About this Research Topic
Empirical evidence posits the synergic interplay of sensory processing and reinforcement learning. Sensory processing is a general term used to describe how, sensory information derived from the environment and one’s own body is processed so experiences may be understood and appropriate responses may be carried out. The underlying neurobiological mechanisms include registration, modulation and integration with corresponding dysfunctions ranging from basal sensory deficit/ sensory discrimination disorder, sensory modulation disorder, and sensory based motor disorder, respectively. Presence of sensory processing disorder has been implicated in autism spectrum disorder, in specific language impairment, and may contribute to the neurodevelopmental delays identified in preterm children.
Based on the foundations of the pioneering work of Jane Ayres, interventions utilizing multisensory strategies for improving sensory processing disorders have emerged. These emphasize the need to offer challenging goal-directed environments rich in specific sensory output. Utilizing the brain’s mechanism to follow Bayesian-optimal strategies for multisensory integration, repetitive yet continuously challenging tasks are exploited to facilitate improved sensory processing and consequent adaptive interactions with the environment. Another requisite process element of the intervention is the need to support the child’s motivation to fully engage. This is achieved by supporting the successful completion of tasks that are intrinsically “fun” yet pose “just the right challenge”.
Reinforcement learning is a fundamental form of learning where learning is governed by the rewarding value of a stimulus or action. Its two forms, model-based and model-free learning differ regarding the use of a world model. Generally, model-based reinforcement learning uses a model to conduct forward-looking simulations for the sake of making predictions and/or optimizing policy in a way that the cumulated sum of the reward is maximized in the long term. A key issue of model-based learning regards how the internal representations of the environment are created as the world model must represent the salient features of the external and internal (interoceptive, viscerosensory, affective and cognitive) environment. Previously, building on the proactive brain concept it was proposed that model-based learning utilizes association-based context frames to build the world model, upon which forward looking mental simulations and predictions may be formulated. Context frames are created by compiling stimuli and their internal (cognitive/affective (including reward-related), interoceptive (physiological and neurohumoral)) and external (spatial, temporal, social or cultural) settings. They contain contextually associated information as an average of similar contexts containing typical, generic representations and signal cue-context associations reflecting statistical regularities and a lifetime of extracting patterns from the environment (related to contingencies, spatial locations, temporal integration, etc.) Hence it may be strongly suggested that context frames are derived as a function of multisensory processing.
To provide a more complete picture of the synergic interplay of sensory processing and reinforcement learning, this Research Topic encourages contributions assessing this potential interaction. Use of rigorous diagnostic criteria for defining sensory processing disorders is a must. Both observational and explorative studies are encouraged.
Keywords: sensory processing disorder, sensory integration theory, multisensory integration, reinforcement learning, autism spectrum disorder
Important Note: All contributions to this Research Topic must be within the scope of the section and journal to which they are submitted, as defined in their mission statements. Frontiers reserves the right to guide an out-of-scope manuscript to a more suitable section or journal at any stage of peer review.