Disentangling signal and noise in autism spectrum disorder

Abstract

a b s t r a c tPredictive coding has recently been welcomed as a fruitful framework to understand autism spectrum disorder. Starting from an account centered on deficient differential weighting of prediction errors (based in so-called precision estimation), we illustrate that individuals with autism have particular difficulties with separating signal from noise, across different tasks. Specifically, we discuss how deficient precision-setting is detrimental for learning in unstable environments, for context-dependent assignment of salience to inputs, and for robustness in perception, as illustrated in coherent motion paradigms.

Increased variability but intact integration during visual navigation in Autism Spectrum Disorder

“…An alternative, but not necessarily mutually exclusive explanation, is that the increased variability in path trajectories and end points does not reflect a heightened sensitivity to noise at the stage of encoding, but rather an increase volatility of beliefs. Specifically, it has recently been proposed that the core abnormality in ASD may reside in perceptual aberrations due to an imbalance in predictive coding (Haker et al, 2016;Van de Cruys et al, 2016). According to this hypothesis, individuals with ASD overestimate their sensory prediction errors, such that the world appears more volatile than it is.…”

“…In response to its pervasiveness, researchers have recently turned their attention to computational and normative tools attempting to identify canonical computations underlying ASD symptomatology (e.g., Robertson & Baron-Cohen, 2018;Rosenberg et al, 2015). A promising candidate family is that of probabilistic inference (Doya et al, 2007), and indeed a large number of Bayesian accounts of ASD have recently been put forward -positing an anomaly in the strength of Bayesian priors (Pellicano and Burr, 2012;Friston et al, 2013), the abnormal updating of these priors Lieder et al, 2019), the aberrant precision in sensory representations (Brock, 2012;Lawson et al, 2015;Zaidel et al, 2015;Karvelis et al, 2018), and the atypical weighting of sensory prediction error (Friston et al, 2013;Haker et al, 2016;van de Cruys et al, 2016).…”

Gravity as a Strong Prior: Implications for Perception and Action

“…One hypothesized reason is that the latter relies more on predictive mechanisms than the former. Last but not least, the study of clinical populations could prove advantageous: if the internal model of gravity is indeed to be considered a prior and the Weak Prior Hypothesis for Autism (Pellicano and Burr, 2012) or similar theories (Lawson et al, 2014; Van de Cruys et al, 2017) hold, persons with ASD should perform better at interception tasks under earth-discrepant gravity conditions. Likewise, Schizophrenia has been linked to abnormalities in the integration of sensory input and prior beliefs and prediction errors (Fletcher and Frith, 2009).…”

The Glasgow Sensory Questionnaire: Validation of a French Language Version and Refinement of Sensory Profiles of People with High Autism-Spectrum Quotient

“…Bayesian theories could explain the presence of both hypersensitivity and hyposensitivity, as sensory reactivity would depend on the context (and particularly, of its predictability). The context should help disentangling sensory relevant information from irrelevant noise, and therefore adapt the level of precision of the sensory input (Van de Cruys et al 2016). Hence, depending on the context, noisy sensory inputs would lead to hyposensitivity, while precise sensory inputs would lead to hypersensitivity.…”