Global Workspace Theory, its LIDA model and the underlying neuroscience

A biologically inspired cognitive architecture must draw its insights from what is known from animal (including human) cognition. Such architectures should faithfully model the high-level modules and processes of cognitive neuroscience. Also, biologically inspired cognitive architectures are expected to contribute to the BICA “challenge of creating a real-life computational equivalent of the human mind”. One unified theory of cognition, Global Workspace Theory (GWT) has emerged as the most widely accepted, empirically supported theory of the role of consciousness in cognition. Recent experimental studies reveal rich cortical connectivity capable of supporting a large-scale dynamic network. We propose that brains in fact cyclically and dynamically form such a network according to GWT. The biologically inspired LIDA cognitive architecture implements GWT conceptually and computationally. Here we argue that the LIDA architecture’s breadth, flexible motivations using feelings, explicit attention mechanism, and continual, incremental and online learning in several modalities provide a significant first step in the direction of the BICA challenge. We also measure LIDA against the architectural features listed in the BICA Table of Implemented Cognitive Architectures. Applying recent brain connectivity results, we go on to elucidate the relationship between LIDA and the underlying and motivating neuroscience, using the language of non-linear dynamics. In particular, we claim that LIDA’s representations correspond to basins of attraction in the non-linear dynamics of neural activation patterns. In addition, we claim that the rhythms of LIDA’s cognitive cycle and of its internal cognitive elements have definite psychophysiological corollaries in the oscillatory patterns observed in the human brain.