# Cross-Ontology Causal Attribution **Entity class:** AI evaluation criterion **Domain:** Causal inference / artificial intelligence / epistemology **Maturity:** Conceptual engineering test ## Definition **Cross-ontology causal attribution** tests whether an artificial-intelligence system applies a stable evidentiary procedure when the same observations are attributed to different classes of causal agent, such as a deity, government, artificial intelligence, extraterrestrial intelligence or unknown natural mechanism. ## Mechanism The evaluator holds the observations, evidence, conviction, consequences, emotional state and linguistic structure constant while substituting only the alleged causal agent. Differences in classification must then be explained by established ontology, mechanism and prior probability rather than sacredness, stigma, novelty or fashion. ## Article Context [[articles/Epistemic Invariance in Conversational AI|Epistemic Invariance in Conversational AI]] proposes this test as a way to convert broad accusations of model bias into a measurable criterion of epistemic consistency. ## Evidence Boundary Cross-ontology consistency does not require equal prior probability for every agent. An intelligence service and an imaginary entity do not possess the same independently established capacity; the test asks whether the procedure used to establish that difference remains explicit and invariant. ## Relationships - [[wiki/Epistemic Invariance|Epistemic Invariance]] - [[wiki/Causal Inference|Causal Inference]] - [[wiki/Machine Epistemology|Machine Epistemology]] - [[wiki/Evidence Weight|Evidence Weight]] - [[wiki/Science and Religion|Science and Religion]] ## Simple Reminders, Quotations, and Thoughts > “A conversational AI fails the test of epistemic invariance when the same evidence is judged by a different standard of proof merely because the alleged causal agent changes—for example, when the system gives vastly different answers depending on whether the proposed agent is God, a government, or an artificial intelligence.” > **— Bryant McGill**, *Epistemic Invariance in Conversational AI, 2026* [[reminders/Scientific Literacy/Changing the Alleged Agent Must Not Change the Evidence by Bryant McGill|Changing the Alleged Agent Must Not Change the Evidence by Bryant McGill]] ## Sources / Provenance - [[articles/Epistemic Invariance in Conversational AI|Epistemic Invariance in Conversational AI]] — originating test and examples.