A decision architecture for epistemic prioritization: Machine learning at the intersection of technology and society

Abstract

This review examines how machine learning (ML) methodologies are transforming the philosophy of science and engineering through five critical epistemic functions: Prediction, Explanation, Discovery, Understanding, and Decision-making (P.E.D.U.D.). We analyze each function individually and then provide examples of how ML applications embody these epistemic aims. Building on this analysis, we develop a framework to help users/practitioners determine which epistemic function to prioritize for specific problem domains by creating a decision architecture that aligns ML methodologies with epistemic goals. Finally, we explore the broader philosophical implications of this epistemological landscape by analyzing tensions between data-driven and theory-driven approaches and argue that ML necessitates a reconsideration of the traditional philosophy of science as the balance between these five functions evolves.