Newton to Einstein: Axiom-Based Discovery via Game Design

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Ma, Pingchuan, Jones, Benjamin Tod, Wang, Tsun-Hsuan, Guo, Minghao, Lipiec, Michal Piotr, Gan, Chuang, Matusik, Wojciech
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915482926317568
author Ma, Pingchuan
Jones, Benjamin Tod
Wang, Tsun-Hsuan
Guo, Minghao
Lipiec, Michal Piotr
Gan, Chuang
Matusik, Wojciech
author_facet Ma, Pingchuan
Jones, Benjamin Tod
Wang, Tsun-Hsuan
Guo, Minghao
Lipiec, Michal Piotr
Gan, Chuang
Matusik, Wojciech
contents 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.
format Preprint
id arxiv_https___arxiv_org_abs_2509_05448
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Newton to Einstein: Axiom-Based Discovery via Game Design
Ma, Pingchuan
Jones, Benjamin Tod
Wang, Tsun-Hsuan
Guo, Minghao
Lipiec, Michal Piotr
Gan, Chuang
Matusik, Wojciech
Computational Engineering, Finance, and Science
Artificial Intelligence
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.
title Newton to Einstein: Axiom-Based Discovery via Game Design
topic Computational Engineering, Finance, and Science
Artificial Intelligence
url https://arxiv.org/abs/2509.05448