Artificial intelligence is entering clinical care through decision support, diagnostics and patient-facing technologies, bringing new opportunities for information access and shared decision-making alongside ethical risks for patient autonomy. A recent analysis published in BMC Medical Informatics and Decision Making assessed evidence on how AI may support, challenge or reshape autonomous patient decision-making in healthcare. The analysis covered studies published between January 2010 and June 2025 and drew on searches in PubMed, Scopus, Web of Science, Cochrane Library and Science Direct. Fourteen studies met the eligibility criteria after screening and quality assessment. Seven themes structured the results, covering philosophical dimensions of autonomy, human-centred design, cognitive agency, inequality, covert influence, human–AI authority and the need for transparency, dialogue and flexibility.
Must Read: AI Use in Hospital Medicine Needs Better Implementation
Autonomy Extends Beyond Consent
Patient autonomy in AI-enabled care involves more than signing a consent form or receiving information before treatment. It includes the patient’s informed, reflective and voluntary capacity to understand, evaluate and make healthcare decisions that reflect personal values, preferences and life goals. AI affects that capacity when it shapes the information available to patients, mediates clinical recommendations or changes how choices are presented.
The evidence links autonomy to authenticity, agency, rational self-construction and the ability to shape one’s own life. In clinical settings, these concepts translate into meaningful participation in treatment planning rather than passive acceptance of algorithmically generated care pathways. Patients need enough understanding of AI’s role to assess whether a recommendation fits their own preferences and circumstances.
AI can strengthen autonomy when it improves access to relevant information and supports shared decision-making. It can also weaken autonomy when patients accept outputs without critical reflection, when algorithms remain opaque or when care pathways move forward without genuine patient engagement. Autonomy therefore becomes a practical requirement for clinical workflows, not only an abstract ethical principle.
Design Choices Shape Patient Agency
Technology design influences whether AI supports or restricts self-determination. Human-centred design, positive user experience and interfaces that support autonomy, competence and relatedness can make AI-enabled tools more empowering. Models based on motivation, engagement and user experience aim to assess how technology affects basic psychological needs across adoption, interface, task, behaviour, life and society.
Transparent “glass-box” systems offer a stronger basis for autonomous decision-making than opaque “black-box” systems because patients and clinicians can better understand how a recommendation has been generated. Meaningful explanations matter because algorithmic opacity can prevent patients from tracing the reasoning behind a diagnosis or treatment option. Without that understanding, consent, refusal or preference-based modification becomes less meaningful.
Cognitive risks also arise when clinicians or patients rely heavily on AI. Prolonged dependence may weaken decision-making skills and encourage patients to defer to automated recommendations. This risk includes cognitive heteronomy, where thinking and decision-making become externally governed rather than self-directed. In clinical care, that pattern can shift patients from active participants to passive recipients of algorithmically structured options.
Inequality and Influence Remain Central Risks
AI does not affect all patients in the same way. Older adults, economically disadvantaged groups, culturally marginalised populations, people with lower digital competence and patients with limited health literacy may face greater barriers to meaningful engagement with AI-mediated care. Biased datasets may also produce recommendations that disadvantage certain patient groups, limiting fair access to autonomy-enhancing benefits.
Algorithmic categorisation may reduce patients to data profiles rather than recognising individual values and preferences. AI systems designed around individual autonomy may also fit poorly with relational, communal or family-centred decision-making models. Language barriers can add further difficulty when systems are primarily designed in English.
Covert influence creates another risk. Recommendation algorithms, implicit prompts and interface features can guide patients towards particular choices without clear awareness. In healthcare, this can blur the boundary between decision support and decision steering. AI may appear to offer neutral options while framing information in ways that favour certain pathways. Preserving autonomy therefore requires transparency not only about AI’s involvement but also about how options are ordered, framed and presented.
AI creates both opportunities and threats for patient autonomy in healthcare. Its value depends on transparent design, meaningful patient engagement, clinician oversight and governance that protects against opacity, overreliance, covert influence and unequal access. Human judgement remains essential when decisions affect values, preferences and life goals. Autonomy-preserving AI requires clear explanations, adaptable levels of intervention, patient education, clinician training, validated measurement tools and institutional policies that keep AI in a supportive role. The central challenge is ensuring that AI strengthens patient empowerment rather than becoming a mechanism of subtle control.
Source: BMC Medical Informatics & Decision Making
Image Credit: iStock
References:
Mohammadnejad S, Raiesifar A, Bazmi S et al. (2026) Ethical challenges to patient autonomy in the era of artificial intelligence: a systematic review. BMC Med Inform Decis Mak. https://doi.org/10.1186/s12911-026-03614-x
