JS Festival in Tochigi 1st National Elderly Welfare Facility Conference

Ethical AI In Care Planning: Lessons From JS Festival

The Panel on Ethical Considerations for Using AI in Care Planning was one of the most relevant conversations at the 1st National Elderly Welfare Facility Conference and Research Meeting, known as JS Festival in Tochigi. Held for professionals across Japan, the event examined how digital tools can support older people without turning care into a purely technical exercise. Learn more about Contact.

The issue matters well beyond Japan. Australian providers are adopting electronic care records, predictive analytics, speech-to-text tools and generative AI while responding to tighter expectations around privacy, dignity and clinical accountability. In Melbourne, regional Victoria or remote Queensland, the central question remains the same: who makes the final care decision, and how can an older person challenge it?

JS Festival brought together care workers, facility leaders, researchers and suppliers, creating a useful setting for that debate. Research presentations, practical workshops, study tours, networking and an exhibition allowed participants to compare policy with what actually happens on a ward, in a residential home or during a home-care visit.

Why AI Needs An Ethical Care Framework

AI can help staff identify changing risks, summarise progress notes and organise information for a person-centred care plan. It may flag a pattern of missed meals, reduced mobility or social withdrawal earlier than a busy team member can. Used carefully, this can give nurses and support workers more time for conversation and hands-on care.

Yet a recommendation is not a diagnosis, and a probability is not a person’s preference. Older people may have dementia, communication difficulties, limited digital confidence or a first language other than English. A system trained on incomplete records can reproduce bias, overlook cultural needs or treat a family’s concerns as secondary evidence.

Australian organisations also need to consider the diversity of the aged-care system. A provider in Sydney may work with multilingual families and complex hospital transitions, while a small service near Cairns may face limited staffing and connectivity. For Aboriginal and Torres Strait Islander older people, culturally safe practice and community authority must remain central rather than being reduced to fields in a database.

Consent, Privacy And Human Choice

Consent for AI-supported planning should be specific, understandable and ongoing. An older person should know what information is collected, why it is used, who can see it and whether declining the tool will affect access to care. Substitute decision-makers can help when capacity is impaired, but their involvement should not erase the person’s known wishes.

Privacy controls need to cover more than the software itself. Providers should examine cloud hosting, supplier access, retention periods, cybersecurity and secondary use for product development. The Australian Privacy Act, health-record obligations and the strengthened expectations arising from the Royal Commission into Aged Care Quality and Safety all point towards clear governance and accountable documentation.

The same instinct that encourages people to inspect terms before using an online service can support better digital procurement; a plain-language deposit terms guide illustrates how visible conditions help people understand what they are agreeing to. In care, that transparency must be adapted to health information, with large-print explanations, interpreters and time for family yarning where appropriate.

Designing Fair And Explainable Recommendations

An ethical care-planning system should offer reasons that a worker can understand and communicate. “The model says so” is not an acceptable explanation for restricting movement, changing medication support or reducing personal choice. Staff need to see which information influenced a recommendation, how reliable it is and when it should be disregarded.

Fairness testing should compare outcomes across age groups, genders, cultural backgrounds, disability types, languages and locations. Australian providers should be especially alert to data gaps affecting rural and remote communities, where a small sample can make a model appear confident without being dependable. A tool that performs well in a large metropolitan service may need substantial validation before use in the bush.

Practical safeguards can make these principles part of ordinary work:

These measures preserve professional judgement. They also recognise that good care involves observation, rapport and context that may never be captured in structured fields. A worker who knows that an older resident is grieving, frightened or simply having a rough day may interpret the same data very differently from an algorithm.

Building Accountability Across The Care Team

Responsibility cannot sit solely with a software vendor. Facility boards, managers, clinicians, care workers and suppliers each need defined duties. A provider should know who approves a system, who monitors performance, who reports an incident and who can turn it off. Contracts should address audit rights, data ownership, model changes and notification when performance shifts.

Training should be practical rather than a one-off compliance module. Teams can work through examples involving falls, nutrition, social isolation and behaviour support, then compare the AI output with their own assessment. They should learn how automation bias works: people may accept a computer-generated suggestion too quickly because it looks objective or saves time.

The JS Festival’s mix of research and hands-on sessions is a useful model for this kind of learning. Professional events work best when participants share what went wrong as well as what worked, and the conference’s approach to lasting professional connections reflects the value of relationships after delegates return to their services.

A local Australian trial might involve a small group of residents, family representatives, direct-care staff and an independent reviewer. Feedback should be gathered in everyday language, including “fair dinkum” concerns from workers who feel a system adds paperwork or undermines trust. Their experience is evidence, not resistance to innovation.

Measuring Outcomes Without Losing Humanity

Evaluation should ask whether AI improves life, not merely whether it processes information efficiently. Useful measures may include unplanned hospital transfers, response times, medication errors, continuity of staff, resident participation and family confidence. Qualitative accounts matter too: a resident feeling heard can reveal an outcome that a dashboard misses.

Research on social connection is particularly relevant. Evidence from peer volunteer programmes shows why human relationships should remain visible in care planning. If an algorithm identifies loneliness, the response should not be limited to a notification; it might lead to a visitor, group activity, culturally familiar event or regular call from someone the person trusts.

Decision-makers can use a balanced set of checks before expanding an AI tool:

For Australian audiences, this means connecting innovation with the practical realities of My Aged Care, home support, residential services and the NDIS where systems overlap. A technology may be impressive in a demonstration yet unsuitable for a provider with high workforce turnover, limited training time or unreliable internet. The best standard is whether it strengthens relationships, choice and safety.

JS Festival in Tochigi demonstrated why an interdisciplinary gathering matters. Researchers can test assumptions, practitioners can expose operational realities, administrators can address governance and suppliers can hear directly from the people affected by their products. The discussion around AI in care planning belongs in that shared space, where efficiency is considered alongside dignity, culture, accountability and the right of every older person to remain more than a data profile.