Similarity requirements for global workspace processes in artificial systems

Ascertain the degree of functional similarity that selection, ignition, and broadcasting processes in an artificial system must bear to the corresponding human brain processes for global workspace theory to support phenomenal consciousness in that system.

Background

Global workspace theory posits that consciousness arises when information is selected for a limited-capacity workspace and globally broadcast. The report raises interpretive challenges about how closely an artificial implementation must match human selection, ignition, and broadcast mechanisms to count as conscious under GWT.

This uncertainty affects assessments of AI architectures with workspace-like features and guides the design criteria for systems intended to instantiate consciousness-related mechanisms.