T. Ekin, L. Shaw
Texas State University,
United States
Keywords: AI, health, adversarial, AI security, agents
Summary:
We introduce a dynamic data-driven health care billing audit framework, leveraging AI and adversarial risk analysis (ARA). ARA frames the relationship between the health care billing process and adversarial decision makers as a game from the perspective of the billing analyst. The uncertainty of the process about the adversarial agents’ uncertainties and goals are modeled using a generative AI model, e.g., large language model that generates scenarios given the fraud context and data. Such integration of domain-specific datasets, adversarial scenarios, and dynamic data streams within ARA enables the proposed proactive framework, e.g., AI agent to address evolving adversarial tactics under uncertainty. We illustrate the framework through a health care billing fraud detection example utilizing clustering-based outlier detection.