Newton to Einstein: Axiom-Based Discovery via Game Design
This work addresses the need for creative, interpretable, and theory-driven discovery in machine learning, though it is incremental as it builds on existing concepts of reasoning and game design.
The paper tackles the problem of machine learning for scientific discovery by shifting from inductive pattern recognition to axiom-based reasoning, proposing a game design framework where agents evolve rules to explain outliers, and demonstrates feasibility through preliminary experiments in logic-based games where agents evolve axioms to solve previously unsolvable problems.
This position paper argues that machine learning for scientific discovery should shift from inductive pattern recognition to axiom-based reasoning. We propose a game design framework in which scientific inquiry is recast as a rule-evolving system: agents operate within environments governed by axioms and modify them to explain outlier observations. Unlike conventional ML approaches that operate within fixed assumptions, our method enables the discovery of new theoretical structures through systematic rule adaptation. We demonstrate the feasibility of this approach through preliminary experiments in logic-based games, showing that agents can evolve axioms that solve previously unsolvable problems. This framework offers a foundation for building machine learning systems capable of creative, interpretable, and theory-driven discovery.