Embodied Prediction

Versions of the “predictive brain” hypothesis rank among the most promising and the most conceptually challenging visions ever to emerge from computational and cognitive neuroscience. In this paper, I briefly introduce (section 1) the most radical and comprehensive of these visions—the account of “active inference”, or “action-oriented predictive processing” (Clark 2013a), developed by Karl Friston and colleagues. In section 2, I isolate and discuss four of the framework’s most provocative claims: (i) that the core flow of information is top-down, not bottom-up, with the forward flow of sensory information replaced by the forward flow of prediction error; (ii) that motor control is just more top-down sensory prediction; (iii) that efference copies, and distinct “controllers”, can be replaced by top-down predictions; and (iv) that cost functions can fruitfully be replaced by predictions. Working together, these four claims offer a tantalizing glimpse of a new, integrated framework for understanding perception, action, embodiment, and the nature of human experience. I end (section 3) by sketching what may be the most important aspect of the emerging view: its ability to embed the use of fast and frugal solutions (as highlighted by much work in robotics and embodied cognition) within an over-arching scheme that includes more structured, knowledge-intensive strategies, combining these fluently and continuously as task and context dictate.