Peter Dalsgaard

8/8
Curious? Skeptical? Inspired?

Read the preprint here: drive.google.com/file/d/1QOBj...

Or reach out if you want to discuss how ambiguity and tension might be designed into GenAI tools—not just engineered out of them.

#ECCE2025 #GenAI #Creativity #HCI #DesignResearch #dkforsk

7/8
This work contributes to ongoing HCI and creativity research by shifting the conversation:
From “How do we reduce surprise?”
To “What kinds of surprise are worth staying with?”

6/8
The takeaway:
Rather than designing GenAI tools for perfect alignment, we should explore how to design for productive friction—interactions that provoke users to think, reflect, and reframe.

5/8
The theoretical part builds on:

Cognitive dissonance (Festinger)

Frame conflict (Goffman, Schön)

Predictive error (Clark)

Together, they explain how misalignment becomes a driver of insight—not just a disruption.

4/8
These aren’t just abstract ideas. I analyze five examples of how this plays out across:
✍️ Writing (ChatGPT)
🎨 Visual design (Midjourney)
🎵 Music (AIVA)
💻 Coding (Copilot)
🧰 Product prototyping (Fusion 360)

Each case shows how “getting it wrong” can reveal new possibilities.

3/8
The paper introduces two core ideas:

Creative ambiguity: when AI output is open-ended, contradictory, or off-mark in suggestive ways

Cognitive tension: when that output clashes with your expectations, creating a moment of interpretive pressure

2/8
We often treat GenAI hallucinations or unexpected outputs as bugs. But in creative work, they can be features: points of friction that prompt reinterpretation, redirection, and conceptual expansion.

🧵 1/8
When AI gets it “wrong” – and that’s exactly the point.

Thrilled to share that my paper Creative Ambiguity and Cognitive Tension in Generative AI Tools has been accepted for ECCE 2025. It’s about how misalignment between human intent and AI output can fuel creativity.

@au.dk #hci #creativity

Når #GenAI som fx Google AI Overview skaber nyhedsoverblikket, ændres spillereglerne for #dkmedier & indholdsproducenter. Færre clicks > færre indtægter.
Måske er tiden kommet til at udvikle danske AI-overbliksservices på vores præmisser? #dkmedier #dkforsk @au.dk

www.linkedin.com/pulse/n%C3%A...

🎨🧠 How do creatives really capture ideas on the fly?

New TOCHI paper finds idea capture for creative practitioners is fast, messy, and wildly different across creative domains. ⚡Yet all prioritize capturing ideas quickly, even if it disrupts order.

📄 Read the full paper: doi.org/10.1145/3727...

Proceedings of the 2025 CHI Conference on Human Factors in Computing Systems | ACM Conferences CHI: Conference on Human Factors in Computing Systems

Proceedings of the 2025 CHI Conference on Human Factors in Computing Systems #chi2025 are now online! dl.acm.org/doi/proceedi...

Great work, thx 🙏

Great article and presentation, thx 🙏

Search LibGen, the Pirated-Books Database That Meta Used to Train AI Millions of books and scientific papers are captured in the collection’s current iteration.

57 af mine publikationer er blandt de tekster, de har hentet fra piratsider som LibGen.
Tjek om dine også er med: www.theatlantic.com/technology/a...

The Unbelievable Scale of AI’s Pirated-Books Problem Meta pirated millions of books to train its AI. Search through them here.

Hvis @theatlantic.com har ret, har Meta – med verdens næstrigeste mand, Mark Zuckerberg, i spidsen – brugt ulovligt kopierede bøger og artikler til at træne AI-modeller. Med åbne øjne. Uden at spørge. Uden at kompensere forfatterne. 1/2
www.theatlantic.com/technology/a...
@au.dk #dkforsk #forskpol

a man and a woman are jumping in the air in a hallway with their hands in the air . ALT: a man and a woman are jumping in the air in a hallway with their hands in the air .

Thrilled that our paper "Micro-Phenomenology as a Method for Studying User Experience in Human-Computer Interaction" is accepted for #chi2025. It is the culmination of extensive cross-disciplinary collaboration with Katrin Heimann, Minke Nouwens and Suneetha Saggurthi that started five years ago.

OpenAI - dem med ChatGPT - benytter sig som mange andre big tech firmaer af lavtlønnet arbejdskraft i udviklingslande, når de skal træne og udvikle deres #AI modeller. Arbejdsforholdene er ofte dybt kritisable, som beskrevet her:

Vuvuzela Removal We’re hearing complaints everywhere about the noisemakers called Vuvuzelas during the world cup. Whether you are a fan of the sport or not, you can appreciate when a fellow hacker gets annoye…

Der var en, der klarede det med vuvuzelaer, da VM i fodbold var i Sydafrika. Måske vedkommende kan gøre verden endnu en tjeneste?
hackaday.com/2010/06/15/v...

Photo of a slide with the title of a talk and a quote that says As we design, design changes us

Photo of a speaker presenting at a lectern with a slide in the background

“Design to be our best human selves with mediocre AI” - insightful reflection on the role of design and participation in an increasingly automated world - closing keynote by Margot Brereton. #ozchi24

Is it Bad to leave Twitter? No. Here are 7+ years of insights from my lab’s research that explain why.

Featuring work w/ @williambrady.bsky.social @killianmcloughlin.bsky.social

🧵

You can download the paper here: tinyurl.com/ozchi24

As more and more of our favorite #HCI venues join Bluesky, I created a starter pack with the ones I found @chi.acm.org @acm-dis.bsky.social @chiwork.bsky.social @mobilehci.bsky.social @ieeevis.bsky.social @acm.org @acm-cscw.bsky.social @ozchi-conf.bsky.social #CHI2025

go.bsky.app/GN935Hq

Massively enjoyed this morning’s @ozchi.bsky.social keynote by Paul Dourish on what comes after #HCI.
On why #AI is not the answer (and maybe even the antithesis), and a call to consider why we have strayed from the ambitions of eg Engelbart (to augment human intellect) & Nelson (to save the world!)

Feel free to put me on the list

Participants expressed curiosity about the AI tools and appreciated their intuitive alignment with existing workflows, but rarely adopted AI-generated ideas directly, using them more for inspiration.

Our findings show that AI-supported teams generated more ideas and achieved higher novelty but at the expense of appropriateness, revealing a trade-off in creative outcomes. AI tools were particularly useful for fostering divergent thinking during moments of creative impasse. >

In the paper, we explore how generative AI software impacts collaborative creativity, specifically idea generation. In a controlled study with our custom-made collaborative tool, 36 participants worked in pairs, with half using AI features to support sense-making, divergent and convergent thinking >

I'm in Brisbane for the OzCHi’24 conference, @ozchi-conf.bsky.social, to present the paper “Exploring the Impact of AI Features on Collaborative Creativity”, co-authored with
@jonasfrich.bsky.social, Jens Emil Grønbæk & Marcel Borowski from @aarhusuni.bsky.social.
Summary + Key points: >

Blue Mountains, Australien. Er her til konference mm.

Du er velkommen til at smide mig på listen.