Designing Ethical Artificial Intelligence-Driven Mental Health Applications: A Cross-Contextual Framework Informed by Developers, Digital Ethics Experts, and Digital Health Specialists in Switzerland and Nigeria

Abstract Background Artificial intelligence (AI)-driven mental health applications are proliferating rapidly across high-income and low- and middle-income countries (LMICs), yet the ethical governance of their design and deployment remains inadequately operationalised. Existing frameworks are largely principle-based and fail to address how ethical obligations are understood and implemented by those building and evaluating these systems. Objective To develop an empirically grounded, context-sensitive ethical design framework for AI-driven mental health applications through qualitative stakeholder research. Methods Ten semi-structured interviews were conducted with three stakeholder groups: mental health app developers, digital ethics experts, and digital health specialists in Switzerland and Nigeria. Data were analysed using iterative thematic analysis guided by six conceptual domains: transparency, privacy, algorithmic fairness, safety, accountability, and contextual adaptation. Dual ethical approval was obtained from the University of Zurich Ethics Committee (Review No. MeF-Ethik-2026-10) and IAMRAT, Nigeria (Protocol No. UI/EC/26/0187). Results Eight themes emerged. AI was conditionally accepted as a bounded, supportive tool rather than an autonomous clinical authority. Human oversight was the strongest cross-stakeholder consensus. Privacy was universally prioritised but governance maturity varied. Fairness required both technical performance parity and contextual cultural fit. Accountability remained fragmented in practice. Nigeria-based participants emphasised scalability, multilingual access, and infrastructure constraints, whereas Swiss participants prioritised regulatory compliance, evidentiary standards, and validation. Conclusions A 'core-plus-context' model is proposed: a universal ethical foundation comprising privacy, fairness, transparency, safety, human oversight, and accountability, coupled with context-sensitive implementation. An eight-pillar ethical design framework is presented to guide developers, clinicians, and policymakers toward responsible AI-driven mental health innovation across diverse global settings.