Introduction

Instrumentalism and realism represent two philosophical positions regarding the nature of scientific laws and their application. Instrumentalism views laws as tools for making predictions, while realism posits that these laws reflect a deeper truth about the universe created by a higher power. The distinction between these perspectives is essential for understanding scientific models, particularly in the context of machine learning. Currently, most practitioners approach machine learning from an instrumentalist standpoint, aiming to develop models that facilitate accurate classification. However, an alternative approach would emphasize the importance of understanding the underlying conditional probabilities that govern the behavior of systems, rather than merely focusing on predictive accuracy. This perspective suggests that while prediction is important, grasping the deeper probabilistic relationships offers a more profound understanding of the complexities in learning and inference.