From Observation to Insight: Mechanistic World Models and the Quest for Autonomous Discovery

2026-07-14

AI Summary

Current AI systems, while proficient at prediction, lack the capacity to model reusable explanatory mechanisms, hindering autonomous scientific discovery. This work proposes a novel paradigm—mechanism-centric world models—that reframes scientific discovery as a knowledge organization problem. By integrating mechanistic interpretability, causal representation learning, symbolic equation discovery, and modular neural architectures, and by incorporating design principles from philosophy of science as inductive biases, the framework formalizes the structure of mechanism-driven world models. It clarifies the computational capabilities required for autonomous discovery and provides a unified theoretical foundation that bridges multiple cutting-edge research directions, thereby advancing AI beyond prediction toward explanation and discovery.

Abstract

Recent advances in foundation models have transformed AI for Science, enabling remarkably accurate predictive performance across domains ranging from protein folding to weather forecasting. Yet prediction alone does not constitute scientific discovery. Scientific understanding depends on uncovering the reusable explanatory mechanisms that generate observations, whereas contemporary machine learning remains fundamentally organised around predictive mappings rather than explanatory structure. In this paper, we argue that scientific discovery is fundamentally a problem of knowledge organisation. To this end, we introduce Mechanistic World Models, a new design paradigm that places reusable mechanisms at the centre of representation, computation and learning. Drawing on insights from the philosophy of science, we derive the computational capabilities required for discovery, identify the design principles and inductive pressures that encourage explanatory knowledge to emerge, and formalise the anatomy of a mechanism-centric world model. Finally, we show how diverse research directions including mechanistic interpretability, causal representation learning, equation discovery and modular architectures capture complementary ingredients of this paradigm while lacking a unified framework. We propose Mechanistic World Models as a conceptual foundation and computational blueprint for moving AI beyond predictive forecasting towards autonomous scientific discovery.

Problem

• scientific discovery
• explanatory mechanisms
• predictive modeling
• knowledge organisation
• world models

Innovation

• Mechanistic World Models
• scientific discovery
• explanatory mechanisms
• causal representation learning
• modular architectures