Recent advances in foundation models have transformed AI for Science, enabling remarkably accurate predictive performance across domains ranging from protein folding to weather forecasting. However, prediction alone does not constitute scientific discovery. Scientific understanding relies on uncovering reusable explanatory mechanisms that generate observations, whereas contemporary machine learning remains fundamentally organized around predictive mappings rather than explanatory structure. This paper argues that scientific discovery is fundamentally a problem of knowledge organization.
To this end, we introduce Mechanistic World Models, a new design paradigm that places reusable mechanisms at the center 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 formalize 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.
Blogger's Review: The proposed Mechanistic World Models offer a refreshing perspective on scientific discovery, emphasizing the importance of reusable mechanisms in knowledge organization. This framework could significantly enhance AI's application in scientific research, particularly in terms of interpretability and sustainability, deepening our understanding of complex systems. By integrating various research directions, future AI systems are expected to become more autonomous and creative.