Introduction

The distinction between instrumentalism and realism highlights two contrasting perspectives in understanding the nature of laws and theories. Instrumentalism views laws as tools for prediction without asserting their absolute truth, while realism posits that these laws reflect a deeper reality governed by a creator's design. In the context of machine learning, the primary goal often focuses on generating rules for classification, representing an instrumental approach. However, an alternative view suggests that machine learning should also aim to understand conditional probabilities, allowing for a richer comprehension of the underlying processes. This duality emphasizes that while predictions are vital for practical applications, understanding the probabilities involved provides a more profound insight into the mechanisms of reality and learning.