New: “AI Agents in the Lab”
A concept paper on how agent systems can support real scientific work in the lab.
medium.com/@valentin.gu...

Thrilled to announce I'll be giving a lecture on the formation of beliefs and trust in the world during the Ecole douteuse.

Excited to announce I'll be teaching NACS 645: Cognitive Science next semester at UMD! This PhD-level course covers computations, representations, decision-making, and more. Can't wait to get started with our amazing grad students!

I know I will.

Still ran into the usual [hallucinations, random loops, repetitive tool use, slow responses] on my pc. Server access is still key.
Once resolved, I can see small academic teams using these to level the playing field in the “publish or perish” environment.

AI agents to design [tools, roles, workflows, agentic collab]; multi-agent systems to run them. No doubt individuals/companies will rely on this kind of setup - once we downsize the need for massive parameters and RAM.

Been experimenting with AI agent systems lately (smolagents, llamaindex, langgraph, crewAI). The ecosystem is getting much more mature, tools for easy implementation are there (massive improvements over 2023 LangChain/Auto-GPT), ability to run python is insane.

Comment la confiance en soi nous expose à la #désinformation. C’est la conclusion que tirent les chercheurs Jean-Claude Dreher et @guigonv.bsky.social. Pour la #Semainedelapresse, pensons contre nous et vérifions si nos certitudes ne sont pas écrans de fumée. www.ouest-france.fr/leditionduso...

Coucou @cortecs.bsky.social !

J'ai eu le plaisir d'écrire un essai sur les enjeux épistémiques des chambres d'écho et leur confusion avec la sphère publique pour le collectif Cortecs (Collectif de recherche transdisciplinaire esprit critique & sciences).

Lien:
cortecs.org/informations...

Excited to share our new preprint on longitudinal changes in infant attention networks and their link to fear! My first paper in Courtney Filippi - Infant Brain & Behavior Lab in New York – grateful for the support! 🧠💫

Check it out here:
www.medrxiv.org/content/10.1...

Pour mes collègues français, un résumé de l'étude disponible ici: www.insb.cnrs.fr/fr/cnrsinfo/...

I am thrilled to share this collaborative work with Marie Claire Villeval and @jc-dreher.bsky.social, which sheds light on the mechanisms shaping how we navigate today’s complex information landscape.

As a result, when faced with ambiguous news, individuals often fail to recognize and address gaps in their knowledge, hindering their ability to evaluate subsequent information effectively.

Ambiguity—defined by the precision of the news content and its capacity to polarize opinions (perceived dissensus)—plays a critical role in shaping veracity assessment. It leads individuals to confidently misinterpret both true and fake news, making them vulnerable to systematic errors.

Our study reveals that metacognition, not beliefs, drives the demand for information. However, individuals’ metacognition is often uncalibrated: their confidence in discerning true from fake news does not reliably reflect their actual accuracy.

📣 New paper! 📣

How do we assess the veracity of ambiguous news, and does metacognition guide our decisions to seek further information?

Read the full article: www.nature.com/articles/s44...

Key findings below 🧵

#Psychology #CognitiveScience #Metacognition #DecisionMaking #FakeNews #Information

After over 10 years of barely using Twitter, I’m moving to Bluesky! Looking forward to sharing ideas, connecting with others, and seeing where this platform takes me. I’ll mostly post about research, science communication, et tout ce qui pique ma curiosité.
À bientôt!