CATB: A Causal Benchmark for Assessing Consciousness-Relevant Capacities in Artificial Intelligence

Abstract

Artificial intelligence systems are increasingly capable of producing convincing self-reports, strategic behavior, and apparent introspection. However, such behaviors do not by themselves provide evidence of phenomenal consciousness. This working paper introduces the Causal Artificial Consciousness Triangulation Battery (CATB), a methodological framework and benchmark proposal for assessing consciousness-relevant capacities in artificial systems. Rather than relying on self-report or behavioral imitation alone, CATB combines theory mapping, privileged-access perturbation tests, no-report paradigms, mechanistic interventions, metacognitive assessment, source monitoring, counterfeit-consciousness controls, and evidence grading into a unified experimental framework. The proposed benchmark is designed to distinguish linguistic simulation from causally integrated internal self-monitoring while explicitly acknowledging the current epistemic limits of consciousness research. CATB does not claim to detect or prove subjective experience. Instead, it provides a structured methodology for accumulating and evaluating converging empirical evidence under uncertainty. The manuscript also proposes falsification criteria, experimental design principles, evidence tiers, governance recommendations, and reporting standards intended to support transparent and reproducible research in artificial consciousness. Version 1.0 (Preprint)