AI can care for older people, but not take the place at their side: the line between indirect and direct care

By El Dínamo · October 31, 2026.

Daniela Alegría, director of the Bachelor's program in Philosophy at Universidad Alberto Hurtado, asked at the university's 2nd Artificial Intelligence Week whether generative AI can care for older people, including those with dementia. Her answer, as reported by El Dínamo: it can facilitate indirect care, but it cannot take the place of direct care.

The demographic context she gave is stark. According to the figures she cited, the share of people over 60 will rise from the current 12% to 22% in 2050. The article does not specify what scope this figure refers to. In Chile, she said, 32.1% of the population will be over 60 by that date. Alegría added that women have an average of 0.9 children and that dementia cases worldwide could rise from 55 million to 139 million by 2050.

The academic distinguished three possible roles for AI. The first is the virtual assistant, which carries out tasks the person requests. The second is the cognitive extension, with devices that help people remember to do things. The third is the companion, which offers interaction. On this last one, she warned that it can create dependence, something “very problematic.”

She also separated two types of care. Indirect care includes organizing medication or buying groceries. Direct care covers face-to-face attention, emotional support and physical contact. According to Alegría, AI can monitor, remind and alert the caregiver to possible accidents, but it cannot replace the latter.

She also raised specific ethical dilemmas. If a person is grieving, they may need to go through that distress, but a system monitoring them might choose to play music to cheer them up. She also asked what happens if the older adult does not know they are interacting with a robot. And she warned that delegating too many decisions to these systems can reduce older people's autonomy.

Finally, she noted that in Chile 85% of care work falls on women, and called for shared responsibility among families, communities and the State.

**Our take.** The distinction between indirect and direct care is useful because it defuses a poorly framed debate. The question is not whether AI “replaces” the caregiver, but which part of the burden it can absorb so that humans can devote their time to what truly is irreplaceable. A system that manages pills, detects a fall and sends an alert frees up hours of routine tasks. Those hours can go back to conversation, affection and presence.

This fits a thesis we hold on employment: long-term care is among the work most resistant to automation, precisely because of its relational component. But resistant does not mean immune. The short-term risk is not that a robot will replace a caregiver, but that technology will be used as an excuse to cut human presence where it is already scarce. A cheap conversational companion is tempting for a system with few resources, and there dependence and deception about the nature of the interlocutor stop being an academic dilemma.

The figure of 85% of care provided by women is, in this sense, the key. In general, a huge share of care work today is unpaid and invisible. If AI genuinely eases it, the social benefit would be enormous, as long as that relief reaches those who carry the burden today and not only those who can afford the device.

In the long term, the outlook is hopeful: a long-lived society where technology covers the routine, medicine delays cognitive decline, and people provide care out of connection rather than obligation. Getting there requires designing with users, as the speaker urged, and setting basic rules now: transparency about when one is talking to a machine, limits on decision-making on behalf of the older person, and verifiable human oversight.

It should be noted that this was an academic talk and that the article provides no studies or data from real deployments. It is a solid conceptual framework, not empirical evidence of what AI does today in care homes or private homes.