Is Using AI to Answer Calls From a Parent With Dementia Ethical? We Think It's a Fair Question.

Is AI ethical in dementia care? It can be, but not because a diagnosis supplies an automatic answer. The better question is whether this particular use has a clear purpose, fits the person and situation, respects participation and privacy, and can be stopped when it is no longer wanted. Families deserve a practical way to think through those questions without turning a difficult relationship into a checkbox.

Start with the person, not the diagnosis

A dementia diagnosis does not automatically establish incapacity. Decision-making ability varies between people, can depend on the specific decision, and may change over time. A person may be able to express a preference about a familiar voice, understand part of an interaction, or make some decisions while needing help with others.

Involve the person in decisions about technology to the extent appropriate to their abilities, preferences, and circumstances. Notice whether they want to interact, whether they show discomfort, and whether their view changes. If the person does not want to use the system, refusal is important information, not a problem to overcome.

An editorial decision framework

This is a practical editorial framework, not a validated clinical or legal test. Use it to ask better questions:

1. Purpose: What specific problem is the technology intended to address?
2. Proportionality: Is it proportionate to the actual need, or would a simpler option do?
3. Transparency: What does the person understand about the technology and the interaction?
4. Preference and assent: Does the person appear willing to interact with it, and can they change their mind?
5. Decision-making authority: Who is authorized to make which decision if assistance or substitute decision-making is actually required?
6. Privacy: What information is collected, who can access it, and when is it deleted?
7. Deception risk: Does the system intentionally cause the person to believe something materially false?
8. Benefit and burden: Does it appear useful for this individual without unnecessary distress, confusion, intrusion, or burden?
9. Exit: Can the person or family stop using it promptly and easily?

Should someone with dementia know they are talking to AI?

There is no single answer that applies to every person or situation. Transparency, the person’s understanding and preferences, the nature of the interaction, the risk of deception, and any applicable decision-making authority all matter.

That does not mean a person must understand every technical detail before any assistive technology can be considered. It does mean that KindredMind should not rely on a claim that dementia removes a person’s interest in transparency or participation. Deliberate deception is not a universal KindredMind policy, and dementia-care ethics does not provide a blanket rule that hiding AI identity is always right.

Families should consider a proportionate, person-centered approach: explain the system in a way the person can understand, invite questions and preferences, watch how the interaction is experienced, and revisit the choice. If disclosure or the interaction itself creates confusion or distress, pause and seek appropriate care guidance rather than assuming that either concealment or disclosure is always correct.

Consent has more than one layer

Creating or managing an account is not the same as every form of consent. Keep these roles distinct:

• Account holder: the person configuring and managing the service.
• Voice owner: the person whose voice is used; that person should provide appropriately documented consent for the voice use.
• User: the person interacting with KindredMind; their willingness, preferences, and assent matter.
• Authorized substitute decision-maker: an appropriately authorized decision-maker where applicable, for a specific decision when assistance or substitute decision-making is actually required.

A spouse, adult child, caregiver, or account holder does not automatically have legal authority to make every decision for another adult. This page does not make legal determinations. If authority is uncertain, obtain advice appropriate to the relevant jurisdiction and circumstances.

What the evidence does—and does not—show

Research on simulated presence therapy has examined particular interventions, populations, settings, and outcomes. For example, the cited randomized trial studied simulated presence therapy and reported differences in agitation, anxiety, and use of protective restraint in its studied population and comparison—not KindredMind, cloned voice, or interactive AI conversation. Those findings should not be presented as product outcomes.

A recorded familiar voice is not automatically equivalent to a cloned familiar voice, and a cloned familiar voice is not automatically equivalent to interactive AI conversation. Research on one does not establish that another creates trust, comfort, recognition, compliance, or any other clinical result. KindredMind may be informed by related familiar-presence research, but does not claim to reproduce a studied clinical mechanism.

Where KindredMind fits

KindredMind can provide responsive conversation using a caregiver-approved familiar voice and family-provided context. Families can also use a screen-based Presence option or scheduled voice and SMS reminders where those features are available. These are support options, not a clinical intervention, emergency response, or a replacement for human relationships and professional judgment.

Review what the person wants, what information the system needs, what it cannot do, and who will monitor the experience. Keep an ordinary way to reach a human helper, and turn the service off if it creates more burden than value. For a broader comparison of fit, information use, evidence, and exit options, read how to evaluate dementia care technology.

Our answer

AI in dementia care is not automatically ethical or unethical. A defensible choice is specific to the person, decision, time, purpose, and safeguards. Start with participation where possible, distinguish the different consent and authority questions, be honest about limitations and evidence, protect privacy, and keep stopping easy.

Frequently asked questions