Investigating Adaptive Learning Strategies Under Uncertainty in Autism Spectrum Disorder

Abstract

As a persistent evolutionary force, uncertainty continuously shape our brains to develop sophisticated mechanisms that account for ambiguity, predict outcomes, and construct world models through learning to guide decision-making. Notably, individuals with Autism Spectrum Disorder (ASD) may process uncertainty differently, often developing rigid routines and repetitive behaviors that potentially serve as coping strategies for managing novel and unpredictable environments. For decades, the cognitive profile of ASD has been conceptualized through various theoretical lenses—from the Theory of Mind deficits to Weak Central Coherence to predictive coding frameworks—converging on fundamentally distinct uncertainty processing mechanisms. Here, we develop an experimental and computational framework to address how uncertainty is processed in the autistic brain. We hypothesize that the autistic brain differentially tracks environmental ambiguity, displaying alternative learning strategies under uncertainty. First, we validate an auditory operant conditioning task where mice learn to associate auditory tones (cues) with licking for reward (actions). By incorporating probabilistic reward delivery and reversing task cue-action associations, we compare uncertainty processing in 16p11.2 deletion mice (an ASD model) and wild-type controls. While both genotypes acquire task associations at comparable rates, 16p11.2 mice display poorer performance after reversal. To identify differences in learning strategies, we performed Markov state analyses and found that 16p11.2 deletion mice rely more heavily on exploitation strategies under familiar rules (increased win-stay behavior) and display perseverative errors after reversal (increased lose-stay behavior). Our findings demonstrate altered learning strategies and uncertainty processing in the autistic brain.