Epistemic injustice as a structural threat to AI-enabled mental health systems: towards cognitive sovereignty and neuro-justice in the global south
Year: 2026
Type: article
Abstract: Artificial intelligence (AI) is increasingly incorporated into mental health research, clinical care, population surveillance, decision support, and health-system management. Ethical debates have emphasised privacy, safety, autonomy, transparency, and algorithmic fairness. However, these frameworks do not fully address a prior question: whose knowledge is represented in systems that define, classify, predict, and govern mental health? This paper argues that epistemic injustice is an under-recognised structural threat to AI-enabled mental health systems, particularly where populations are underrepresented in data, research infrastructure, model development, validation, and governance. This conceptual and normative analysis synthesises literature on epistemic injustice, global mental health, cultural psychiatry, AI ethics, computational psychiatry, neuroethics, cognitive liberty, mental privacy, mental integrity, relational autonomy, critical psychiatry, and neurorights. The analysis distinguishes three forms of epistemic injustice—testimonial, hermeneutical, and contributory—and examines how AI may encode, amplify, automate, and institutionalise existing inequalities. Cognitive sovereignty is developed as a complementary normative construct linking epistemic agency, interpretive authority, and protection against technologically mediated determination of mental life. Two illustrative cases; a published clinical encounter involving an Afro-Brazilian child and a constructed perinatal depression screening scenario in Kaduna, Nigeria are used to trace these mechanisms. The paper is not a systematic or scoping review. AI-enabled mental health systems can transform historically situated epistemic inequalities into scalable sociotechnical processes. Testimonial injustice may be embedded in datasets, labels, and predictive systems when credibility deficits are computationally reproduced. Hermeneutical injustice may arise when culturally situated understandings of mental distress are excluded from computational categories and interpretive frameworks. Contributory injustice may persist when affected communities and local researchers are excluded from research priorities, dataset construction, model development, validation, and governance. These processes can become amplified and institutionalised through automated systems and may influence population-level decision-making. Epistemically just AI-enabled mental health requires more than algorithmic fairness. It requires epistemic pluralism, meaningful community participation, equitable research capacity, representation and validity audits, culturally responsive model development, transparent governance, cognitive-sovereignty safeguards, independent epistemic auditing, and equitable benefit-sharing. The paper maps twelve governance recommendations to specific forms of epistemic injustice, mechanisms, and system stages. AI-enabled mental health systems should therefore be understood as components of broader population health systems, making epistemic justice a condition of the validity, legitimacy, and equity of the intelligence they generate. Not applicable
Source: BMC Medical Ethics
Institution: Nasarawa State University
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: gold
Estimated APC paid: $2,290
OpenAlex ID: W7225826390
FWCI:
Cites:
Cited by:
Related to: