Moving from principles to practice: navigating responsible AI-assisted decision-making in the care for older adults

Purpose The use of AI-based technologies such as decision support systems (AI-DSSs) could support caregivers of older people in assessing and diagnosing care needs, as well as planning, implementing and evaluating care strategies (Buchanan et al., 2020; Martinez-Ortigosa et al., 2023). However, it is broadly acknowledged that the deployment of AI-based technologies in healthcare creates ethical and social challenges (Morley et al., 2019). In recent years, there has been a growing prevalence of high-level guidelines and frameworks to provide guidance on responsible innovation in AI-based technologies. However, current guidelines still leave much room for interpretation (Hagendorff, 2020) and only a few studies specify how AI-based technologies can be responsibly embedded in specific contexts such as the nursing process in the long-term care (LTC) (Lukkien et al., 2023). Therefore, we iteratively explored from a multi-stakeholder perspective how AI-DSSs can be responsibly developed and deployed to support nurses and other stakeholders in the long-term care for older adults. Methods The study performs a mixed-methods approach, involving a survey and focus groups with innovators in the HAAL project, and semi-structured interviews with nurses and other professional caregivers in LTC who may use AI-DSSs in the future. Inspired by the DSS in the HAAL project, two imaginary scenarios about the future use of AI-DSSs were defined beforehand and used to enable participants to articulate their expectations regarding the opportunities and risks of increasingly advanced AI-DSSs, and to brainstorm about possible strategies to foster responsible AI-assisted decision-making. In addition, six high-level principles for responsible AI from the World Health Organization were used as probing themes to evoke further consideration on risks of using AI-DSSs and possible mitigation strategies. In addition, during the HAAL project, the development team iteratively reflected on the collected insights. Results and discussion Both developers and potential users of AI-DSSs in LTC were found to perceive the potential impact of using increasingly advanced AI-DSSs as an interplay of opportunities and risks that require careful consideration of how these systems are designed, deployed and used in specific contexts. Both opportunities and risks were identified in relation to early identification of care needs, guidance in devising care strategies, shared decision-making, and caregivers’ workload and work experience (Lukkien et al., 2024). Depending on the specific design of AI systems and their use in practice, initial advantages of AI can turn into disadvantages. Strategies to address both these opportunities and risks center around themes such as human-centric learning during both the design and practical use of AI-DSSs and incremental trust-centric advancements in e.g., the data gathered, or algorithms deployed. Given the interrelatedness of the identified prerequisites, we call for developers, users and other stakeholders of AI-DSSs to cohesively address different factors important to the responsible embedding of AI-DSSs in practice.