The core of their argument is deceptively simple: AI does not merely respond to existing challenges in long-term care, it actively reshapes them. By determining which data are collected, which risks are classified, and which outcomes are prioritized, AI systems influence how care needs are understood and addressed in the first place. Ethical risks, the authors argue, arise not only from technical performance failures but from something far subtler: the possibility that structural and relational problems in care systems are reframed as individual conditions that can be monitored and managed through technology. When loneliness, neglect, or inadequate staffing become data points on a dashboard, the underlying causes can escape scrutiny entirely.

These risks operate at several interconnected levels, the authors explain. At the individual and relational level, AI may affect autonomy, dignity, privacy, consent, and the relationships between older adults and their caregivers. AI-enabled systems can collect and process sensitive data without sufficient transparency about how that information is used, protected, or retained. Excessive reliance on AI may also weaken the human connection and reciprocity that lie at the heart of care itself. At the level of social justice, biased data, opaque systems, and weak accountability may deepen existing inequalities, particularly for older adults living with frailty, socioeconomic disadvantage, or limited AI literacy. And at the environmental level, scaling data-intensive AI carries material costs through energy consumption, water use, mineral extraction, and the expansion of data infrastructures.

Perhaps the most provocative claim concerns the cultural and epistemic level. Categories such as dependency, risk, efficiency, and quality of care, the authors stress, are not neutral descriptors. They reflect assumptions shaped by medical, administrative, commercial, and policy institutions, and those assumptions can become embedded in datasets, classification systems, platform design, and performance indicators. Decisions about what data to collect and which outcomes to optimize therefore encode particular ideas about normal aging, desirable behavior, appropriate care, and where responsibility for care should lie. The authors warn that these embedded assumptions may reinforce ageist and deficit-oriented representations of older people while marginalizing diverse experiences of aging and social relations. In this process, older adults risk being reduced to data subjects or measurable profiles, while their identities, relationships, life histories, and social belonging receive less and less attention.

To make this concrete, the authors offer a hypothetical scenario that reads like a cautionary tale for care facilities everywhere. Imagine an AI-powered chatbot introduced to reduce loneliness among residents of a long-term care home. In an underfunded facility facing chronic workforce pressures, staff may come to rely on the chatbot to monitor residents’ emotional states, flag signs of social isolation, and initiate conversations. The facility may then treat the technology as a substitute for investment in staffing, communal activities, and sustained human relationships. The system itself may rest on datasets that reflect narrow assumptions about normal social interaction, potentially misinterpreting culturally diverse forms of communication and connection. Its collection of personal data creates privacy and governance risks, while the computing infrastructure behind it carries environmental costs. Loneliness, in this scenario, has been reframed as an individual condition to be detected and technologically managed, while the organizational conditions that produced it remain untouched.

Existing regulation, the authors acknowledge, has made real progress. The European Union’s AI Act has established a risk-based approach to governing artificial intelligence, and China has been building an expanding framework for AI industry standards and technical standardization. These governance frameworks provide important safeguards. Yet compliance-oriented governance, Xu and Teerawichitchainan argue, tends to focus on the capabilities and risks of individual AI systems, framing responsibility mainly in terms of technical compliance rather than institutional responsibility for staffing, social infrastructure, and equitable care. Such approaches may leave unexamined the most important questions of all: why particular AI systems are adopted in the first place, which problems they are expected to solve, and how organizational arrangements, power relations, and sociocultural assumptions shape their use.

The authors’ proposed remedy draws on systems thinking, a framework pioneered by researcher Donella Meadows, whose influential work on leverage points described the places where intervention can alter how an entire system behaves. Systems thinking understands complex challenges as outcomes of dynamic relationships among technical systems, organizational arrangements, professional practices, and societal values. The key insight is that changes to technical parameters, such as improving model accuracy or tightening consent procedures, may reduce particular risks but leave the system’s underlying goals and structures intact. Deeper interventions instead seek to change how problems are defined, whose knowledge carries authority, what organizations are designed to achieve, and who holds the power to redirect the system. Applied to AI-enabled long-term care, this means that the most powerful levers are not in the code but in the context surrounding it.

From this perspective, the authors identify four domains in which deep leverage points may operate. The first encompasses societal paradigms and value systems at the level of policy and public discourse, which define what counts as aging well, technological progress, and quality of care. The second concerns epistemic authority in knowledge production and technical design, determining whose knowledge is used to define care needs, relevant data, and desirable AI outputs. The third involves systemic goals and models of AI-enabled care at the organizational level, which determine whether AI strengthens relational care or primarily advances efficiency, surveillance, and risk management. The fourth addresses distributed decision-making power and user agency, asking whether affected groups can shape, contest, modify, or discontinue the AI systems that touch their lives.

Each domain suggests a corresponding intervention pathway. At the policy level, the authors argue, ethical AI governance requires challenging the assumption that care challenges are best addressed through greater prediction, automation, and individual risk management. Quality of care could instead be defined in terms of dignity, reciprocity, inclusion, sustainability, and opportunities for meaningful social connection. In knowledge production, participation should extend well beyond token consultation: older adults, family caregivers, care workers, and community members should have a meaningful role in defining the problems to be addressed, determining which data are relevant, prioritizing outcomes, and deciding whether an AI application is needed at all. At the organizational level, procurement, funding, and evaluation arrangements should reward uses of AI that strengthen relational care and support the workforce, while facilities should assess whether AI applications redirect resources away from human care, intensify surveillance, or transfer institutional responsibilities onto residents and caregivers. Finally, affected groups need practical mechanisms to question, modify, suspend, or discontinue AI applications.

The article’s ultimate message is both a warning and an invitation. AI is transforming not only how long-term care is delivered but how care needs, social problems, and desirable outcomes are understood. Governance should therefore look beyond whether individual applications are safe or compliant and examine the care contexts into which they are introduced. This demands ongoing multidisciplinary learning about how workforce conditions, organizational priorities, social relationships, and inequalities shape AI implementation, and how AI, in turn, reshapes those conditions. Given the limited evidence on long-term effects, the authors conclude, governance should remain precautionary and adaptive. Most importantly, AI should never be allowed to reframe structural and relational problems as individual conditions to be technologically managed, particularly when doing so may reproduce underinvestment, weaken care relationships, or leave inadequate social infrastructure unaddressed. The future of ethical AI in elder care, in other words, depends less on smarter machines than on wiser choices about what we ask those machines to do.

Subject of Research: Ethical and responsible governance of artificial intelligence in long-term care for older adults

Article Title: Deep leverage points for ethical and responsible AI in long-term care

Article References: Xu, W., & Teerawichitchainan, B. (2026). Deep leverage points for ethical and responsible AI in long-term care. PLOS Aging and Health, 1(3), e0000041. https://doi.org/10.1371/journal.page.0000041

Image Credits: AI Generated

DOI: 10.1371/journal.page.0000041

Keywords: artificial intelligence, long-term care, aging, AI ethics, systems thinking, leverage points, elder care, AI governance, care workforce, ageism, loneliness, health policy