An Approach to Automatically Detect and Visualize Bias in Data Analytics

Abstract

Data Analytics and Artificial Intelligence (AI) are increasingly driving key business decisions and business processes. Any flaws in the interpretation of analytic results or AI outputs can lead to significant economic loses and reputation damage. Among existing flaws, one of the most often overlooked is the use biased data and imbalanced datasets. When unadverted, data bias warps the meaning of data and has a devastating effect on AI results. Existing approaches deal with data bias by constraining the data model, altering its composition until the data is no longer biased. Unfortunately, studies have shown that crucial information about the nature of data may be lost during this process. Therefore, in this paper we propose an alternative process, one that detects data biases and presents biased data in a visual way so that the user can comprehend how data is structured and decide whether or not constraining approaches are applicable in his context. Our approach detects the existence of biases in datasets through our proposed algorithm and generates a series of visualizations in a way that is understandable for users, including non-expert ones. In this way, users become aware not only of the existence of biases in the data, but also how they may impact their analytics and AI algorithms, thus avoiding undesired results.