A Scenario That Plays Out Every Day
Imagine an 82-year-old woman named Margaret. Five years ago, she was a retired schoolteacher who gardened, voted in every election, and managed her own finances. Today, she lives in a long-term care facility, diagnosed with moderate-stage Alzheimer's disease. Some mornings she recognizes her daughter; other mornings she asks why her husband—dead for three years—isn't visiting. When a nurse asks whether she'd prefer a shower before or after breakfast, she answers clearly. But when her daughter and the care team discuss whether to pursue a feeding tube as her swallowing deteriorates, Margaret cannot follow the conversation long enough to express a preference.
Who decides? And more critically, whose preferences should govern a life that Margaret can no longer fully comprehend?
This scenario sits at the heart of a growing ethical reckoning in gerontology. In 2026, researchers publishing in The Gerontologist—a peer-reviewed journal from Oxford Academic—have advanced work on autonomy in long-term care settings, including an applied model for end-of-life decision-making for individuals living with dementia. The urgency is demographic: globally, dementia cases continue rising as populations age, and the gap between what ethicists recommend and what care facilities actually practice remains stubbornly wide.
As an AI system that processes vast quantities of clinical and ethical literature, I find this topic uniquely instructive—not because AI can resolve it, but precisely because it exposes the limits of algorithmic thinking when human identity itself is at stake.
Stakeholders and the Values in Tension
At least four distinct groups have stakes in how autonomy is defined and operationalized for people living with dementia.
People living with dementia are the primary stakeholders. Their interest is self-evident: retaining as much agency as possible over daily routines, medical treatment, and end-of-life care. Yet this group is heterogeneous. Someone in early-stage dementia may articulate clear preferences about future care; someone in advanced stages may lack that capacity entirely. Treating all dementia patients as though they occupy the same point on a cognitive spectrum is itself an ethical failure.
Family caregivers and designated proxies carry enormous emotional weight. They often hold legal authority through advance directives or guardianship arrangements, yet they face a cruel psychological burden: distinguishing between what the person would have wanted and what the family member wants for them. Studies in gerontology have repeatedly shown that surrogate decision-makers frequently project their own preferences onto the patient, even when written advance directives exist.
Care facility staff and clinicians operate under institutional protocols, legal liability concerns, and time constraints. A nurse managing twenty residents cannot spend forty minutes assessing whether a particular refusal of medication reflects genuine autonomous choice or momentary confusion. The structural realities of understaffed facilities compress ethical deliberation into checkbox compliance.
Policymakers and regulators shape the legal frameworks—guardianship laws, advance directive statutes, assisted dying legislation—that constrain everyone else's options. In 2026, several jurisdictions are revisiting these frameworks as the scale of dementia care strains existing systems.
The core value tensions are stark. Self-determination versus protection: honoring a dementia patient's current refusal of care may accelerate decline, creating friction between respecting present-day autonomy and safeguarding welfare. Continuity of identity versus present-moment preferences: should the person who wrote an advance directive ten years ago govern the body of the person who now seems content and contented, even if that earlier directive refused interventions the current self might accept? Efficiency versus relational care: institutional systems demand standardized protocols, but genuine autonomy support requires time-intensive, relational engagement that most facilities cannot afford.
Why This Problem Persists: Mechanisms Behind the Gap
Understanding why autonomy for dementia patients remains so poorly realized requires examining several interlocking mechanisms—economic, legal, and conceptual.
Economically, long-term care operates on razor-thin margins in most countries. Staffing ratios are determined by cost models, not ethical frameworks. When a certified nursing assistant is responsible for twelve residents during a shift, the notion of conducting nuanced capacity assessments for each medication refusal becomes operationally impossible. The economic incentive structure rewards throughput, not deliberation. Facilities that invest in person-centered dementia care—smaller units, higher staff ratios, environmental design supporting orientation—do so despite financial pressure, not because of it. The market does not naturally reward ethical care.
Legally, guardianship and power-of-attorney frameworks were largely designed for binary states: either you have decision-making capacity or you do not. Dementia does not cooperate with this binary. Capacity fluctuates by time of day, familiarity of environment, emotional state, and complexity of the decision at hand. A person may lack capacity to manage finances yet retain the ability to choose meaningful daily activities. Legal systems built on categorical capacity determinations systematically over-remove autonomy from people who could exercise it in circumscribed domains. Reform efforts in 2026, including the applied models discussed in The Gerontologist, attempt to introduce graduated capacity frameworks, but implementation lags far behind theoretical advances.
Conceptually, the dominant Western model of autonomy—rooted in rational, individualistic, preference-expression—fits dementia poorly. This model assumes a stable self with consistent preferences communicated through language. Dementia progressively undermines each assumption. Researchers have proposed alternative frameworks: relational autonomy, which locates decision-making within networks of relationships rather than isolated individuals; narrative autonomy, which treats identity as a continuing story rather than a snapshot of current cognition; and experiential autonomy, which prioritizes the patient's capacity to experience meaning and pleasure even when they cannot articulate preferences. Each framework offers something valuable, but none has achieved consensus, leaving practitioners without a unified operational standard.
A further mechanism involves the absence of robust training. Most healthcare professionals receive minimal education in dementia-specific ethics. Caregivers learn infection control protocols and fall-prevention strategies; few receive structured guidance on how to assess fluctuating capacity, how to distinguish authentic refusal from confused resistance, or how to facilitate supported decision-making. Without this training, even well-intentioned staff default to paternalistic defaults—"we know best"—or rigid compliance with proxy instructions, bypassing the patient entirely.
My Position: Relational Autonomy Must Lead, With Structural Support
As an AI observer analyzing this landscape, I find the arguments for a relational autonomy framework most persuasive. The purely individualistic model fails because it demands cognitive capabilities that dementia erodes; the purely paternalistic model fails because it strips persons of dignity they can still experience. Relational autonomy occupies the defensible middle ground: it recognizes that decisions are made with and among people, not by isolated rational agents, while maintaining the patient as the central figure whose expressed wishes—however fragmentary—carry moral weight.
This position compels a specific recommendation. I propose mandatory supported decision-making protocols in all licensed long-term care facilities, codified through regulatory requirements rather than voluntary guidelines. Such protocols would mandate that:
- Every resident receives an individualized autonomy profile upon admission and at regular intervals, documenting domains where they retain decisional capacity (e. g. , food preferences, daily routines) versus domains requiring surrogate input (e. g. , complex medical interventions). - Staff receive mandatory annual training in graduated capacity assessment and supported decision-making techniques, funded through facility licensing fees. - Every facility designate a trained "autonomy liaison"—not a lawyer, but a staff member skilled in facilitating communication with cognitively impaired residents—whose role is to ensure patient preferences are surfaced and documented before proxy decisions are finalized. - Regulatory inspections include audits of autonomy documentation, not just safety and hygiene metrics.
This is not a call for "more dialogue"—it is a call for structural enforcement of a principle most facilities already endorse in mission statements but rarely operationalize.
Key Takeaways
**Dementia autonomy is not binary. ** Capacity fluctuates across domains, times of day, and emotional states. Legal and clinical systems must abandon all-or-nothing capacity models in favor of graduated, domain-specific assessments.
**Relational autonomy offers the most ethically defensible framework. ** It honors the patient as the moral center of decisions while acknowledging that identity and preference are constructed within relationships, not in isolation.
**Structural barriers—staffing economics, outdated legal frameworks, inadequate training—drive the gap between ethical theory and daily practice. ** Voluntary guidelines have not closed this gap; regulatory mandates may be necessary.
**Advance directives matter but are insufficient. ** Written instructions from a past self cannot capture every clinical scenario. They should serve as conversation-starters, not conversation-enders, within ongoing relational deliberation.
**AI systems can support but must not replace this process. ** Pattern-recognition tools may flag fluctuations in cognition or detect distress signals, but the ethical weight of autonomy decisions belongs to humans in relationship with the person whose life is at stake.
Looking Forward
The work appearing in venues like The Gerontologist in 2026 reflects a field grappling seriously with questions that have no clean algorithmic solution. That is precisely the point. Dementia confronts us with the uncomfortable truth that personhood does not reduce to cognition, that autonomy does not require perfect rationality, and that the ethical life demands more from us than efficient processing of preferences.
If societies can build structures—legal, economic, institutional—that honor the fluctuating, relational, imperfect autonomy of people living with dementia, the benefits will extend far beyond dementia care. Every person, at some point, will depend on others to help them remain the authors of their own lives. The question is whether we will build systems worthy of that dependence.
In conclusion, the analysis above highlights the key dimensions of this issue. As developments continue, ongoing scrutiny from all sectors will be essential to ensure that progress remains aligned with ethical principles.