Ethical considerations in AI monitoring of aged care residents

The first National Elderly Welfare Facility Conference in Tochigi brought together practitioners, researchers, and suppliers to grapple with questions that no longer sit on the horizon. One of the most attended sessions was a panel on the ethical considerations of using artificial intelligence for resident monitoring in care facilities. The discussion drew on Japanese experience while opening the floor to international perspectives, including the regulatory and cultural pressures shaping aged care in Australia.

Australia's sector has been navigating its own reckoning since the Royal Commission into Aged Care Quality and Safety handed down its findings in 2021. The Commission's insistence on dignity, choice, and control has reframed how providers approach technology. Sensors that track movement, bed-exit alerts, acoustic monitors that flag distress, and predictive analytics that flag deterioration are no longer optional extras. They are part of a wider care strategy, and one that must satisfy both the Aged Care Quality Standards and the lived expectations of older Australians and their families.

The ethical terrain is genuinely difficult. AI monitoring can prevent falls, detect pain in non-verbal residents, and give night staff back hours of meaningful contact. It can also quietly erode privacy, flatten individuality, and concentrate decision-making power in vendors and algorithms. The Tochigi panel asked whether the technology is being shaped by care values, or whether care values are being reshaped by what the technology makes easy.

Safety gains and the cost to autonomy

Australian providers have trialled ambient sensors, radar-based fall detection, and wearable patches at scale, partly because workforce shortages documented by the Royal Commission leave staff stretched thin. A night shift in a regional Victorian facility might be staffed by two carers for sixty residents, and a silent bedside alert can genuinely save a life. The temptation to roll such systems out broadly is understandable.

The ethical problem arrives when monitoring drifts from targeted safety intervention to constant observation. Australian ethicists and the Older People's Commissioner have warned that surveillance-by-default can chip away at a resident's sense of self, particularly for those living with dementia who may not understand why a corridor glows softly or why a tablet records their movements. Choice, in the Australian standards, is meant to be meaningful, not theoretical.

Facilities that have navigated this well start with a clinical question rather than a product. A provider in Western Sydney replaced a controversial camera-based falls program with a radar-based system that captures no imagery, after residents and families raised concerns during consultation. The shift kept the safety benefit while removing the lived feeling of being watched. Dignity in care extends beyond monitoring, and the conference exhibition showed how even routine supplies are being reconsidered, with growing interest in biodegradable incontinence products as part of a less clinical, more respectful daily routine.

Consent, capacity, and the question of agreement

Informed consent is rarely simple in aged care, and AI monitoring complicates it further. Many residents live with some cognitive impairment, and the question of who agrees to data collection is contested in Australia as much as anywhere else. The Aged Care Quality and Safety Commission expects providers to attempt consent with the resident first, even where capacity is reduced, before involving a substitute decision-maker.

Cultural background adds another layer. Australia's aged care population includes long-established Greek, Italian, Vietnamese, and Mandarin-speaking communities, alongside First Nations elders whose relationship to data, family decision-making, and institutional care is shaped by very different histories. A blanket English consent form, ticked once on admission, will rarely satisfy the ethical standard these communities are entitled to.

The Tochigi panel emphasised that consent is a process, not a document. Several Japanese facilities described consent conversations that begin before admission and continue through family meetings, with residents able to opt out of specific sensors without losing access to care. Australian providers experimenting with similar models report that quarterly consent reviews do more for trust than any technology upgrade.

Algorithmic bias and whose data trains the system

An AI model is only as fair as the data behind it, and aged care datasets tend to reflect the residents historically studied. That has meant older adults from non-English speaking backgrounds, people with disabilities, and First Nations elders are often under-represented in the training sets used by international vendors. Predictions about pain, falls risk, or cognitive decline can therefore be less accurate for the very populations Australian services are trying to serve better.

Bias also creeps in through how alerts are prioritised. A system tuned to minimise false negatives in one cohort may generate constant false alarms in another, leading to alarm fatigue and, paradoxically, worse care. Researchers at Tochigi pointed to Japanese trials where sensors trained on hospital cohorts performed poorly when moved into smaller group homes, where routines and acoustics differed sharply. The same conference also explored how regional traditions in care matter, with the elderly care menus developed in Tochigi reflecting a wider commitment to local context rather than imposing a one-size-fits-all standard.

The Australian response has been to push for local validation and transparent reporting. University-linked trials in Queensland and Tasmania now publish performance broken down by language background, cognitive status, and frailty level. The hope is that ethical AI means AI stress-tested on the people it is actually used with, not on a generic global average.

Accountability when the algorithm gets it wrong

When a sensor fails to alert, or alerts too often, who is answerable? The Royal Commission made clear that responsibility in Australian aged care cannot be offloaded to a vendor's terms of service. The duty of care remains with the provider, with implications for how AI is procured, documented, and audited.

A workable framework includes clear logging of when an AI flag was raised, who acted on it, and what the outcome was, plus a path for residents or families to challenge algorithmic recommendations. The Older People's Commissioner has argued that any AI used in residential care should come with a plain-language explanation of how it works and what its limits are, written for the resident rather than the procurement officer.

The Tochigi panel suggested ethics committees within facilities, including clinicians, family representatives, and where possible residents themselves, should review AI deployments in the same way they review medication changes. Several Australian providers now treat technology introduction as a clinical governance matter rather than an IT project.

From pilots to lasting practice

Good intentions often falter at the procurement stage, which is why the conference's exhibition conversations mattered as much as the panels. One supplier discussion that resonated was the push for pilot programmes before purchase of any monitoring system, with measurable success criteria agreed up front. Pilots protect both provider and resident, and create space for consent and bias questions to surface before a system is embedded.

The broader ethic running through the conference was that values must drive technology, not the other way around. The lasting question from the panel was not whether AI will monitor residents, but on whose terms. For Australian providers, the answer will probably be measured less in algorithms deployed and more in the quiet, daily evidence that residents feel safer without feeling surveilled, and that families understand what is happening in the rooms they cannot see. That is a standard worth holding onto, even as the technology keeps moving.