Title Pending 28

The social sciences are facing a credibility crisis driven by two interrelated tendencies: an overreliance on abstract theorizing (“theory fetishism”) and the ritualistic application of Null Hypothesis Significance Testing (NHST), which often lends false authority to inconsequential p-values. Through a satirical ‘Health Communication Criminology’ thought experiment and real-world examples from psychology, nutrition, and criminology, we illustrate how theory fetishism encourages post hoc rationalizations, while NHST, model fit indices, and machine-learning-derived ‘importance scores’ create an illusion of certainty. These practices routinely exaggerate the significance of findings while accounting for minimal variance at the population level. To address these issues, we propose two correctives: (1) integrating effect size and prevalence using population attributable fractions to better assess societal impact; and (2) interpreting the proportion-mediated statistic (Tₚ) as an upper limit—not definitive evidence—of mediation. We further identify common analytical pitfalls, such as comparing coefficients across different models, indiscriminate use of matching and mediation analyses, and conflating interpretability with causality. A path forward requires rigorous causal designs, transparent reporting of uncertainty, and a commitment to epistemic humility—essential steps toward improving replicability and generating evidence that is truly meaningful for policy and real-world outcomes.