Rethinking Explanation, Causality, and Prediction in Social Science
European Political Science Society Annual Meeting – 18-20 June 2026 – Belfast
CERSP Co-Directors, Matthew Loveless and Chiara Binelli, organized a panel titled, “Rethinking Explanation, Causality, and Prediction in Social Science’ [ME01] to bring together leading-edge thinking on empirical social science research and its direction.
Description: Across the social sciences, causality has become an organizing principle but also a polarizing force in research methodology. Entire research programs now pivot on identifying causal effects, which some argue has led to the expense of broader explanatory or theoretical ambitions. This panel brings together work that takes on these questions by pointing to novel ways of gaining credible, generalizable, and meaningful social inquiry. At the center lies a shared question: What does it mean to learn from data?
The papers approach this from different angles – rethinking how uncertainty can be measured and preregistered rather than merely controlled; how the accumulated record of studies might be reanalyzed to reveal generalizable patterns; how prediction can serve as a route to stronger theory rather than a substitute for it; and how the very concept of causality has drifted from a theoretical ideal into a methodological end in itself.
Taken together, these interventions suggest that methodological innovation is necessary to balance competing goals of research design. To move forward, these methods help make causality, prediction, and description as complementary methods that aim for the full explanation of complex human realities.
Prof. Loveless and Prof. Binelli proposed the following:
Social science has increasingly treated causal identification, prediction, and explanation as if they belong to the same conceptual hierarchy. While they intermingle and are related theoretically, methodologically they are distinct. Causal identification estimates effects under assumptions while prediction evaluates performance across observations or contexts. Explanation continues to resist simple methodological capture requiring greater conceptual understanding of mechanisms, scope, meaning, and data-generating processes.
Thus, even when causal effects are credibly identified, they do not necessarily explain the broader social processes that generate outcomes, a limitation that prediction further exposes – but does not resolve – by showing that many causal models predict poorly. We therefore argue not only for a clearer methodological distinction but also more formal conceptual separation between causality, prediction, and explanation that repositions the potential roles that causality and prediction can play in social scientific explanation.