A Modeling Framework For Resolving Conflicts Between Experts’ Intuition And Data-driven AI Models

Abstract

In this work, we incorporate experts' intuition into existing machine learning models. We define expert intuition as an expert’s insight that is not statistically explained by available data and thus, it cannot be captured statistically from the available training data, but it is also not a random guess. We aim at making the best of both worlds; AI data-driven prediction models and experts’ intuition, by resolving conflicts between the two and using mathematical methods to model the intuition within the prediction models. We apply our method to a real-world use case directly aligned with the topic.