Can LLMs do science, how would we know, and how would they know? A lab note about David Louapre’s ( @dlouapre ) Eleusis benchmark for scientific reasoning, from a cognitive psychologist's perspective. thomaschazelle.fr/en/how-psychop…
English
Thomas Chazelle
122 posts
Thomas Chazelle
@Thomas_Chazelle
PhD in cognitive psychology, Postdoc at @durham_uni / I (sporadically) share my work in 🇬🇧 & complain about pseudoscience in 🇫🇷
Katılım Ekim 2020
191 Takip Edilen163 Takipçiler
Thomas Chazelle retweetledi
GPT 5.2 drains your time being overcautious.
Grok 4 wastes resources being reckless.
This is what we found benchmarking 16 AI models on Eleusis: a game where you form hypotheses, test them, and refine theories, just like real science 🧪♠️
New video on @huggingface channel ⬇️
English
New lab note: Comparing multisensory integration across different sensory cues is tricky because the size of the multisensory benefits depends on the variance of individual cues. How can cue normalisation solve this problem? thomaschazelle.fr/en/cue-normali…
English
Faut que quelqu'un explique à notre "chère consœur" que le centenaire de 1933 c'est dans quelques années, pas besoin de prendre autant d'avance.
Florence Bergeaud-Blackler 🎓@FBBlackler
Et derrière LFI, il y a les fréristes qui infiltrent les loges maçonniques depuis longtemps.
Français
un sondage comparant "musulmans" et "autres", comme si catégoriser ensemble évangélistes, néopaïens et bouddhistes avait du sens. Pas surprenant : les pseudosciences accompagnent les préjugés de leur époque. Avec l'Ifop, la sondologie ne fait pas exception. Poubelle !
Français
L'Ifop (le sérieux de la voyance, le racisme en plus) poursuit maintenant en justice quiconque ose remettre en question ses enquêtes pleines de préjugés. Peut-on encore critiquer leurs dérives ? Et en l'occurrence, (1/2)
bfmtv.com/police-justice…
Français
Just read Paul Bloom's criticism of (some of) developmental psychology and wanted to share my thoughts. Tldr: We do need more data to find out, in the future, whether we were trying to fit cylinders (data points) in square holes (bad theories) all along thomaschazelle.fr/en/data-matter…
English
Thomas Chazelle retweetledi
Now out in @NatureNeuro
What happens to the brain’s body map when a body-part is removed?
Scanning patients before and up to 5 yrs after undergoing an arm amputation, we discovered the brain’s body map is strikingly preserved despite amputation
nature.com/articles/s4159…
🧵1/18
English
What a lovely 3 days in Utrecht for #BRNet2025. Thanks everyone for the thought-provoking talks and chats. I think the computational modelling theme for this year was spot-on, congrats to the organisers @bodyrepnetwork ! Looking forward to seeing some of you again at #IMRF !
English
Brilliant workshop, imho definitely a/the direction our field should move towards!
English
1st day at #BRNet2025 - I thought my nerdiness was sated with the computational modelling workshop, until Pieter Medendorp mentioned the El Greco fallacy 🥹
English
Je veux pas dire mais je l'avais call (et c'est une mauvaise idée) thomaschazelle.fr/enlever-le-dro…
sean@_sn_n
the 2024 presidential election if this was actually in place
Français
Can #BRNet2025 answer the cartographer / tuner question? If you're in Utrecht in the next few days and even if you're not coming, feel free to share and, most of all, comment!
@bodyrepnetwork
thomaschazelle.fr/en/the-cartogr…
English
Some healthy skepticism about the "Your brain on ChatGPT" paper you may have heard of these days.
Ben@BenShindel
Okay I read the MIT "Your Brain on ChatGPT" preprint. (well, it's 140 pages long, so I read about 1/3 of it and skimmed the boring parts) Here are my takeaways: 🧵
English
Les L1 psycho : participent aux expériences
Les M2 psycho : les font passer
Les profs de psycho : les publient
omg@ohemjei_
Les S : fabriquent la drogue Les ES : la vendent Les L : la consomment
Français
🇫🇷 Version française : thomaschazelle.fr/weight-labels-…
Français
Based on some of my PhD work that resulted in a paper a few months ago in Psychological Research with Michel Guerraz and Richard Palluel-Germain link.springer.com/article/10.100…
English
Latest lab note on our latest attempt at describing weight-related words with probability distributions thomaschazelle.fr/en/weight-labe…
English