User Driven Model Adjustment via Boolean Rule Explanations

AI solutions are heavily dependant on the quality and accuracy of input training data, however data may not always fully reflect most up-to-date policy landscape or be missing business logic. The advances in explainability have opened possibility allowing users to interact with interpretable explanations ML predictions order inject modifications constraints that more accurately current realities system. In this paper, we present a solution which leverages predictive power models while user specify decision boundaries. Our interactive overlay approach achieves goal without requiring model retraining, making it appropriate for systems need apply instant changes their making. We demonstrate feedback rules can layered provide immediate turn supports learning less data.