Patrick Cooper

Towards a Criteria‐Based Approach to Selecting Human‐AI Interaction Mode Artificial intelligence (AI) tools are now prevalent in many knowledge work industries. As AI becomes more capable and interactive, there is a growing need for guidance on how to employ AI most effec...

🧪🤖Our latest paper: Towards a Criteria-Based Approach to Selecting Human-AI Interaction Mode is now out.

We made a handy rubric to help decisions around adding AI to your work. AI isn’t always the solution but if it is, thinking about how your workflow changes is vital!

dx.doi.org/10.1002/hfm....

I completed my first game jam on Sunday. Two weeks to make a game from scratch around the theme of “replicate.”

If you have s&box (or want to download it) you can check it out: sbox.game/veggiepatty/...

Good luck to all the DECRA applicants, hoping you get some helpful comments today!

Trust in AI is dynamically updated based on users' expectations | Proceedings of the Extended Abstracts of the CHI Conference on Human Factors in Computing Systems

Link to the paper now available: dl.acm.org/doi/10.1145/...

Screenshot of an error message in RStudio that says "R Session Aborted. R encountered a fatal error. The session was terminated." with an icon of a bomb"

Gentle error messages have nothing on R

This is such a great package!

I am I'm this picture and I don't like it

CHI Talk - Descript ​ Deepfakes can be hard to spot, and as the technology that generates these artificial images improves, it may get even harder. So what if we had some help? A decision support tool that helps us iden

No time to read it? Here’s a 3 min video I made to accompany the paper here: share.descript.com/view/7N59sry...

Instead, how much someone trusted the AI greatly influenced the agreement with AI advice. We explored how this developed by modelling how trust evolved over time. We found expectations of the AI’s accuracy changed over time and the violations of these expectations predicted trust in AI.

Participants performed a deepfake detection task with advice from AI presented at different stages of the decision making process. We found the timing of advice had minimal impact on the rate of agreement with the AI or the accuracy of their classifications.

Trust in AI is dynamically updated based on users' expectations DOI: https://doi.org/10.1145/3706599.3719870 CHI EA '25: Extended Abstracts of the CHI Conference on Human Factors in Computing Systems, Yokohama, Japan, April 2025

We’ve just had our work on assessing how trust in AI support develops accepted as late-breaking work at CHI25. DOI is still to go live but you can check it out here in the meantime:
camps.aptaracorp.com/ACM_PMS/PMS/...
(I’ll update the thread when the DOI is live).

🧠 💡 PhD scholarships available! 💡🧠

The CogNeuro group at @marcsinstitute.bsky.social is on the hunt for talented students to work on face and object perception using EEG and neural decoding. Learn more here 👇 and get in touch!

Project 1: shorturl.at/0NwQl
Project 2: shorturl.at/yTdqG

Ever wonder how well frontier models (Claude 3.5 Sonnet, Gemini 1.5 Flash & Pro, GPT-4o, o1-mini & o1-preview) play Atari, chess, or tic-tac-toe?

We present LMAct, an in-context imitation learning benchmark with long multimodal demonstrations (arxiv.org/abs/2412.01441).

🧵 1/N

I’d like to be added please.

Why do I think these instances are interesting? To me, these are not just random instances that LLMs are sometimes wrong (like people), these are *diagnostic* that LLMs do not have a world model and are not “reasoning”, it exposes the basic architecture. Why? 1/n

🚨 If you haven’t seen it yet…

📢 The NIH BioArt Source provides a library of FREE professionally designed illustrations and icons, available for anyone to use. They can be Downloaded in High Definition.

Check it out at bioart.niaid.nih.gov

Does brain activity cause consciousness? A thought experiment The authors of this Essay examine whether action potentials cause consciousness in a three-step thought experiment that assumes technology is advanced enough to fully manipulate our brains.

The craziest paper I have ever done is this thought experiment with Albert Gidon and Matt Larkum.

In the first journal, reviewer 1 recommended that we should not try to publish this; reviewer 2 called it "wacky". Thanks for the motivation: A sequel is coming up!

journals.plos.org/plosbiology/...

Examining the replicability of online experiments selected by a decision market - Nature Human Behaviour This study finds that decision markets can be a useful tool for selecting studies for replication. For a sample of 26 online experiments published in PNAS selected by a decision market, the authors fi...

New in NHB: We systematically replicated 26 of 41 PNAS social science studies that used MTurk. The approach provides a proof-of-concept for using decision markets to select findings to replicate.

Replication effect sizes were 45% of original effect sizes.

www.nature.com/articles/s41...

Anthropic posts something a lot of social scientists studying AI have been thinking - you need to apply basic statistical methods to AI evaluations!

A solid attempt to lay out how to do better tests with basic methodology. https://buff.ly/3OijwX0

I’d love to be added 🙏

Congratulations Dr Jarvis! 🎉🎉

Our latest work, up as a preprint to enjoy.

We gave LLMs tasks adapted from experimental psychology to see how they contribute to teamwork.

We found LLMs were reasonable at monitoring tasks but were poor at tasks requiring planning and strategising.

Frontiers | Users do not trust recommendations from a large language model more than AI-sourced snippets BackgroundThe ability of large language models to generate general purpose natural language represents a significant step forward in creating systems able to...

Our latest paper: "Users do not trust recommendations from a large language model more than AI-sourced snippets” www.frontiersin.org/journals/com...

Looking up some cognitive task analysis figures this morning and kept finding those slides with the blue background and yellow text.

Took me down a rabbit hole about why these things were everywhere. Someone else had the same idea.

en.rattibha.com/thread/12942...

New preprint! Mind the blank: behavioral, experiential, and physiological signatures of absent-mindedness

With Esteban Munoz Musat, @corcorana.bsky.social, Laouen Belloli and Lionel Naccache

biorxiv.org/content/10.1...

1/n

Emerging researcher @bellabower.bsky.social giving a fantastic overview of her work on how room characteristics influence brain and behaviour!

Dan F @danfeuerriegel.bsky.social had “results so nice he found them twice” finding null effects for visual ERPs scaling with probability.

Huw Jarvis suggesting the link between reward prediction errors and vigour is more complicated than first thought! #ACNS2023

Julian Matthews @quining.bsky.social suggesting a reframing of some classic metacognitive effects as motivation related! #ACNS2023