The Minimal Viability Test (MVT): An Axiomatic Framework for Structural Integrity and a Sanitation Protocol for the AI Era

The replication crisis in contemporary science suggests a systemic failure not merely in statistical method, but in the logical architecture of hypotheses. R.A. Fisher famously warned that consulting statistics post-experiment is merely a “post-mortem examination”; yet, current protocols overlook fundamental structural deficits. Here, we derive the Minimal Viability Test (MVT) from the first principles of Information Theory, Category Theory, and Bayesian Logic. We posit that a scientific argument is a lossy compression of reality that must preserve topological structure to be valid. By operationalizing concepts from Lakatos and Gentner, MVT provides a deterministic protocol to detect “Axiom Smuggling” and “Isomorphism Violations.” Crucially, as generative AI begins to saturate the academic corpus, MVT serves as a necessary immunological filter to prevent “Model Collapse”—ensuring that future intelligences are not trained on logically incoherent data.