Informed consent in the age of value-laden medical AI

Purpose The purpose of this paper is to argue that the value structures and biases embedded in medical artificial intelligence (AI) systems must be explicitly disclosed as part of the informed consent process to uphold patient autonomy and ethical care. Design/methodology/approach The authors examine how AI value structures arise from objective functions and training data, illustrating their impact through real-world cases in health care where algorithmic misalignment led to harm. The authors then connect these issues to the ethical and legal standards of informed consent in medicine, as well as patient autonomy and well-being. Findings Medical AI systems differ from other medical technologies owing to their opaque value-laden decision frameworks, scalability and intelligent capacity. Furthermore, these value systems have led to real-life negative consequences for patients’ health and well-being at both small and large scales, with the potential for further effects as these systems become more capable. When related back to the principles of informed consent and appropriate disclosure of information to patients, this makes it necessary for providers to have a dialogue with patients about these complex, value-laden systems when they are involved in patient care. Originality/value While prior work has explored informed consent for AI, few have specifically addressed the ethical implications of disclosing AI’s values and biases. This paper highlights these biases as a core risk, warranting direct engagement in the consent process to ensure ethically sound use of AI in patient care.