Adversarial alignment enables competing models to engage in cooperative theory building toward cumulative science

Abstract

Crises in science concern not only methods, statistics, and results but also, theory development. Beyond the indispensable refinement of tools and procedures, resolving crises would also benefit from a deeper understanding of the concepts and processes guiding research. Usually, theories compete, and some lose, incentivizing destruction of seemingly opposing views. This does not necessarily contribute to accumulating insights, and it may incur collateral damage (e.g., impairing cognitive processes and collegial relations). To develop a more constructive model, we built on adversarial collaboration, which integrates incompatible results into agreed-on new empirical research to test competing hypotheses [D. Kahneman, Am. Psychol. 58, 723-730 (2003)]. Applying theory and evidence from the behavioral sciences, we address the group dynamic complexities of adversarial interactions between scientists. We illustrate the added value of considering these in an "adversarial alignment" that addressed competing conceptual frameworks from five different theories of social evaluation. Negotiating a joint framework required two preconditions and several guidelines. First, we reframed our interactions from competitive rivalry to cooperative pursuit of a joint goal, and second, we assumed scientific competence and good intentions, enabling cooperation toward that goal. Then, we applied five rules for successful multiparty negotiations: 1) leveling the playing field, 2) capitalizing on curiosity, 3) producing measurable progress, 4) working toward mutual gain, and 5) being aware of the downside alternative. Together, these guidelines can encourage others to create conditions that allow for theoretical alignments and develop cumulative science.