Research Group
in Analytic Philosophy

Abstract

The social sciences have had to deal with serious crises, criticisms, and calamities in recent decades. Far too many research results do not replicate; far too often do social scientists fail to predict social events of massive reach; far too many ‘evidence-based’ policies have failed to bring about the desirable changes taxpayers were promised; and far too many disciplines within the social sciences have been charged with ideological bias, which comes, or so argue the critics, at the expense of scientific standards.


Adversarial collaboration (Kahneman 2011) is a relatively new research strategy that is hoped to help with a range of methodological problems the social sciences face and restore trust in the field. The basic idea is that (teams of) scientific opponents agree on the formulation of a contested hypothesis, a research design, a neutral data collector, and to publish results regardless of what they are. Proponents of adversarial collaboration predict that the practice will lead to numerous desirable outcomes, among which to increase accountability and minimise bias among scholars, lead to more moderate, nuanced, and therefore more likely true claims, promote tolerance of genuine academic freedom and weed out scholarship with an agenda (Clark and Tetlock 2022).


The goal of this paper is to take a philosophical look at adversarial collaboration, to evaluate the strong claims that have been made in its support, and, in particular, to assess potential roles for philosophers of (social) science in adversarial collaboration. One main line of criticism will be that opponent researchers will have to share many background assumptions in order to be able to agree on research strategies. That will be difficult in cases where value judgements have wide ranging implications for how social phenomena are conceptualised, measured, and accepted as genuine. I use the Cambridge Capital Controversy to illustrate this point. The upshot is that, despite all its promise, adversarial collaboration might not deliver in cases for which it was designed, viz. cases in which disagreements about the degree of support of hypotheses derive in part from deep disagreements about value judgements.