Unemployment from AI disruption: could it solve the care crisis?

Artificial intelligence may make many people redundant. But with enough preparation, these workers could help to address the growing care crisis. An effective solution would require wealth redistribution, proper training and reforms to ensure that care work is good work.

This article is part of a collection on good jobs by the Policy Hub for the Huth Initiative for a New Political Economy.

There is not yet a consensus on whether the development of artificial intelligence (AI) will bring about large-scale technological unemployment. But the possibility is there (Dobbs et al, 2015; Acemoglu and Restrepo, 2019). For example, some experts argue that automation will put 47% of the jobs in the United States at high risk of redundancy over the next two decades (Frey and Osborne, 2017).

There is consensus about the existence of a care crisis – that is, of a significant gap between care needs and the capacity to meet them (Dowling, 2021). Many societies experience significant shortages of caregivers. It is a growing problem: healthcare systems around the world are short of personnel; the population in many economically developed countries is ageing rapidly (OECD, 2025); and it is hard to see how it will be possible to provide sufficient care for the elderly.

An important aspect of the care crisis is the what’s often called the loneliness epidemic. Around the world, large numbers of people are suffering from social isolation (Surkalim et al, 2022; US Surgeon General, 2023). Some worry that advances in the automation of care work might lead to an excessive decrease in human caregiving (Sharkey and Sharkey, 2012; Santoni de Sio and van Wynsberghe, 2016). Such a shift could further aggravate loneliness.

Both crises have attracted significant attention, but the two debates are not in dialogue with each other. Instead of thinking about these two challenges separately, researchers and policy-makers should address them in tandem, playing each issue against the other. Technological unemployment will make available the necessary workforce that, with sufficient training and labour market reforms, could in part become the care force in the wake of the AI revolution.

The value of work

Research shows that subjective wellbeing depends in part on employment (Donovan et al, 2002). Some objective goods can also be found in work. These include: attaining various types of excellence; making a social contribution; experiencing community; and gaining social recognition (Gheaus and Herzog, 2016).

These goods are central to people’s conception of a successful life. And crucially, they cannot be bought. The first two require long-term involvement in a productive activity, and all of them are such that one can most typically enjoy them in the context of work – whether paid or not.

For these reasons, and because workers themselves value their employment above and beyond the material gains it entails, some researchers argue that states should invest in preserving employment opportunities (see Parr, 2022, for example). But surely it would be best for states to invest in types of work that produce socially necessary goods and services of the kind that either could not or should not be (fully) automated.

Not all care can or should be automated

Better robotics will help to deliver better care. Automating many caregiving tasks can improve quality, including the activities left to be done by people. One argument would be to aim at a new division of labour: between robot care for bodily needs, such as showering; and human care for meeting more sophisticated, relational needs for companionship, affection and mutual understanding.

Robots cannot fully replace personal care – that is, the care that fulfils some of our relational needs. It is certainly possible to imagine machines emulating the behaviours through which people express interest, attention, warmth, consideration and understanding towards one another. But robotic behaviour of this kind is unlikely to be enough to dispel loneliness, let alone to deliver the full value of personal relationships, which depends on receiving affection, concern and companionship from other people.

An attempt to automate caregiving completely – if that were even possible – would involve great loss of value compared with a world in which people are there to meet each other’s relational needs. So, policy-makers should be cautious about how much caregiving is automated. It might be important that even some physical care continues to be provided by other human beings because, while primarily aimed at meeting a material need – showering, for example – it is possible that such care plays just as crucial a role in meeting the emotional needs of the person receiving care.

While the question of how much caregiving it is desirable to automate is far from settled, the AI revolution will not make human care obsolete. We will continue to value having our needs – emotional and, in many cases, physical – met by other people in face-to-face encounters. Insofar as other kinds of work become automated, we might be able to find enough people to do nursing and social work, care for children, the elderly, sick and disabled people, and provide palliative care and psychological therapy.

A fair and efficient world of care work

Many of the professions listed above require qualified care work. So, the creation of a large workforce that can provide such care requires planning. Part of the task of policy-makers is to encourage the creation of opportunities in the form of education for caregiving. Another part of it is to encourage people to avail themselves of such opportunities, both by encouraging sociability in general and by helping to elevate the reputation of caring professions.

Caregivers in many societies are over-burdened, rushed and micro-managed (Dowling, 2021). This means that many care jobs do not provide workers with opportunities to enjoy the non-monetary goods of work. High care loads are detrimental to the job: they deplete workers and make them increasingly unable to do their work well. Care work that is drudgery fails to fulfil relational needs both because it is rushed and because excessive burdens can bear on the caregiver’s motivation.

This affects care recipients too. People resent feeling like a burden on others. When those who need care know that they are putting an excessive burden on the caregiver, the care can be perceived as humiliating for the cared-for.

It is also hard to be the recipient of any genuine attention or affection when the relationship in which these are offered is rushed and micro-managed. Care provided in bad working conditions is unlikely to meet the need for valuable human connection, since that need is only met when the relationship is at least somewhat rewarding, and mutually so.

But if caregivers have control over their work, and caregiving happens at a pace that allows the formation and maintenance of caring relationships, care work can be intrinsically rewarding to the caregiver, as good relationships usually are.

High care loads corrode the quality of work: it is difficult for caring relationships to emerge, and the relational needs of care recipients as well as the goods of work for the worker are mostly compromised: caregivers cannot make a full social contribution, nor develop and exercise their skills fully.

This suggests that sharing care work among more individuals can increase the net goods of work. We have overwhelming reason to create numerous caring jobs, through schemes that would simultaneously solve much of the unemployment problem created by automation and allow workers to partake in (some of) the non-monetary goods of work.

There is a big but to this proposal. The elephant in the room concerns the issue of wealth redistribution. Care work is currently overburdening because it is insufficiently funded. Without redistributing the gains of the AI revolution – as well as wealth more generally – neither states nor individuals will be able to afford to pay for enough, and good enough, care. The fairness and feasibility of the solution outlined in this article depend on significant wealth redistribution to pay for the additional care work.

Even so, the shortage of care workers and the possible unemployment crisis triggered by automation can offer mutual solutions. The current task is to prepare future workers for much needed, as well as potentially rewarding, care jobs. To be rewarding, such jobs would require manageable working loads, worker training and, possibly, a changed ethos of sociability. Policies should be put in place to help with all these factors.

Where can I find out more?

On the care crisis in general: Emma Dowling’s 2001 book The Care Crisis: what caused it and how can we end it?
On demographic change and the ageing of societies: OECD Employment Outlook 2025.
On the loneliness crisis: John Cacioppo and William Patrick’s 2008 book Loneliness: human nature and the need for social connection.
On technological unemployment: the International Labour Organization (ILO) regularly publishes studies of AI-induced unemployment.
On a fuller version of the argument in this article: One crisis to solve another? The place of care in a world of automated work, a recent paper by Anca Gheaus, published in the Journal of Social Philosophy.