Efficient and trustworthy decision making through human-in-the-loop visual analytics: A case study on tax risk assessment

Abstract

Data mining and AI techniques are increasingly being used to automate data analysis. Ideally, one may wish to completely automate the data analysis process, but in many real-world applications a full automation may pose significant risks. In these cases, human analysts must be directly involved to refine the analysis or to make the final decisions. A challenging problem, therefore, is how to perform efficient and trustworthy decision-making when humans are an integral part of the analysis pipeline. We propose a “human-in-the-loop” methodology that leverages data mining, machine learning, and visual analytics to improve and speed up the analysis. A key feature is the use of a dashboard that integrates intuitive visual tools, which aid analysts to efficiently discover hidden data patterns or to get helpful insights. We describe in particular how this methodology has been successfully applied to support Revenue Agency officers in tax risk assessment.