Rationalizing sensorimotor deficits in autism spectrum disorder

Autism spectrum disorder (ASD) is a neurodevelopmental condition marked by socio-communicative, sensory, and motor anomalies. Recent studies have utilized probabilistic models to explain the diverse symptoms observed in ASD. However, these models have mostly concentrated on simple tasks that separate action from perception, and they assume that both neurotypical and ASD individuals aim to optimize the same cost function.

To address this, researchers conducted a study where both ASD patients and neurotypical (NT) individuals navigated to and stopped at targets presented briefly in a virtual reality (VR) environment, simulating a closed action-perception loop.

The researchers found that while the overall performance on the task did not differentiate between ASD and NT groups, data-rich virtual trajectories did. This suggests that ASD and control participants employ distinct navigational strategies and/or cost functions. To interpret the sensorimotor differences in ASD, the researchers applied the Inverse Rational Control (IRC) framework.

IRC does not presume that agents are optimal but rather that they aim to maximize reward based on an internal model of the task, which may not align with the actual task and environment.

The IRC analysis revealed sensory and motor differences between ASD and NT participants. Sensory differences were consistent with Bayesian inference, where altered weighting of sensory evidence was observed. Motor differences manifested as more frequent and abrupt actions in ASD, which were attributed to a reduced subjective action cost in ASD.

In conclusion, this study demonstrates the value of model-based characterization of psychopathology in ASD and suggests that computational research on ASD should consider not only perceptual anomalies but also subjective costs.