In the concluding part, we highlight connections among these challenges and also suggest ways forward drawing on these connections.

The third challenge concerns *model performativity* as a deep problem for predictive policy advice. We focus on the question of how performative models ought to be evaluated and drawn upon for predictive policy advice.

The second one concerns navigating inherent, but often implicit, *value influences* in predictive policy advice. We discuss issues of value transparency and legitimacy and highlight shortcomings in current practice.

The first challenge revolves around the question of *epistemic pluralism*: which disciplines should – and indeed can – be drawn upon in providing input for predictive policy advice? We explore field-specific differences, unearthing practical and normative challenges for science-informed policy.

Philippe and I address three fundamental challenges predictive policy advice is faced with and that are inextricably linked to each other.