RT @umdnacs: 🧠At NACS Seminar tomorrow, Dr. Dorit Kliemann (@DoritKliemann) will tell us about "Behavioral and Neurobiological Basis of Soc….

Come be our new colleague in Computational Psych/Cog Sci/Neuro at @umdpsyc. More details and apply here:

Inaugural SLD lab posters today at #cpconf2024. Check out poster #6 by Xiaodi (Sara) Hu on heterogeneity in models of social influence on risk-taking & poster #42 by Sean Maulhardt on the effect of uncertainty on temporal credit assignment computations.

So proud of @QianyingWu1 for her amazing work on this project, and honored to be one of her mentors!.

And last but not least, all data and code is up on Github: Do not hesitate to reach out if you have questions! (the end).

If you want to try out the task, a demo of the instructions, practice and 10 example trials is available here: (10/n).

Massive credit also to @Alexandra_Pike, @GuillermoHorga, and anonymous reviewer for their help, suggestions, time and a very enjoyable review process. The review file is available online: (9/n).

Huge thanks to co-authors for endless brainstorming sessions to help make sense of this (messy) data: @QianyingWu1, @seokyoung_min, Weilun Ding, @jeffcockburn, and John O'Doherty. (8/n).

This work constitutes a proof-of-concept that, especially with large-scale online studies, heterogeneity should be embraced rather than ignored, though more work is needed to test and improve classification/clustering approaches and their relevance to psychopathology (7/n).

Finally, we provide some insights into the potential relevance of this classification to transdiagnostic symptom dimensions, with groups differing significantly along autism-related traits and trait anxiety, typically associated in the population (and in our sample). (6/n)

To examine heterogeneity, we classified participants into 5 groups based on which model explained each participant’s data best, rather than assuming a ‘best-model fits all’. We found that groups differed in all key behavioral signatures, thus validating our classification (5/n)

We then performed robust simulations and posterior predictive checks to ensure each model could be appropriately recovered and captured expected behavioral signatures in the data. (4/n)

We fit a set of 5 parsimonious models to the data to capture individual strategies: experiential learning (EL) only, observational learning (OL) only, fixed mixture, dynamic arbitration, and a baseline model containing non-learning heuristics. (3/n)

We first showed (and replicated) that participants’ behavior is a mixture of the two strategies, and varies with changes in uncertainty in the environment, consistent with theories of arbitration. Yet, we found huge individual differences that we wanted to further examine. (2/n)

A bit late on this, but our new paper is out in @NatureComms : In this study, we embraced the noisiness on large-scale online data to characterize heterogeneity in experiential vs observational learning strategies. (1/n).

Any paper format (opinion letter, review, empirical research, etc) is welcome. Don't hesitate to reach out if you have question!.

Truly enjoyed my first @AutismINSAR conference #INSAR2024 - such an inspiring mix of basic science, clinical research and amplifying autistic voices!.If you are interested in the heterogeneity of autism, consider submitting a paper to this upcoming @PNeuroscience special issue 👇.

RT @aniladmello: Excited to be at #CNS2024 in gorgeous Toronto! If you’re also here, be sure to catch our symposium on Tuesday at 1:30 on “….

RT @NASAMoon: Oops I did it again 🤭 . #TotalSolarEclipse

RT @MattRoesch_UMD: Congrats to my grad student Daniela Vazquez for defending yesterday and publishing our ACC opto paper today! @umdnacs @….