A Comprehensive Exploration of Predictive Processing: An fMRI Meta-Analytic Approach to Investigate “Predictive” Networks Across Multiple Cognitive Domains, Including the Development of a New Tool for the Evaluation of Risk of Bias in Cognitive Studies

This work further explores the existence of a network specifically devoted to prediction in the brain. Building on previous knowledge, multiple meta-analyses on General Prediction, Prediction Encoding and Prediction Violation are performed, both within and across different cognitive domains (Cognitive Control, Attention, Language, Motor, and Social Cognition). Domain-specific prediction appears to be mediated by different brain regions depending on the considered domain, which closely resemble those that are involved in signal processing and elaboration in that domain. Moreover, the processing of prediction violation seems to involve the insula and the inferior frontal gyrus, as previous meta-analytic work suggested. Finally, previous findings on the existence of a prediction network in the brain are partially replicated, once again highlighting the role of the insula and the inferior frontal gyrus in said network. Due to unforeseeable computational limitations, these results can be considered only partial and need replication. In parallel with this meta-analytic work, a new tool for the evaluation of risk of bias in cognitive studies is developed, as the literature lacks such an instrument. This tool is therefore created by adapting a checklist for the assessment of quality in cognitive studies and then applied to a subset of papers from the Social Cognition meta-analyses. Its statistical properties are assessed starting from these preliminary data: although its inter-rater reliability is not good, this instrument shows potential to be widely used in research, once a few adjustments to its structure are made.