Modeling Consciousness Through Attention: A Predictive Coding Approach to the Social Brain

Abstract

This thesis explores the Attention Schema Theory (AST) as a computational framework for understanding consciousness and social cognition. Through a series of behavioral experiments, we test whether humans can distinguish real from artificial attention sequences based on gaze-like motion patterns. Participants consistently performed above chance, even under degraded visual conditions, and confidence ratings tracked performance—suggesting the presence of an internal model guiding judgments. Open-ended commentary further revealed partial introspective access to this modeling process. A follow-up fMRI study is discussed in support of AST’s proposed neural architecture, though the primary emphasis remains on behavioral evidence. I propose a multi-level refinement of AST, integrating predictive coding and a hierarchical account of awareness.

Keywords: attention schema, consciousness, predictive coding, social brain, fMRI, TPJ, meta-consciousness