Introduction

The discourse emphasizes the distinction between two philosophical positions: instrumentalism and realism. Instrumentalism views laws as tools for predictions, while realism posits that these laws reflect deeper truths about the universe, believed to be ordained by a higher power. The conversation highlights the tension between using machine learning models for predictive purposes versus seeking a more profound understanding of the underlying probabilities and conditions that govern behaviors. In the realm of machine learning, the goal is often framed as finding rules for classification, aligning with instrumentalist thought. However, a more profound objective can be framed as learning about conditional probabilities, leading to a deeper comprehension of the intricacies of the world, rather than merely achieving accurate predictions.