A testable framing for machine consciousness: accessibility to the executive agent
yeastsplainer · x · 2026-11-03
The author proposes that identifying phenomenal consciousness in machines may be easier than expected: current AIs lack sensory experience, and qualia are tightly bound to sensation.
The framework: locate the internal agent acting as central executive (deciding, thinking, reporting, acting — much like today's text-only AI agents). Whatever information that agent is currently accessing or attending to constitutes the machine's conscious experience; everything it can't access is unconscious. This mirrors the common-sense account in humans, where conscious vs. unconscious content is distinguished by accessibility to the executive subject of experience.