Jess Thompson

The lesson for all the students out there is that science is a community project. Most of us make individually small contributions to this project. Success is measured at the collective level. Many of our professional (and personal) dysfunctions could be fixed by more fully embracing this view.

We audited one of the most critical pieces of modern AI alignment: reward models. We find consistent and persistent biases and trace them back to the pretraining stage, challenging the premises of common approaches to alignment based on finetuning on human preferences. Accepted at #ICLR2026

Reward models (RMs) are supposed to represent human values. But RMs are NOT blank slates – they inherit measurable biases from their base models that stubbornly persist through preference training. #ICLR2026 🧵

Oxford University did not suspend professor investigated for rape - Cherwell An emeritus professor at the Saïd Business School, who is also a fellow of an Oxford college, was investigated for rape by Thames Valley Police (TVP).

"the academic alerted the University of the situation, and asked for protection as well as restrictive measures to be implemented. The University did not formally investigate the professor, and did not restrict his access to premises or remove his emeritus status." cherwell.org/2026/02/01/o...

Person: say, i am alive. Computer: I am alive. Person: oh my god.

once again being driven insane by ML conference submissions

Bluesky > X for science

an envelope containing ballot papers being put into a royal mail mailbox

I'm voting YES on my UCU ballot for industrial action
• to tell UCEA that I want negotiation, not imposition.
• to call for the broken market system to change.
• to call for decent pay, conditions, and job security for all university staff.

#ucu @oxforducu.bsky.social @ucu.org.uk

When does new learning interfere with existing knowledge in people and ANNs? Great to have this out today in @nathumbehav.nature.com

Work with @summerfieldlab.bsky.social, @tsonj.bsky.social, Lukas Braun and Jan Grohn
www.nature.com/articles/s41...

Super happy to see this review out! We ask why people are so reluctant to abandon goals and how this commitment could be understood computationally. Work with Jill O'Reilly & @yaelniv.bsky.social

I'll be in Montreal October 26-27! Let me know if anyone wants to meet up!

the post office is a public service. it doesn’t need to make money. public transit doesn’t need to make money. the library doesn’t need to make money. some things exist for the public good and we desperately need lawmakers to stop thinking about them in terms of capitalism. these are not businesses.

I had an illuminating interaction with a colleague who said that they try to find time to do a few paper reviews per year. It made me realize that different researchers have widely varying expectations about reviewing, probably because there are no public norms.

New at the university: Prof. Weber! Osnabrück University welcomes Prof. Dr. Lilian Weber! Since August 15, she has held the professorship "Cognitive Modeling" at the School of Human Sciences. A warm welcome!

In other news, I am now a ✨ COXI ✨ (cognitive scientist in Osnabrück, Germany)

www.uni-osnabrueck.de/en/campus-li...

I'll be looking for PhD students & postdocs later this year, so watch this space if combining cognitive modelling, EEG & tFUS to study flexible cognition sounds exciting to you 🚀

I’m recruiting committee members for the Technical Program Committee at #CCN2026.

Please apply if you want to help make submission, review & selection of contributed work (Extended Abstracts & Proceedings) more useful for everyone! 🌐

Helps to have: programming/communications/editorial experience.

Expression of Interest in Serving as a PsyArXiv Moderator As you might have heard, PsyArXiv is having some issues with an increase in low-quality submissions, ranging from AI generated manuscripts to inflate citation metrics, incoherent or nonsensical docume...

PsyArXiv is seeking new moderators to help combat an increase in AI submissions! If you've ever posted a preprint to PsyArXiv, please consider joining. Minimum commitment 1h/month, there's a training session this Monday @ 1pm ET. More info here: forms.gle/9LB1rEtxHAeZ... #PsychSciSky

Forms of explanation and understanding for neuroscience and artificial intelligence | Journal of Neurophysiology | American Physiological Society Much of the controversy evoked by the use of deep neural networks as models of biological neural systems amount to debates over what constitutes scientific progress in neuroscience. To discuss what co...

For no reason, I thought now might be a good moment to reshare my 2021 article arguing that the #CCN and #NeuroAI communities can avoid reinventing the wheel by building on existing philosophy of science about scientific progress and explanation.

journals.physiology.org/doi/full/10....

#CCN2025

Check out this great summary of our satellite event on science for social good by @georgiaturner.bsky.social! ♥️

I also wrote up a transcript of my opening remarks thompsonj.github.io/CCN_for_soci...

We are the gardeners of our scientific ecosystem. What will we grow?

Blown away by Ellie Pavlick's talk at #CCN2025 about compositionality in transformers. More work like this at CCN please.

@davidpoeppel.bsky.social points out the elephant in the conference: whichever direction of the planet we look, things are bad.

Einstein, chomsky and arendt point out to us that silence is not neutral

👏 #CCN2025

Inspiring quotes about the state of the world and our role in it as scientists at the beginning of David Poeppel's keynote at #CCN2025

RNNs making decisions The goal of this CCN workshop is to host a scientific discussion on the study of decision-making using recurrent neural networks. The selected speakers will cover all aspects of decision-making, from ...

If you are planning to attend #CCN2025 in Amsterdam next week and are interested in recurrent network models of decision making, please join our Satellite Workshop on Monday at 3pm. We'll have @summerfieldlab.bsky.social , @veronicachelu.bsky.social and many more! sites.google.com/view/rnnsmak...

Waiting for the train to Amsterdam for #CCN2025. Looking forward to catching up with old friends and making new connections. Hit me up if you want to chat about (modeling) collective intelligence, philosophy of science, communication, care, reward models, or the epistemic risks of scientific AI.

"If we want a psychological science that continues to serve our deepest interests, we must be wary of the temptation to trade the difficult quest for humanistic understanding for an expedient facsimile which merely looks close enough, on average."

New preprint by William D'Alessandro and myself:

The promise and peril of AI surrogacy in psychological research

osf.io/preprints/ps...

In this world on fire, with misinformation, science and democracy under threat, how can scientists be a force for social good?

This Monday in Amsterdam, join @weijima.bsky.social @tsonj.bsky.social Ili Ma + me for an interactive workshop as part of #CCN2025

anneurai.net/2025/05/05/c...

Cartoon image of me looking at a map, with a stadium behind me and a hotel and ferris wheel across the river in the background. I am thinking about going to the ferris wheel

My first PhD paper - with @lhuntneuro.bsky.social and @summerfieldlab.bsky.social - is now out in @plosbiology.org! We ask: how do humans (and deep neural networks) navigate flexibly even in unfamiliar environments, such as a new city? Link: plos.io/45uSwNm 🧵 (1/6)

UCL – University College London UCL is consistently ranked as one of the top ten universities in the world (QS World University Rankings 2010-2022) and is No.2 in the UK for research power (Research Excellence Framework 2021).

📢 Job announcement: Two (!) 3-year postdoc jobs in our lab at UCL 📢

🧠💫🔊 We are looking for postdocs interested in the abstract mechanisms underlying social cognition. Modelling, fMRI and non-invasive ultrasound, a new deep-brain stimulation method.

Please RT

www.ucl.ac.uk/work-at-ucl/...

What *is* metascience? Issues, inclusion and future public value - Warren's blog: Publics, politics, science and technology This week, the biennial Metascience Conference came to London’s “Knowledge Quarter” with around 800 participants from a wide range of roles, including researchers, librarians, research funders, publis...

After two really enjoyable days at #Metascience2025, I wrote about how one 🌶️ moment encapsulated the controversy over metascience's issues 📑, inclusivity 🤗and future public value 🔮 #STS #AcademicSky

All feedback welcome on here, or by email 😀

warrenpearce.pika.page/posts/what-i...