Keep Your Enemies Close: Adversarial Collaborations Will Improve Behavioral Science

Behavioral scientists enjoy vast methodological freedom in how they operationalize theoretical constructs. This freedom may promote creativity in designing laboratory paradigms that shed light on real-world phenomena, but it also enables questionable research practices that undercut our collective credibility. Open Science norms impose some discipline but cannot constrain cherry-picking operational definitions that insulate preferred theories from rejection. All too often scholars conduct performative research to score points instead of engaging each other's strongest arguments-a pattern that allows contradictory claims to fester unresolved for decades. Adversarial collaborations, which call on disputants to codevelop tests of competing hypotheses, are an efficient method of improving our science's capacity for self-correction and of promoting intellectual competition that exposes false claims. Although individual researchers are often initially reluctant to participate, the research community would be better served by institutionalizing adversarial collaboration into its peer-review process. General Audience Summary Behavioral and social scientists have long enjoyed vast discretion in data-analysis choices. This permissive regime has enabled scholars to engage in many deceptive analytic techniques that facilitated false claims and undercut the field's collective credibility. Recent adoption of new transparency norms has slowed these trends and has shed light on the nonreplicability of many once-prominent empirical findings. But these transparency norms lack means of constraining researchers' flexibility to cherry-pick how they define their variables and design their empirical tests that make it easier for scholars to support their preferred hypotheses. This freedom continues to facilitate false claims in the social and behavioral sciences and enables contradictory conclusions to persist for decades with little to no convergence. We propose that adversarial collaborations, which call on disagreeing scientists to codevelop tests of their competing hypotheses, are a vital supplement to current scientific norms for improving science's capacity for self-correction. Adversarial collaborations disincentivize performative research aimed at like-minded colleagues, instead favoring intense intellectual competition designed to winnow false claims. We explain why it is in the best interests of the research community to incorporate adversarial collaboration as a routine component of science.