The Evidence Genome: A Generalized Taxonomy for Auditing Autonomous Systems

The Evidence Genome is presented as a generalized metadata-classification taxonomy designed to audit and map empirical evidence in complex autonomous systems. Its mathematical and structural utility is decoupled from consciousness research. To demonstrate generalizability, the manuscript evaluates the framework within a fleet of self-driving vehicles experiencing edge-case system degradation. A failure event is processed through the taxonomy, classifying each diagnostic telemetry into provenance, reproducibility, mechanistic, and predictive evidence categories. For example, if packet loss exceeds 0.2 and V2V latency exceeds 100 ms, the failure is classified as mechanistic communication breakdown rather than sensor noise. This validates the framework's diagnostic value independently of any consciousness hypotheses. LIMITATIONS, FAILURE MODES, AND ALTERNATIVE HYPOTHESES (H0): The Evidence Genome is a classification taxonomy and does not establish causal relationships by itself. Failure modes include: 1. Taxonomy Incompleteness: The categories may not cover all failure types. Mitigation: Iterative expansion and community feedback. 2. Misclassification: Telemetry may be ambiguous. Mitigation: Probabilistic classification and expert adjudication. 3. Overfitting to Specific Domains: The taxonomy may not generalize beyond tested systems. Mitigation: Cross-domain validation studies.