Beyond the p < 0.05 trap: the adaptive integrity model for preventing and detecting P-hacking

When statistical significance becomes the currency of publication, the incentive to reach p < 0.05 can subtly shape research behavior. While many researchers report their findings with integrity, others face implicit pressures that increase the likelihood of selective reporting or post hoc adjustments. This article introduces the Adaptive Integrity Model (AIM), a conceptual framework that synthesizes explanation, prediction, and detection to illuminate how p -hacking tendencies arise within specific institutional and cognitive environments. Whereas existing tools such as p -curve and z -curve offer retrospective diagnostics based on p -value distributions, AIM complements these by embedding detection within a broader model of behavioral and structural influences. Its explanatory component quantifies the structural incentives and psychological biases that shape research behavior. Its predictive component flags statistical irregularities such as clustering near the significance threshold, omitted test reporting, and boundary inflation. Its detection component evaluates transparency through replication outcomes, preregistration adherence, and analytic completeness. Validated across five real-world datasets, including studies later retracted or disputed, AIM generates a P integrity score that captures statistical anomalies and contextual vulnerabilities. By modeling research integrity as a layered system, AIM offers journal editors, funders, and reviewers a scalable tool for credibility assessment that promotes retrospective research audits and prospective safeguards.