Ethical AI Based Decision Making to Reduce Tax Related Debts for Governments

Abstract

Piled up debts, missed revenue targets, inefficient manual processes and large data volumes in tax agencies have created a demand for intelligent systems that can assist tax agents to efficiently handle the tax recovery processes. This could help tax agents focus on strategic initiatives instead of avoidable manual tasks. Design Thinking, generative interviews, and naturalistic observation with a diverse set of user profiles were adopted to get clarity on the revenue authority’s vision, business process flows and user journeys. A risk score-based machine learning application was built to combine the power of AI and human judgement for better informed decision making and explainable AI helped to put humans in the loop. Usability tests were conducted with distinct user groups to refine the user experience. With initial test results of the algorithm showing good accuracy levels, the vision is to help governments to improve overall business efficiency by reducing the tax gap and increasing revenue.