Active Inference and the Circuits of Precision and Prediction

The human brain exhibits remarkable cognitive abilities, such as imagery, attention and mental simulation, which along with perception reflect our capacity to infer properties of the world. Perceptual inference involves a statistical hierarchical synthesis of information from the senses with prior knowledge of the world. There is empirical evidence in favour of Predictive Processing where functional interaction of descending pathways carrying top-down prediction and precision signals interaction with the ascending pathways to generative prediction errors. PREDICTION will develop a mechanistic and computational understanding of the hierarchical processing underlying inference in vision-related human brain areas and will establish how feedback connections enable generative cognitive functions. Unravelling the neural mechanisms of high-level visual cognitive functions requires novel ways of thinking about biological information processing and an integrated effort to measure and model cognitive functions at multiple brain scales. PREDICTON accomplishes this by integrating concepts and advanced methods from multiple disciplines: 1) Molecular biology to characterise projection cell-types. 2) Psychophysics and sub-millimetre ultra-high-field functional imaging of cortical layers and columns in the human brain, 3) Precision viral tracing to investigate the anatomical origins and targets of generative networks in non-human primates (NHP), 4) Spatio-temporal controlled activation of genetically identified feedback pathways with optogenetics together in multi-areal laminar recordings in NHPs, 4) Biologically realistic multi-scale computational modelling. PREDICTION will describe the functional circuits supporting hierarchical processing and their role in active vision. This will take us to the heart of cortical function and its causative role in perception, cognition, and decision-making, relative to a range of pathologies as well as to understand what makes us human.