<p>Non-epistemic values are an inextricable component of scientific research, yet their legitimacy remains a contested issue, particularly in contexts of scientific advice to policymakers. Dominant arguments for their legitimacy draw on deliberative democratic theory, establishing as legitimate those non-epistemic values that align with the outcomes of due processes of public deliberation. However, I argue that alignment strategies – particularly Lusk’s (2021) “compatibilism” – implicitly rely on a degree of consensus, entailing questionable consequences in contexts marked by pervasive diversity and dissent. In response, this paper outlines an alternative framework consistent with the core insights of agonistic democracy, in an attempt to overcome these consequences. Drawing on Wenman (2013), I identify three central elements of agonistic democracy: constitutive pluralism, a tragic worldview, and the value of conflict. I suggest that these elements have epistemic implications, leading to (i) a form of epistemic pluralism akin to van Bouwel’s “interactive” pluralism; (ii) engagement with multiple forms of uncertainty, viz., aleatoric, epistemic, and relativistic; and (iii) a form of epistemic relativism that admits critical appraisal. Together, these implications form what I call an “agonistic stance”, which affords distinctive perspectives on the legitimation of non-epistemic values. To illustrate this, I engage with de Melo-Martín’s (2024) argument on the demarcation of legitimate non-epistemic values in science, showing how an agonistic stance offers complementary, but also divergent, insights.</p>