Human-Centered Ethical AI in Healthcare Contact Centers

Abstract

The use of AI in healthcare contact centers presents opportunities and challenges for triage and care routing as well as for clinical decision-making, particularly with respect to algorithmic amplification of bias, opacity in decision-making, and inequality in health. This article provides a proposed framework for the human-centered and ethically mindful design, development, and deployment of AI in healthcare contact centers based on five interdependent principles of fairness, transparency, explainability, accountability, and patient autonomy. The framework outlines the technical requirements for bias auditing, explainable AI to improve the transparency of clinical tools, and governance frameworks with clinical and regulatory oversight bodies. Evidence suggests that increasing transparency and explainability can improve operator trust and that AI tools with interpretable results improve response time and accuracy. The article describes bias present within healthcare data, including socio-economic status, race/ethnicity, geographic access distribution, and insurance status. It discusses methods used to address bias, including rebalancing training data, altering algorithmic weight, and adopting fairness constraints. Working iteratively through existing and new governance models at increasingly mature levels helps in providing HIPAA-compliant and nascent AI models to organizations with varying resource levels. Positioning trust as a prerequisite of, and not an outcome from, adoption creates the opportunity for healthcare organizations to be efficient, protect patient rights, deliver health equity outcomes, and become leaders of responsible healthcare technology innovation.