Problem-solving ontologies on steroids? The space for critique in policy research on artificial intelligence

This article critically examines social sciences research on AI and policy as a politics of knowledge production, offering three key contributions. First, it debates the merit of systematic reviews in critical policy research in revealing epistemic hierarchies and biases. Second, it introduces a heuristic to map the disjointed methodologies of AI-policy scholarship across two ontological dimensions: whether AI features as a neutral problem-solving tool or a socio-technical system shaping power relations; and whether technology serves as a governance instrument or a regulatory target. This framework provides a reflexive, yet systematic, tool to expose epistemic hierarchies and the relative space for critical social scientific engagement. Third, I present findings from a systematic review of 229 English-language articles published in social sciences journals through 2025. Results show a pronounced imbalance: three-quarters of studies focus on AI regulation, while empirical research on technology uses in governance remains marginal. Moreover, the majority of the corpus operates within problem-solving ontologies, framing AI either as an optimizer of policymaking or as an objectively manageable risk. In the discussion, the article therefore calls for greater scholarly reflexivity and debate on the role of social sciences knowledge production in examining power asymmetries, techno-solutionist ideologies, and social inequalities in the AI-policy complex.