Diagnosing Inequality: The Promise and Limits of AI in Bridging Healthcare Gaps

Abstract

Healthcare disparities in the United States continue to disproportionately affect low-income individuals, ethnic minorities, migrants, and people with disabilities due to systemic barriers such as limited access to care, provider shortages, and financial constraints. This paper evaluates the effectiveness and limitations of artificial intelligence (AI)-based diagnostic tools in addressing these inequities. Drawing from a comprehensive literature review, AI simulation testing, sensitivity analyses, and qualitative case studies, the study assesses diagnostic accuracy across diverse demographic profiles. Findings indicate that AI systems perform well in identifying single-cause conditions when provided with detailed symptom inputs, but exhibit reduced accuracy in complex or multi-causal cases, especially when training data lack demographic diversity. Disparities in diagnostic outputs by race and age further underscore embedded algorithmic biases. Integration of mobile diagnostic devices and culturally representative datasets improves AI performance in under-resourced settings. The paper concludes with strategic recommendations for algorithm refinement, infrastructure investment, workforce training, and policy integration to ensure that AI technologies contribute meaningfully to healthcare equity.

Key Words: healthcare disparities; artificial intelligence; diagnostic accuracy; social service integration; algorithmic bias; health equity; underserved populations; digital health ethics