor-based modules that classify a specific input pattern to trigger a specific action. All links were then combined into a complete sequence, as shown in the image below, and presented in a group performance.

<swiper-container navigation="true" cssMode="true" grabCursor="true" loop="true"> <swiper-slide> <cite>Impressions of the final AI-based Rube Goldberg machine.</cite> </swiper-slide> <swiper-slide> <cite></cite> </swiper-slide> <swiper-slide> <cite></cite> </swiper-slide> <swiper-slide> <cite></cite> </swiper-slide> <swiper-slide> <cite></cite> </swiper-slide> </swiper-container>

This workshop format not only aimed at increasing students' technical literacy but also fostered their ability to communicate about AI topics and collaborate effectively with data scientists and AI experts. The course materials, available as open-source resources on GitLab, provide valuable tools for educators and students interested in similar interdisciplinary explorations. In June 2024, we also published a paper on the workshop format, explaining the process and content of the workshop in more detail. You can download the author's version of the paper by clicking the yellow button on the top left of this article. Additionally, a video documenting the workshop offers a visual summary of the activities and outcomes. With our teaching format, “Making (Non-)Sense”, we contributed to the still open question of how to to teach AI in design. By combining playful role-playing with practical, hands-on experience, we see great potential for preparing future designers to integrate AI into creative and meaningful interactions.

reshape forum @ HfG Schwäbisch Gmünd

Tue, 02 May 2023 00:00:00 GMT

It’s summer again! ⛱️ Join us for the 3rd KITeGG summer school, which is hosted for the first time by HfG Schwäbisch Gmünd and organized by AI+D Lab.

From the 10th to the 12th of May various talks and panels on the future of AI and Design will be held at the Auditorium of the HfG Schwäbisch Gmünd.

In advance, from the 8th to the 10th of May, students are able to participate in a diverse range of AI-focused workshops held by educators from the KITeGG partner universities (Cologne, Mainz and Trier)!

Wednesday

We will kick off the symposium on Wednesday afternoon with a warm welcome from the organizer team, followed by a guided tour of the university to get to know the host university of this year's KITeGG summer school. After we have warmed up our limbs in this way, we will gather in the Aula for the most exciting event of the year - a “Prompt Battle” (5-7 pm), a live event where people compete against each other using text-to-image software and the audience is the judge (prepare to move your feet)! We end the evening with a relaxing (and vegan-friendly) barbecue in the courtyard of HfG.

Thursday

The schedule on Thursday will be packed! We are looking forward to an inspiring list of talks from international experts working and thinking in the field of the intersection of AI and design: From 👩‍🔧Nadia Piet, the founder of the flourishing online community AIxDesign we will hear her thoughts on the “UX of AI” and the various challenges posed by these adaptive, intelligent, and semi-autonomous systems for designers. Followed by 👩‍💻Catherine Breslin, a renowned expert in the field of Conversational AI, who has been in the founding team of Amazon Alexa: How does the way people converse with each other inspire the design of voice assistant technology and how do Large Language Models (LLM) reshape this field of application? Before the lunch break, we will hear from 👩‍💻Ploipailin Flynn on how she got tired of well-designed tech products and AI-based services that turned “accidentally racist” and get to know resources, tools and design principles for addressing discrimination in tech.

On Thursday afternoon, we will hear from Brooklyn-based UX researcher and strategist 👩‍💻Emily Saltz on how generative AI technologies and synthetic media demand new mechanisms for safety considerations and UX best practices. We will continue with a panel discussion 👯 “AI Industry - professional requirements and fields of activity for designers”. The day will be rounded off by a talk by members of 👨‍💻oio studio (London) in which they present their fun and thought-provoking work on “future products and tools for a less-boring future”.

Friday

On Friday morning we will start the programme with a talk by 🤹‍♂️Simon Maris (from our KITeGG partner university Trier) on sustainability, design and AI. He will be followed by 👨‍💻Tom White, an artist based in New Zealand, who has been active in computational art for more than two decades with themes of artificial intelligence, interactivity, and computational creativity.

At 12 pm we will do a bit of self-reflection in the panel 👯“KITeGG: Learnings from one year of AI education at Art & Design schools”. For more than a year, the 5 participating universities have now gathered concrete experience in teaching, demystifying and unpacking AI. What can they report? To close off the symposium we will enjoy a presentation of the results of the four student workshops from the beginning of the week while enjoying some 🍕 Pizza!

We're looking forward to welcoming you in Schwäbisch Gmünd!

🌐 Find the full programme at: reshapeforum.de

✅ The event is free and open to everybody. Thank you for registering online here

Video recordings:

Emily Saltz: "AI Safety & Disclosures in Product Design: Practices in Flux"

oio: "Future Everyday Products"

Simon Maris: "A priori: Automated Pattern Recognition Inspiring Overdue Renewable Inventions"

Tom White: "Machine Abstractions"

Panel KITeGG: "Learnings from one year of AI education at Art & Design School"

Panel AI Industry: "Professional requirements and fields of activity for designers"

AI Is Not a Wildcard: Challenges for Integrating AI into the Design Curriculum @ EduCHI23

Fri, 28 Apr 2023 00:00:00 GMT

We were more than happy to present and discuss our experiences and thoughts on AI education for design students at the EduCHI symposium 2023 in Hamburg. Our paper 'AI Is Not a Wildcard: Challenges for Integrating AI into the Design Curriculum' [1] contributed to the as-yet unsolved challenge of how to structurally integrate AI into the design curriculum. You can find the paper here or download an authors' version of the paper by clicking the button on the left and we provide a comprehensive summary of the paper in this article.

The EduCHI symposium brings together an international community of scholars, practitioners, and researchers to work on the future of Human–Computer Interaction (HCI) education. Especially given the rapid developments in artificial intelligence in recent months and the increasing impact of the technology on our professional practice, we found it particularly exciting to discuss current and future challenges facing HCI educators and to be inspired by teaching practices from the community.

Ressources

[1] Rahel Flechtner and Aeneas Stankowski. 2023. AI Is Not a Wildcard: Challenges for Integrating AI into the Design Curriculum. In EduCHI 2023: 5th Annual Symposium on HCI Education (EduCHI ’23), April 28, 2023, Hamburg, Germany. ACM, New York, NY, USA. DOI: 10.1145/3587399.3587410.

AI Is Not a Wildcard!

Wed, 26 Apr 2023 00:00:00 GMT

Within the last year, we observed a growing interest among students in incorporating AI into design. However, we have also noticed that AI technologies are sometimes used as “wildcards” in the design process. The insufficient understanding of AI‘s capabilities and limitations often results in unrealistic design proposals. Design education should equip students with the necessary knowledge and skills to engage with AI technologies in a meaningful way in their design processes. However, advising students on this topic requires a comprehensive understanding of AI, which places significant demands on educators. In the paper “AI is not a wildcard: Challenges in integrating AI into the design curriculum”, submitted to EduChi23, we shared and discussed our thoughts and first experiences on integrating AI in the design curriculum. In the following article, we highlight the most important points from the paper. For more information, please read the paper [1].

The use of AI technologies is becoming more important in the design of human-computer interactions and user experiences. Designers are well-positioned to drive stakeholder-centered adaption of AI technology, but this also poses new challenges for design professionals. The pace of development and complexity of AI technology makes it difficult for designers to stay abreast of new developments and envision novel solutions. There is a lack of structured integration of AI technology into design education, and designers often lack technical knowledge to effectively collaborate with AI engineers. Future designers need to learn new skills when working with AI, such as assessing the capabilities and limits of machine-learning technologies, evaluating the cost of development and computation, and effectively communicating possible harm and unintended consequences. While there are many resources available to learn about AI technologies, integrating them into design education remains a challenge.

The aim of our project is to integrate AI technologies into the design curriculum to give students the necessary intuition for the technology to create meaningful and technically feasible AI-based systems and user experiences. This intuition involves that students should be able to make informed decisions about the use of AI technologies, assess their capabilities and limitations, and understand their cost in terms of development and computation. They should also be able to use existing technologies in creative and unconventional ways, and use entry and intermediate-level scaffolding tools to prototypes and test the AI-based interactions they design. However, they should also be aware of the possible consequences of implementing AI technologies and evaluate if it is possible to mitigate or avert possible side effects by design. They must be able to use speculative and conceptual approaches consciously instead of using AI as a wildcard, and should develop a precise language for communicating about AI topics.

Our observations have shown that approaches to developing and applying AI intuition are often taught in separate course formats that may not be connected. As a result, foundation courses often lack meaningful connections to other formats such as studio courses. Follow-up formats for the application of AI, in turn, can only be effective if educators can supervise them in a qualified way. We see the high demands that AI technology places on educators as one of the biggest challenges to be overcome when structurally integrating AI education into the design curriculum.

At the University of Design Schwäbisch Gmünd we test the format of an AI-lab as a place of specialization to overcome this challenge. The AI+Design Lab offers entry-level formats for building an AI intuition in primary teaching and is linked to studio education and project-based teaching. The lab provides students with technical support and advice for their projects and serves as a contact point for educators to learn and discuss the technology. With our paper, we aim to inspire further research and discussion on integrating AI into design education.

Ressources

[1] Rahel Flechtner and Aeneas Stankowski. 2023. AI Is Not a Wildcard: Challenges for Integrating AI into the Design Curriculum. In EduCHI 2023: 5th Annual Symposium on HCI Education (EduCHI ’23), April 28, 2023, Hamburg, Germany. ACM, New York, NY, USA. DOI: 10.1145/3587399.3587410.

This article was written with the support of ChatGPT

Prompt Battle

Thu, 30 Mar 2023 00:00:00 GMT

On Wednesday we officially opened the AI+Design Lab at the HFG and invited colleagues and students to come by and connect. Following a brief introduction to the research project and the project members working with the AI+D Team, we offered pizza and drinks. The highlight of the evening was the Prompt Battle - an event invented by a Team at the HTW Dresden. It's inspired by rap battles, but instead of rhymes, participants battled each other by generating images with Stable Diffusion. The tournament was held in a 1v1 single elimination bracket.

Each match consisted of a best of 5: Up to five challenges where given, according to which the students then took their best shot at prompting Stable Diffusion to produce a witty image. Each round the audience judged both pictures and decided the better image by gathering behind their contestants of choice.

The Prompt Battle was a fun and exciting way to showcase the capabilities of AI software and its applications. The competition was intense, and the participants were fearless in their approach to the contest. The keyboards were working overtime, and the algorithm was put to the test, pushing its parameters to the limit.

It was a great way to engage with the audience, showing them the potential of AI technology and its ability to enhance the design process. The event provided an opportunity for participants to explore the relationship between AI and design, and how these two fields can work together to create innovative and exciting new applications.

The event was a huge success, and the AI+Design Lab is looking forward to working with the participants to develop new applications at the intersection of design and AI. The lab is committed to exploring the potential of AI technology to revolutionize the design process and create new and innovative solutions to some of today's most pressing design challenges.

This article was written with the support of ChatGPT

Pathways for integrating AI education in the design curriculum @ IxDA education summit 2023

Mon, 06 Mar 2023 00:00:00 GMT

Interaction design is a field that continually evolves alongside technological advancements, facing ever new possibilities and challenges. The same applies to interaction design education, which must adapt to meet the changing demands of the discipline. Especially the emergence of AI-based learning systems that change their behavior over time and with use pose new questions for the discipline. At the IxDA interaction design education summit 2023 we had the opportunity to discuss this exciting topic with other educators, learners and practitioners. With our talk titled "Pathways for integrating AI education in the design curriculum" we contributed our experiences and thoughts on AI education in design from the KITeGG project and the establishment of the AI+D Lab at HfG Schwäbisch Gmünd.

We also compiled our experiences and thoughts into a comprehensive paper. For further information and access to the paper, please follow this Link.

This article was written with the support of ChatGPT

SandwichNet

Sat, 04 Mar 2023 00:00:00 GMT

<div class="inline-block relative overflow-hidden rounded-full transition-all font-medium default-text bg-white text-black hover:bg-primary-yellow" style="box-sizing: border-box; border: 0.25rem solid black;"> <audio class="relative" id="player" src="/audio/sandwichNet_website_artile.mp3" onended="document.getElementById('pauseButton').style.display='none'; document.getElementById('playButton').innerHTML = 'read aloud again ... '; document.getElementById('playButton').style.display='block';"></audio> <div class="" style="width: 100%; height: 100%; border-radius: inherit;"> <button id="playButton" class="cursor-pointer pt-2.5 pb-2 px-[18px]" onclick="document.getElementById('player').play(); document.getElementById('pauseButton').style.display='block'; document.getElementById('playButton').style.display='none';" style="display: block; width: 100%; height: 100%; border-radius: inherit; overflow: hidden"> <span style="position: relative; top:0.1rem">▶</span> ... read article aloud </button> <button id="pauseButton" class="cursor-pointer pt-2.5 pb-2 px-[18px]" onclick="document.getElementById('player').pause(); document.getElementById('pauseButton').style.display='none'; document.getElementById('playButton').innerHTML = 'continue reading ...'; document.getElementById('playButton').style.display='block';" style="display: none; width: 100%; height: 100%; border-radius: inherit; overflow: hidden">pause reading ...</button> </div> </div>

Computers can crunch numbers in seconds that would take us humans hours to process. But for a long time, the opposite was also the case. In many simple classification problems, we humans were faster and more accurate than computers, for example, in the domains of image and sound classification. Today, we still outperform computers in some complex classification problems, in others we don’t stand a chance against modern machine learning algorithms. The back propagation algorithm is one of the main reasons this is possible. It gradually improves a randomly initiated model until it reaches the programmed goal.

Teaching neural nets and machine learning to students of non-technical disciplines often gets tricky when we look at what goes on within the training process of a model. While a parametric model with thousands to billions of parameters configured to output correct approximate predictions is already pretty complex, understanding the overarching principle that makes modern machine learning algorithms possible is key. And that's where the back propagation algorithm comes in.

Inspired by a particularly simple and memorable example from the textbook: "You Look like a Thing and I Love You" from Janelle Shane (Wildfire, 2020), we created an interactive web application that lets students copy the actions of a training algorithm.

We're using the tool in our foundational machine learning courses to help students understand the possibilities and limitations of the technology better down the road. We really believe that teaching these highly technical concepts in a hands-on way will benefit the students when they start working on technologically complex problems and projects in the future.

Classification algorithms are pretty amazing, but when you get down to it, the magic is just a bunch of familiar multiplications and additions, plus some pretty simple methods that introduce non-linear functionality. The history of neural nets, like the mathematically simple Perceptron, can help design students see that deep learning is totally within their grasp.

But what is needed to put our neurons to use is some training data and an algorithm, which is able to bend our neurons to our will.

We're going to start talking about derivatives and slopes, which might get a bit technical. But we think it's worth zooming in on this optimization process instead of treating it as a black box. To do this, we've implemented a very simple neural net:

SandwichNet

In a fictional universe, mysterious random sandwiches emerge from a magic portal. Our goal is to train an algorithm that can estimate whether a sandwich that just entered our dimension will be tasty or disgusting.

This is where our tool comes into the game: SandwichNet visualizes a simple neural network consisting of a few fully connected layers. The tool has three levels, which cover the different levels of complexity in the learning content.

<swiper-container navigation="true" cssMode="true" grabCursor="true" > <swiper-slide> <cite>First level of SandwichNet, similar to perception</cite> </swiper-slide> <swiper-slide> <cite>Second level of SandwichNet with a hidden layer</cite> </swiper-slide> <swiper-slide> <cite>Third level of SandwichNet with the semi-automatic backpropagation process</cite> </swiper-slide> </swiper-container>

When you click on the “magic portal” (the black oval in the center of the bottom), data samples appear, represented by sandwiches with randomly selected ingredients. The ingredients are translated into numerical values for the input layer of SandwichNet. The output layer has one neuron that predicts how tasty each sandwich combo is. To show not only the bare number at the end, we get help from a "Taste-O-Meter", a circular scale with a needle, to indicate how tasty a sandwich combination is. The needle moves up into the green area if the prediction is a delicious sandwich, or down into the grey area if it's inedible.

So, in level one and two, we're going to train our own neural network by tweaking the weights of each individual neuron by increasing or decreasing their values and immediately classify the randomly generated sandwiches.

In the third level of SandwichNet, we've moved away from manually adjusting the weights. Now, we're optimizing SandwichNet in a semi-automated process using backpropagation. We can use the plus and minus buttons on the "Taste-O-Meter"-scale to specify the error by determining a value that should be the correct result of the prediction. This means calculating the output with randomly initialized weights and then calculating the difference between the resulting and expected values. Once we know how much the deviation is off, we adjust the weights by clicking the train button to get the error down. We can keep doing this until the neural network makes the right predictions.

Feel free to give it a try on the links below. You can use it in your own classes and adapt it to suit your needs. The code is provided via a GitHub repository.

Ressources

Shane, Janelle. You Look like a Thing and I Love You. Wildfire, 2020.
p5js (https://p5js.org/)
Links to the individual different levels of SandwichNet:
- Level 1:
  - https://hfg-gmuend.github.io/SandwichNet/index.html?level=1&lang=en
- Level 2:
  - https://hfg-gmuend.github.io/SandwichNet/index.html?level=2&lang=en
- Level 3:
  - https://hfg-gmuend.github.io/SandwichNet/index.html?level=3&lang=en
GitHub repository of SandwichNet:
- https://github.com/hfg-gmuend/SandwichNet

Discussion @ German UPA UX Chat Feb. 23

Fri, 10 Feb 2023 00:00:00 GMT

At the German UPA's UX Chat on "Artificial Intelligence and UX", we discussed new opportunities and challenges for design at the intersection of AI and user experience together with Matthias L. Chavanne, Lead AI & UX Designer at IBM. A recording of the discussion is available here.

Contextual Interactions

Mon, 19 Dec 2022 00:00:00 GMT

Contextual Interactions

In this lab-week workshop, we explored sensor-based interactions using machine learning - hands-on and through simple and accessible tools. Machine learning enables us to recognize contexts or events from images, sound, motion, and other sensor data, which can serve as an interface to systems or objects. The potential for design, however, goes far beyond these technical possibilities: it enables us to create new, diverse qualities of interaction. Guided by experimentation, we approached exciting or unusual forms of interaction with objects and explored different qualities of interaction. The aim of this workshop was to gain an insight into the basics of machine learning and to explore the potential of the technology for sensor-based interactions.

The concepts were prototyped by the students using the teachable machine web service and webcams. The prototypes demonstrate various interaction modalities. <swiper-container navigation="true" cssMode="true" grabCursor="true" loop="true"> <swiper-slide> </swiper-slide> <swiper-slide> </swiper-slide> <swiper-slide> </swiper-slide> <swiper-slide> </swiper-slide> <swiper-slide> </swiper-slide> <swiper-slide> </swiper-slide> </swiper-container>

Correlations Forum 2022 @ HfG Offenbach

Mon, 14 Nov 2022 00:00:00 GMT

Join us at the 2nd KITeGG winter school at the HfG Offenbach.

🌐 Find the full programme at: correlationsforum.de

Breaking the Pattern - AI as a Tool in the Context of Technology Design @ correlations forum, HfG Offenbach

Mon, 14 Nov 2022 00:00:00 GMT

Our talk ‘Breaking the pattern - AI as a tool in the context of technology design’ at the 2nd KITeGG Symposium, the Correlations Forum for AI in Art and Design at the HfG Offenbach, was a great opportunity to reflect on AI and design. Is AI just another tool for the designer’s toolbox, or does it introduce paradigmatically different ways of interacting with technology? We tried to make an educated guess based on what we've observed in design education and applied research on machine learning in recent years. We talked about opportunities for designers to take advantage of the amazing new possibilities this technology offers and possible strategies for shaping design education to have a real impact on how AI technologies are developed and used.

The project KITeGG is a joint project of Mainz University of Applied Sciences, Offenbach University of Applied Sciences, Schwäbisch Gmünd University of Applied Sciences, Köln International School of Design (TH Köln) and Trier University of Applied Sciences. Within the project's duration until the end of 2025, the universities will alternately organize two symposia each year on the topic of AI and design.

Inaugural Lecture Visiting Professorship Creative AI

Sun, 16 Oct 2022 00:00:00 GMT

We, Aeneas Stankowski and Rahel Flechtner, are very happy to be part of the KITeGG project and to start our position as visiting professors for Creative AI at the University of Applied Sciences Schwäbisch Gmünd! On 26/10/22 at 3:30pm, we will introduce ourselves and our ideas and topics for the next three years of teaching and research in AI and Design in our inaugural lecture in the auditorium.

Hiden Layers @ Köln International School of Design

Mon, 18 Jul 2022 00:00:00 GMT

Join us at the 5th KITeGG summer school at the Köln International School of Design.

🌐 Find the full programme at: hiddenlayers.de