Sovereign 8s.is mirror of standardgalactic/Social-learning-in-autism
  • R 35.4%
  • Stan 35%
  • MATLAB 29.6%
Find a file
Repository files (latest commit first)
Filename Latest commit message Latest commit date
2026-03-30 15:52:11 +02:00
data Add files via upload 2026-03-16 15:59:32 -04:00
hpc Add ppc data 2026-03-30 15:52:11 +02:00
scripts Delete scripts/experiment_2/standard_regression directory 2026-03-26 17:17:19 -04:00
README.md Add ppc folder 2026-03-30 13:53:39 +02:00

Modeling how autistic and non-autistic groups learn about their own and each other’s preferences

This repository contains code and data for the paper: "Modeling how autistic and non-autistic groups learn about their own and each other’s preferences", by Shannon Cahalan, Raphael Perla, Sophia Block, Mikaila Loughlin, Christoph W. Korn, Gabriela Rosenblau, To appear in Nature Mental Health


Abstract

Background

Recent theories propose that altered social understanding in autism stems from a mismatch in social representations and inferential processes between non-autistic and autistic groups. In a cross-sectional design, we examined whether non-autistic and autistic individuals apply different types of social knowledge and learning strategies when inferring preferences of individuals from their own versus a different diagnostic group.

Methods

Experiment 1 compared food and activity self-preferences in large samples of non-autistic adults, non-autistic adolescents, and autistic adolescents. Aggregated group preference structures were tested as knowledge for learning in Experiments 2 and 3. In Experiment 2, non-autistic young adults (18–30 years) learned about non-autistic and autistic adolescents, while Experiment 3 applied this design in autistic adolescents (12–17 years). A computational modeling framework examined how groups integrated prior knowledge during learning.

Results

Autistic adolescents’ preferences showed more group-level variability (Experiment 1). Non-autistic adults and autistic adolescents relied on the same learning strategy to infer the preferences of others. They applied fine-grained knowledge but showed reduced accuracy and efficiency when learning about autistic individuals (Experiments 2 and 3). Variability in autistic traits, including rigidity, in the autistic group scaled with lower learning rates and overall accuracy.

Conclusions

Participants in both groups integrated feedback according to reinforcement learning scaled by fine-grained social knowledge reflecting the preferences of their own diagnostic group. Since autistic adolescents’ self-preferences are more variable, aggregate knowledge may be less predictive. Our novel social learning models captured variability in autistic traits in the autistic sample, highlighting the potential of our social learning framework to specify autistic subgroups.


Repository Structure

Social-learning-in-autism
│
├── data/
    ├──experiment_1/             #cleaned self-preference dataset from Exp.1
        ├──raw_data/             #pre-cleaned raw self-preference output
    ├──experiment_2/
    ├──experiment_3/
├── hpc/                         #High compute scripts not intended for local execution
    ├──experiment_3_modeling_covariates/
        ├──data/                 #processed inputs for experiment 3 covariate modeling
        ├──stan/                 #Stan model used by the experiment 3 covariate workflow
    ├──model_comparison/
        ├──data/                 #processed inputs for model comparison
        ├──stan/                 #Stan models used by the model comparison workflow
    ├──ppc/                      #posterior predictive checks for the winning model
        ├──data/                 
        ├──stan/                 
├── scripts/                     #Analysis scripts that can be run locally
    ├──experiment_1/ #code used for standard regresson analyses for all three experiments
    ├──experiment_2_edit/
    ├──experiment_3/



Acknowledgements

We are grateful to all of the families in SPARK, the SPARK clinical sites, and SPARK staff, and appreciate obtaining access to recruit participants through SPARK Research Match on SFARI Base.

We also thank Dr. Ann Clawson, PhD, ABPP, for advice on which clinical measures to include in our individual differences analysis.

We acknowledge access to computational resources provided by the state of Baden-Württemberg through bwHPC and supported by the German Research Foundation (DFG) via grant INST 35/1597-1 FUGG.

Furthermore, this project was endorsed by the German Center for Mental Health (DZPG).

Contact

scahalan@gwmail.gwu.edu grosenblau@gwu.edu Raphael.Perla@med.uni-heidelberg.de