Abstract: Autism Spectrum Disorder (ASD) is a neurodevelopmental condition characterized by various sensory processing abnormalities, particularly in the visual domain. In recent years, Bayesian and predictive coding theories have been widely applied to explain these aberrations in visual processing. These theories propose that the brain constantly generates predictions about incoming sensory information based on prior knowledge and updates these predictions based on the discrepancy between the predicted and actual sensory input. However, the application of Bayesian and predictive coding theories to ASD has generated significant debate within the scientific community. This debate highlights the need for a comprehensive examination of the nuances of these theories and a consolidation of empirical evidence to assess their validity in the context of ASD.
This paper focuses on non-social visual information processing in ASD and presents a detailed analysis of Bayesian and predictive coding theories across three key dimensions: Bayesian inference, predictive coding processes, and predictive coding precision. The study aims to provide a comprehensive understanding of the theoretical nuances and empirical evidence supporting or refuting these theories in the context of visual processing abnormalities in ASD. By examining these dimensions, the paper seeks to shed light on the mechanisms underlying the atypical visual processing observed in individuals with ASD and to evaluate the explanatory power of Bayesian and predictive coding frameworks. The analysis will consider the strengths and limitations of each dimension, assessing their ability to account for the heterogeneity of sensory experiences in ASD and to generate testable predictions. Through this comprehensive approach, the study aims to contribute to the ongoing debate surrounding the application of Bayesian and predictive coding theories to ASD and to provide a foundation for future research in this area.
The hypo-priors and sharper likelihood hypothesis based on Bayesian inference are evaluated for their descriptive insights into visual processing abnormalities in ASD. The descriptive nature of this hypothesis limits its explanatory power and ability to generate testable predictions. The perspectives emphasizing the predictive coding process are examined for their ability to enhance the specificity of visual processing aberrations. These perspectives propose that individuals with ASD have impairments in generating accurate predictions about incoming sensory information and updating these predictions based on prediction errors. While these perspectives provide a more specific account of visual processing abnormalities in ASD, they still fall short of offering a fully explanatory framework that can account for the heterogeneity of sensory experiences in ASD. Finally, hypotheses centered on predictive coding precision are assessed for their theoretical foundations and the need for further refinement and empirical validation. These hypotheses propose that individuals with ASD have an imbalance in the precision of prediction errors, leading to an over-weighting of sensory evidence and a failure to contextualize sensory information. While these hypotheses provide a promising theoretical foundation, they require further refinement of the theoretical details and empirical validation through carefully designed studies.
This comprehensive review of Bayesian and predictive coding theories in understanding visual processing abnormalities in ASD provides a foundation for future research directions. To advance the understanding of predictive coding mechanisms in ASD, future research should adopt a multi-faceted approach. First, researchers should focus on examining predictive processing within specific sensory domains, such as auditory, tactile, or social visual information, to identify the unique characteristics and divergences within each domain. Second, based on the findings from these domain-specific investigations, researchers should conduct comparative analyses and integrate the results to identify commonalities and differences in predictive coding mechanisms across the various sensory domains. This approach will provide a more comprehensive understanding of the predictive processing abnormalities in ASD and help develop a unified framework that accounts for the heterogeneity of sensory experiences observed in individuals with ASD. In addition to this two-pronged approach, incorporating the subjective experiences of individuals with ASD is crucial for gaining a more comprehensive understanding of the lived experience of sensory processing abnormalities and developing ecologically valid models of ASD. Moreover, adopting a developmental perspective is essential for understanding how predictive processing abnormalities in ASD emerge and change over time. Longitudinal studies that track the development of predictive processing in ASD from infancy to adulthood can provide valuable insights into the developmental trajectories of sensory processing abnormalities and inform early intervention strategies.
By addressing these research gaps and integrating findings from multiple approaches, scientists can develop a more comprehensive and explanatory framework for understanding visual processing abnormalities in ASD. This framework will not only contribute to the theoretical understanding of ASD but also have practical implications for developing targeted interventions and support strategies to improve the quality of life for individuals with ASD.

Key words: Autism spectrum disorder, visual perception, non-social information, Bayesian, predictive coding