Truth vs. Opinion: An Event-Valued Test for Objective Discernment

This paper introduces an event-valued kernel framework for distinguishing truth from opinion. Unlike classical theories, which ground truth in correspondence, pragmatism, or deflationary semantics, the kernel requires invariance across admissible observers and typal contexts under guardrails. Propositions that fail under enlargement accumulate scars, while robust truths converge under guardrail-consistent families. This discriminator is stricter than correspondence, blocks “useful falsehoods,” and resists calibration relativism through meta-audits and latent-flag detection. Extensive adversarial audits—including statistical pilots with high power (≥0.8) and AI adversary stress-tests—produced no reproducible false positives: false propositions scar or remain provisional, while only invariant truths persist. The result is a firewall separating truth from well-organized opinion, unifying and extending earlier epistemological accounts.