From cacophony to hierarchy: a principled framework for assessing AI consciousness

Summary

Proposes a principled framework for AI consciousness assessment: five functional-description levels plus Bayesian aggregation of theoretical credences and behavioral indicators.

Keywords

AI consciousness · philosophy · evaluation

Topics

AI consciousness, philosophy, evaluation

Research notes

• Discovery: Shared in #random-papers as an arXiv link.
• Method: Defines five functional levels of description; aggregates credences from theories of consciousness and empirical indicators via Bayesian updating.
• Key findings: Illustrative LLM consciousness estimates range from below 0.01 to roughly 0.8 depending on assumptions; the authors advocate structured agnosticism rather than a single number.