Adaptive Representation: A Moderate Stance on Predictive Processing

Predictive processing (PP), emerging as a novel research paradigm in contemporary cognitive science, offers a departure from both traditional computational representation views and 4E+S cognition perspectives. This theory advocates that the brain is a hierarchical prediction model based on Bayesian inference, which aims to minimize the difference between the predicted world and the actual world to prediction error minimization. In recent years, the problem of representation has emerged as a focal point in the philosophical examination of PP. This article introduces two primary strands of PP theories: conservative predictive processing (CPP) and radical predictive processing (RPP). Building upon these frameworks, it outlines three distinct positions regarding the representation problem within PP: representationalism, anti-representationalism, and a moderate stance on representations. Lastly, the article proposes a new perspective on representation: Adaptive Representation. Adaptive representation highlights the fact that generative processes are adaptive processes, and that adaptation is not necessarily optimal, whether based on natural selection or natural drift; and that generation is at the same time a representational process. By advocating for a form of weak representationalism grounded in adaptive processes, this perspective supports a moderate stance on representations within PP.