Predictions in a dynamic world: the adaptive role of the Active Prediction Network

Human beings evolved in a complex and constantly changing environment, in which effective behaviour depends on the capacity to anticipate the unfolding of future events. Accordingly, predictive mechanisms have been investigated across multiple cognitive domains, including visual perception, social cognition, and language comprehension. While extensive research has focused on automatic linguistic predictions based on long-term statistical regularities, the neural mechanisms supporting active prediction, the ability to flexibly anticipate semantic information in novel or atypical contexts, remain poorly understood. Across three functional magnetic resonance imaging (fMRI) studies, this thesis demonstrates that such predictive processes are supported by a left-dominant network incorporating both semantic and domain-general control regions: the Active Prediction Network (APN). This distinctive topographical overlap suggests that the mechanism for active prediction repurposes the neural machinery typically involved in flexible semantic retrieval, applying it proactively to anticipate the semantic content of upcoming events. In parallel, domain-general control regions may contribute by integrating newly encountered regularities into internal models of environmental statistics and by coordinating the sequence of operations required to sustain active prediction. Predictive computations carried out by the APN facilitate processing within semantic representational regions belonging to the default mode network (DMN), revealing a dynamic interplay between these two networks where the former support the latter in processing the content of novel stimuli. Taken together, these findings underscore the importance of the APN for ecologically relevant forms of prediction, in which anticipations are embedded within rich and dynamic contexts rather than generated in isolation. By supporting flexible, proactive, and context-sensitive behaviour, the APN may constitute a key neural mechanism through which humans rapidly adapt to novel circumstances and continuously changing environments.