Cross-Disciplinary Evidential Substitution in AI Consciousness Research: An Epistemic Boundary Integrity Framework

Abstract

Claims about artificial consciousness increasingly combine behavioural performance, computational description, neuroscientific theory, philosophical possibility, physical analogy, and religious or spiritual interpretation. Interdisciplinarity is valuable, but it becomes scientifically unstable when evidence from one domain is used to validate a conclusion in another without an explicit, testable bridge. This focused methodological synthesis names that error cross-disciplinary evidential substitution (CDES) and develops an Epistemic Boundary Integrity Framework (EBIF). Unlike theory-derived indicator approaches, which ask whether candidate properties are instantiated, or governance principles, which regulate how research should be conducted, EBIF audits the warrant connecting evidence to attribution. It separates three evidential levels—observable behaviour, implemented mechanism, and theory-linked indicators—from a fourth target claim, phenomenal attribution, and subjects every upward inference to five checks: domain declaration, operational definition, causal bridge, rival discrimination, and independent audit. The framework does not deny the relevance of philosophy, theology, phenomenology, or physics; it assigns each a legitimate role while rejecting their use as substitutes for computational specification, empirical measurement, and falsifiable prediction. The central rule is proportionality: a conclusion should not exceed the evidence and warrant that support it. Two worked audits, operational review decisions, and a preregisterable validation strategy make the framework testable in review practice.