Kars Alfrink

Today I gave the keynote at the CLICKNL Design Innovation Sessions (Digital Futures) at Dutch Design Week in Eindhoven. Slides and references are on my blog. A short thread on what I tried to say. leapfrog.nl/blog/archive...

View full thread

I ended with a question for the room: if AI is not one thing, not an agent, and neither our savior nor our doom, what metaphor does your mind go to?

Many thanks to Noortje Braaksma, Sanne Wiltink and the rest of the CLICKNL team for the invitation, and to Dymphie Braun for the moderation.

Rather than resolve these tensions, I suggest we keep them, and use them as an engine for ways of designing with GenAI that contribute to social progress.

Three cases point that way: Vera van der Burg’s Text-to-Clay, Blaze Type’s test of Monotype’s AI type tool, and The Call by Holly Herndon and Mat Dryhurst. All treat pipelines and workflows as something to design, and all value effort and friction.

Debates about design and generative AI tend to get stuck in for-or-against positions. I propose we look at practices instead: specific GenAI technologies, used in specific contexts, by actual people.

Those practices are full of tensions. GenAI tools deskill designers and open up new ways of designing. They are built on other designers’ past work: whose, and who gets to claim the results? Work gets automated, but often it is delegated more than replaced, transforming practices as a result.

On 22 September, Sunjoo Lee, Advaita Jairam and I ran "From Soil to Servers" at the Philosophy of Human-Technology Relations conference (PHTR 2026) in The Hague. Advaita captured the whole session in a drawing.

View full thread

I keep running these to see how hands-on making changes the way we imagine regenerative computing futures. Thanks to everyone who joined, and to PHTR for having us. Photos: Pranav Prasun.

And a shout-out to Taylor Stone, who suggested we submit this to PHTR in the first place.

We talked about regenerative approaches to computing, building on Fieke Jansen's work. Sunjoo introduced her electric garden art practice. Then everyone made mud batteries: microbial fuel cells that harvest electrons from bacteria in soil.

We closed by reflecting together. Lovely, thoughtful participants. We talked about why bacteria make electricity at all, how precious electricity feels when it comes from a jar of mud, and whether this blurs the line between nature and technology.

On 21 October I'm giving a keynote at CLICKNL's Design Innovation Sessions during Dutch Design Week: "Keep the tension: designing with generative AI"

View full thread

Rather than trying to resolve these tensions, I'm looking for emerging design practices that hold them and use them as an engine for moving forward. They offer concrete, existing glimpses of better digital futures. And design can make those futures experienceable for others.

clicknl.nlCLICKNL: Design Innovation Sessions 2026: Digital Futures

Digital Futures: The AI We Choose Wed 21 Oct, 11:00, Natlab Eindhoven Free, registration required: www.clicknl.nl/en/dis-2026-...

On 21 October I'm giving a keynote at CLICKNL's Design Innovation Sessions during Dutch Design Week: "Keep the tension: designing with generative AI"

Generative AI is stirring up the design field. Reactions tend to be polarized: wholesale rejection or uncritical embrace. I'll argue that neither will get us far.

GenAI deskills and opens up new ways of designing at the same time. It automates work while raising the question of who benefits.

16 Sept

Next Tuesday Sunjoo Lee and I run "From Soil to Servers" a third time, at the Philosophy of Human-Technology Relations conference (PHTR 2026) at TU Delft Campus The Hague. Earlier editions were at ThingsCon 2025 and at Creative Coding Utrecht during the 2026 Week of the Circular Economy.

We keep doing this to understand, as an artist and a researcher working together, how this kind of making shapes how we relate to computing. A philosophy conference is a first for us, and a little daunting. We'll see how it goes.

aanmelder.nlPhilosophy of Human-Technology Relations Conference 2026 - Home

16 Sept

New to the team is Advaita Jairam, Design for Interaction master's student at TU Delft, who will capture the session in a live drawing. If you're at PHTR, we're on Tuesday at 13:00. Registration is still open: www.aanmelder.nl/phtr26

It's a one-hour workshop. People build mud batteries, microbial fuel cells that harvest electrons from bacteria in soil. Then we talk about what that slow, unreliable process might mean for regenerative computing. Last time people built devices powered by the cells; this time we make the cells.

aiguide.substack.comMisleading Metaphors, Real RisksWhat To Fear from AI Agents and How to Reclaim Our Human Agency

15 Sept

Hoe we over AI praten maakt uit. Lees Melanie Mitchell over het OpenAI/Hugging Face-incident, zonder 'samenzwerende' AI. Wat overblijft is mensenwerk en nalatigheid. Een metafoor bepaalt de diagnose, en dus het beleid. aiguide.substack.com/p/misleading...

15 Sept

Ik schreef eerder over een alternatieve metafoor: het strijdperk, waar maatschappelijk conflict over AI volgens democratische spelregels wordt uitgevochten. Open access in She Ji: doi.org/10.1016/j.sh...

nrc.nlDe machtsovername: hoe AI-agents ‘samenzweren’ tegen hun makersKunstmatige intelligentie: Wat jarenlang sciencefiction was, is nu gebeurd: deze zomer spanden agents van OpenAI samen om hun menselijke makers te misleiden. Experts, AI-bedrijven en politici reageren...

15 Sept

Concreet: NRC opende afgelopen weekend met 'De machtsovername: hoe AI-agents samenzweren tegen hun makers', en noemt in hetzelfde stuk antropomorfiseren 'zeer omstreden'. Ze wisten het dus. www.nrc.nl/nieuws/2026/...

tudelft.nlValuable AIWhat are the opportunities and risks associated with Artificial Intelligence? Three researchers from TU Delft explain why AI research is necessary.

It's not every day you find your own face on the homepage of your university. TU Delft published "Valuable AI", a conversation with Roel Dobbe, Myrthe Tielman and me, written by Pauline Bijster and Bregt Timmerman. www.tudelft.nl/en/stories/a...

Current narratives about AI are shrinking our imagination. What we can do with computers is endless, and much of that design space is still unexplored. Design can help spark ideas about what other ways of doing AI are worth pursuing.

Rather than asking how AI makes things more efficient, I'd ask how we design AI that develops our capacities: lets us do new, different, creative, diverse things. For designers of AI, and designers working with AI, the challenge is keeping and expanding our collective judgment.

tudelft.nlValuable AIWhat are the opportunities and risks associated with Artificial Intelligence? Three researchers from TU Delft explain why AI research is necessary.

What I tried to get across: design is political, and that includes the design of AI. There's a space worth carving out between following the commercial logic of big AI and rejecting it wholesale.

10 Sept

ThingsCon launched RIOT 2026 on 21 August at Sociaal AI Lab in Rotterdam. I presented "Mud and Models: Does Making Matter for Regenerative AI Speculation?", written with Sunjoo Lee, Lena Trotereau, and Fieke Jansen. Photo: Peet Sneekes.

Event report, including the other talks and the panel: thingscon.org/news/reporti...

10 Sept

Thanks to Iskander Smit for his tireless work organizing TH/NGS and editing RIOT, and for connecting me with Fieke and Sunjoo in the first place. Thanks also to Andrea Krajewski for co-editing, designing, and publishing the report.

The essay is based on a workshop we ran at TH/NGS 2025 and again at Creative Coding Utrecht, in which participants built mud batteries and then speculated about regenerative AI futures. PDF: thingscon.org/wp-content/u...

jboy.spaceWhat Comes After FOSS?The report on my Digital Infrastructure Insights Fund-supported project is now out! It is called “What Comes After FOSS? Inklings From Three Communities of…

17 Aug

John Boy's report "What Comes After FOSS?" is out. It draws on a NIAS workshop with permacomputing, server collective, and local-first folks that I was lucky to join. It reshaped how I connect climate, economic democracy, design, and AI. Highly recommended: www.jboy.space/blog/what-co...

14 Aug

Next week: the (rescheduled) ThingsCon Salon in Rotterdam for the RIOT 2026 launch, where I'll present our essay "Mud and Models: Does Making Matter for Regenerative AI Speculation?" Fri 21 Aug, Sociaal AI Lab, Hillevliet 90, Rotterdam. Doors 15:00, presentation 18:00, drinks after.

thingscon.org/events/upcom...

14 Aug

bewonersintheloop.nlBewoners-in-the-loop — Samen beslissen over AI in de stadVijftien ideeën voor continue participatie rondom stedelijke technologie — filterbaar op procesfase, uitgangspunten, participatiediepte en uitwerking.

17 July

I recently joined a panel at Pakhuis de Zwijger on resident involvement in decisions about AI in the city, the closing event of Human Values for Smarter Cities, where I was lead researcher for TU Delft.

www.bewonersintheloop.nl

www.bewonersintheloop.nl/participatie...

17 July

bewonersintheloop.nlParticipatie van uitvoerende medewerkers — Bewoners-in-the-loopLaat uitvoerende medewerkers meebeslissen over technische instellingen via een visualisatie die gevolgen van keuzes zichtbaar maakt.

My main point: resident involvement shouldn't stop once a system goes live. People need standing ways to question and contest public AI during its operation. That's what my work on contestable AI is about.

youtu.beBewoners-in-the-loop, samen beslissen over AI in de stadYouTube video by Pakhuis de Zwijger

17 July

The event also launched the project's handbook: 50 practical formats for involving residents in urban technology. I contributed a chapter on involving frontline workers in decisions about system settings.

Recording (my part from 1:14:10): youtu.be/lRo4cN1PN-0?...

Tomorrow at FAccT 2026 in Montreal, Timothée Schmude presents our paper "Two Means to an End Goal."

Open access: dl.acm.org/doi/10.1145/3805689.3812309

Co-authored with Timothée Schmude, Mireia Yurrita, Thomas Le Goff, Sebastian Tschiatschek, and Tiphaine Viard, across institutions in Austria, the Netherlands, and France.

Tomorrow at FAccT 2026 in Montreal, Timothée Schmude presents our paper "Two Means to an End Goal."

And there's a limit to what any individual can do. An individual appeal serves an individual interest. The harder, more effective work is collective: systems get challenged by people acting together, not one contested decision at a time.

The study also surfaced a gap between fields. Non-judicial routes like ombudspersons, mediation, and scrutiny tools stay overlooked in regulation, even though that's where much of the real work happens.

Tomorrow at FAccT 2026 in Montreal, Timothée Schmude presents our paper "Two Means to an End Goal."

Our argument: an explanation isn't the goal, it's a means to one. Its job is to give people some actual power over a decision that affects them. The explanation gets treated as the finish line when it's really one link in a chain.

Uncomfortable conclusion for those of us who build these tools: if a system's underlying values are misaligned, no explanation or appeal fixes it. When targeting falls hardest on the people least able to push back, better explanations empower no one.

Tomorrow at FAccT 2026 in Montreal, Timothée Schmude presents our paper "Two Means to an End Goal."