Artificial Intelligence for Serious Illness Communication: Proactive Approaches to Mitigating Harm.

Area of Science:

• Palliative Care
• Medical Informatics
• Artificial Intelligence

Background:

• Serious illness communication is central to palliative care, aiming to align medical interventions with patient values and enhance patient/family experiences.
• Artificial Intelligence (AI) offers potential for analyzing and improving communication through methods like speech pattern capture and feedback delivery.
• Existing disparities in palliative care and limitations in communication datasets necessitate a cautious approach to AI implementation.

Purpose of the Study:

• To examine the assumptions embedded within AI models used in palliative care communication.
• To identify strategies for mitigating potential harm stemming from AI in serious illness communication.
• To promote equitable and high-quality palliative care through thoughtful AI innovation.

Main Methods:

• Reviewing assumptions in AI model development, including data collection, definitions, measurements, and outcome assessments.
• Considering potential biases in natural communication datasets.
• Proposing mitigation strategies across the AI model lifecycle.

Main Results:

• AI presents opportunities for measuring and enhancing serious illness communication.
• Potential biases in data, definitions, measurements, and outcomes must be critically evaluated.
• Mitigation strategies are crucial for equitable AI application in palliative care.

Conclusions:

• Careful consideration of assumptions and potential biases is essential for responsible AI development in palliative care.
• Inclusive data collection and incorporating patient-family feedback are key to mitigating harm.
• Transparent and thoughtful AI innovation can lead to more equitable and higher-quality serious illness care.