Modeling Working Memory to Identify Computational Correlates of Consciousness
Open Philosophy October 6, 2019 A. Reggia James, E. Katz Garrett, P. Davis Gregory
A framework from recent philosophical work on consciousness, including cognitive phenomenology and mereological analysis, can help guide computational modeling. The authors argue that studying the computational mechanisms of working memory and cognitive control is especially likely to reveal computational correlates of consciousness. They describe their own computational models of human working memory and propose three such correlates: itinerant attractor sequences, top-down gating, and very fast weight changes. Current work tests whether these three correlates can support more complex models involving compositionality and basic causal inference. The authors conclude that computational models of working memory offer a promising path for understanding consciousness generally and for assessing the long-term potential of artificial consciousness.