Stefano Palminteri

Adaptive biases in the wild: Advancing our understanding of the nature of biases - Mind & Society Bias is most often seen as a flaw: people are said to “suffer” from biases and need to be “debiased.” Yet a bias, defined simply as a systematic deviation from a norm or standard, can in principle hav...

"Adaptive biases in the wild: Advancing our understanding of the nature of biases"
The introduction (by Jochen Reb, Natalia Karelaia & Tomás Lejarraga )to the "Mind and Society" special issue on "adaptive biases" I had the pleasure to contribute to
link.springer.com/article/10.1...

Our latest preprint where we show (among other things!) that the main effect of complete feedback information is increase risk (not performance) in experience-based show. We also show that the description experience gap is not due to sampling issue
osf.io/preprints/ps...

The winner’s curse — Behavioral economics anomalies, then and now.
The presentation of Richard Thaler's (Chicago Booth) latest book, followed by a roundtable discussion, organized by the "An integrated approach of economic decisions" project.
www.parisschoolofeconomics.eu/en/events/th...

Also found in the old sci-fi stash recently purchased in Bologna
The plot’s crux is an illustration of the alignment problem (an all-powerful AI with wildly misaligned goals). Basically, the paperclip maximiser has gone rogue.
(but do not expect great writing and depth of reflection)

🚨 New preprint 🚨

Are reinforcement learning models complete accounts of decisions from experience if they ignore explicit memory?

In this new preprint, we show that people indeed form robust explicit memory representations that flexibly guide later decisions.

🔗 Preprint: doi.org/10.1101/2025...

🇪🇺 I am a bit late for this, but is important:

R.I.P. Sofia Corradi (1934 – 2025), the beautiful mind behind the ERASMUS project, one of the most successful and beloved EU programme.

It has changed the life (and mind) of ~15 million Europeans (including mine).

en.wikipedia.org/wiki/Sofia_C...

🇪🇺 I am a bit late for this, but is important:

R.I.P. Sofia Corradi (1934 – 2025), the beautiful mind behind the ERASMUS project, one of the most successful and beloved EU programme.

It has changed the life (and mind) of ~15 million Europeans (including mine).

en.wikipedia.org/wiki/Sofia_C...

Just read this old-school sci-fi gem I found in a vintage bookstore in Bologna, where a Practical Philosopher Corps is deployed across the galaxy to assess sentience and cognition in alien species.
I guess the dream job for @birchlse.bsky.social @petergs.bsky.social

At a time when prominent thinkers like @anilseth.bsky.social Seth and Ned Block advocate a "strategic withdrawal" toward biologism in considering consciousness beyond the human case, our contrarian proposal is a methodological behaviourist computationalism.
www.linkedin.com/posts/stefan...

🚨 New publication: How to improve conceptual clarity in psychological science?

Thrilled to see this article with @ruimata.bsky.social out. We discuss how LLMs can be leveraged to map, clarify, and generate psychological measures and constructs.

Open access article: doi.org/10.1177/0963...

I think this is what we would have observed in Germain's and Constance's paper respectively if decay were true

Very thought-provoking post by @prakhargodara.bsky.social. Is confirmation bias/positivity bias a statistical "ghost" of model specification? Specifically not including temporally decaying learning rates? The evidence suggests this is not the case and here is why (1/n)

Thanks for the pointer Vinny!

Exhibit #3: back to "stable" tasks, @constancedestais.bsky.social conditioned learning rates on confidence over time, and show that the asymmetry is still there. Indeed, it increases over time. Note that the model structure would have perfectly allowed for a "symmetric decaying" pattern 4/n

Exhibit #2: learning rate bias has been reported (by us and other groups) in volatile tasks or conditions where, normatively, learning rates should not decay, and, perhaps more importantly, empirically, they indeed do not decay - if not, accuracy would not be above chance 3/n
doi.org/10.1016/j.ti...

Exhibit #1: I was aware of this possibility since our first paper on the topic, and this is why we fitted separate learning rates in the first half and the second half of the learning phase. We found no evidence of decay and robust bias in both phases 2/n

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

Very thought-provoking post by @prakhargodara.bsky.social. Is confirmation bias/positivity bias a statistical "ghost" of model specification? Specifically not including temporally decaying learning rates? The evidence suggests this is not the case and here is why (1/n)

The associated online tool, however definitely nerdy, is addictive. Many kudos to @dirkwulff.bsky.social and co for setting this up and opening it to the community!

It was a real pleasure to be involved in the meta-scientific collaboration about the (historical, semantic and to some extent sociological) structure of the behavioral reinforcement learning field. Check @annaithoma.bsky.social thread below for more info!

2/2 this is the second paper by Vidal and Moran.

osf.io/preprints/ps...

Something similar has been going on concerning learning bias versus perseveration. Initially fuelled by Kentaro Katahira and co, and more recently by Juan Vidal and Rani Moran. (see the discussion, about what counts as a plausible computation)
See this 1/2 and...
psycnet.apa.org/buy/2023-182...

I think that is where I disagree with the author. Asymmetry is apparent, only if we assume people are Bayesian (which I do not believe is the case). But then I like that a "rational" analysis (Bays) of the task leads to the emergence of asymmetry, which may explains why the it is used by the brain

I guess the missing link here is "However, we find that even if the agent updates its belief via, arguably objective, Bayesian inference, fitting the above model demonstrates both the biases". I working under the assumption that the Bayes solution is understood as normative given the task here

If you want to know more about the reinforcement learning biases framework, I summarised it here:

www.researchgate.net/publication/...

Apparent learning biases emerge from optimal inference: Insights from master equation analysis | PNAS Recent studies [S. Palminteri, G. Lefebvre, E. J. Kilford, S. J. Blakemore, PLoS Comput. Biol. 13, e1005684 (2017); G. Lefebvre, M. Lebreton, F. Me...

Very cool study showing that "apparent" asymmetric update in reinforcement learning can emerge from Bayes optimal principles by Prakhar Godara in @pnas.org

www.pnas.org/doi/abs/10.1...

I am very humbled that during the past years so many smart people took seriously our research questions and results to push forward our understanding.
On the specific subject matter (bias or optimal) I am still persuaded that it is a bias, that just happens to be generally optimal 😉

In-context learning agents are asymmetric belief updaters We study the in-context learning dynamics of large language models (LLMs) using three instrumental learning tasks adapted from cognitive psychology. We find that LLMs update their beliefs in an asymme...

Also quite coherent with what has been found by @akjagadish.bsky.social and co that asymmetric update emerges as an optimal solution in a neural-network meta-reinforcement learning agent arxiv.org/abs/2402.03969

It resonates perfectly well with what we (with @isabellehoxha.bsky.social and Léo Sperber) have shown, that the bias emerges as an evolutionary stable solution in multi-agent simulations

www.pnas.org/doi/abs/10.1...

Apparent learning biases emerge from optimal inference: Insights from master equation analysis | PNAS Recent studies [S. Palminteri, G. Lefebvre, E. J. Kilford, S. J. Blakemore, PLoS Comput. Biol. 13, e1005684 (2017); G. Lefebvre, M. Lebreton, F. Me...

Very cool study showing that "apparent" asymmetric update in reinforcement learning can emerge from Bayes optimal principles by Prakhar Godara in @pnas.org

www.pnas.org/doi/abs/10.1...

Thought experiments such as the Blockhead and Super-Super Spartans are often taken as “definitive” arguments against behavior-based inference of cognitive processes.
In our review -with @thecharleywu.bsky.social- we argue they may not be as definitive as originally thought.