Higher-Level Cognition Under Predictive Processing: Structural Representations and Grounded Cognition

Predictive processing explains perception, action, and cognition as prediction-error minimization, but it is unclear how this supports abstract reasoning. Combining predictive processing, structural representations, and grounded cognition addresses this. Structural representations are isomorphic to the world, retaining its relational patterns. Grounded cognition contributes three mechanisms: hierarchical organization abstracts from sensory qualities; language binds disparate sensory qualities into representations and acts as a social tool; metaphoric mapping uses fragments of concrete percepts to represent abstract concepts. Transplanting these into a hierarchical generative model explains higher-level cognition through detached simulations of perception and action isomorphic to actual behavior. This expands life-mind continuity by specifying how principles driving life's emergence also account for sophisticated human cognition.

Study at a glance

Abstract

Predictive processing posits that prediction-error minimization underlies all perception, action, and cognition. Yet, despite its considerable popularity and explanatory scope, it is unclear how this enables higher-level cognitive abilities, such as representing and reasoning over abstract concepts. We combine insights from predictive processing, structural representations and grounded cognition to address this issue. It has been argued from predictive processing and the free energy principle that an anticipatory model of the person-relevant environment is simulated. Structural representations state that these representations are isomorphic to, i.e., retain the relational pattern of the world. Building on this assembly, grounded cognition research provides three insights into how abstract concepts are represented. First, a hierarchical organization allows abstracting from specific sensory qualities. Second, language glues together sensory qualities into representations that share no intrinsic properties, and acts as a social tool. Third, metaphoric mapping allows fragments of concrete percepts to represent abstract concepts. By transplanting these three insights to predictive processing’s (structural) hierarchical generative model, we explain higher-level cognition through detached models of perception and action simulations, isomorphic to actual behavior. This constitutes a significant expansion to life-mind continuity approaches by providing specific mechanisms for how the principles driving the emergence of life also account for sophisticated higher-level cognition in humans. By synthesizing insights from these literatures, we generate a coherent description of higher-level cognition under predictive processing.