What is work for? Human wellbeing and the value of work in the AGI era
How do people value their jobs, and how does losing one's job affect wellbeing? We examine the empirical evidence, and what it implies if AI changes the nature of work.
How much do people value the work that they do every day?
Speak with any number of friends, and many would surely volunteer to work fewer hours for the same pay, if they could. Economics often treats work as a ‘disutility’ — something that weighs negatively on overall wellbeing. Yet others love what they do, often because they find their work important and meaningful, for its own sake. What explains the huge divergences in the ways that people relate to their work?
Today, there is rightfully significant public concern about the potential for AI to displace workers, shifting the demand for labor across tasks. Existing evidence suggests that, as AI capabilities improve, the technology will diffuse across the economy, disrupting some forms of work while creating others. This transition brings with it risks of increasing inequalities and stagnation of real wages, as well as declining material wellbeing for displaced workers, as the history of technological shocks suggests. In response to AI’s potential impacts on the labor market, proposals generally emphasize the importance of income transfers. Yet the non-monetary social, cultural, and psychological value of work is often under-explored in these discussions.
In a series of new papers from Google DeepMind, we study the value that work provides people and what it might mean for policymakers. In “What is Work For? Labor, Politics, and Meaning in the Age of AI” (Section 1), we study the non-monetary value of employment through a large representative survey across four countries. Meanwhile, in “Work, Wellbeing, and Choice: Empirical Lessons for AI Futures” (Section 2), we investigate the existing empirical literature on work and wellbeing in real-world transitions (e.g. unemployment, winning the lottery). And in another upcoming paper (Section 3), we explore the ways in which work offers a bundle of benefits, and its implications for human dignity.
Collectively, our research explores how people’s work is intimately connected with values beyond material needs. Beyond ensuring that people's material needs are met, managing an AI economic transition will need to account for the potential loss of agency, socio-economic participation, community, mobility, and sense of purpose. Prior research from our team and others shows that programs guaranteeing incomes only are less popular with American adults. Our newest findings suggest that, if AI induces labor disruption, supplementing incomes is no doubt valuable, but it is by no means sufficient to compensate people for the non-monetary value they derive from their work. We also explore the possibility of alternate sources of wellbeing, and the possibility that societal norms around work — a major driver in the loss of wellbeing that is associated with unemployment — may change over time.
1. What do people say about their work and values?
Through a survey across the U.K., U.S., Denmark, and India, “What is Work For?” measures the non-monetary value that people derive from their employment, and the socio-psychological factors that predict high levels of job attachment. As we can see below, people place significant importance on their job/career, equal to their country and their hobbies, and situated just below their family (note: India’s sample is online-only, and so is over-weighted to “white collar” occupations).
Why do people value their work so highly? The chart below shows that income unsurprisingly ranks as the most important piece of what people look for in their employment. However, across all countries, the vast majority of respondents say that work comes with substantial non-monetary benefits, including a sense of fulfillment and belonging. A majority across all countries also say they get a sense of social status, as well as opportunities to meet people and contribute to society. Broadly speaking, the perception of these non-monetary benefits is present across the income and occupational distribution.
One objection to these results might be that, when workers are actually confronted with an opportunity to work less for the same pay, most people would indeed reduce hours or leave employment altogether. To test this, “What is Work For?” used an auction-style mechanism asking respondents what percentage of their current salary they would need to receive as a guaranteed income in order to be willing to leave their job (shown in the figure below). In the scenario, participants were offered 60-120% of their salaries as guaranteed income, increasing sequentially by 10 percentage point increments. The results suggest that about 20% of survey respondents in the U.S., U.K., and Denmark would not leave their work, even if offered 100-120% of their salaries as guaranteed income. India is again an outlier, perhaps partly driven by the online-only population the survey samples.
And so, a substantial fraction of people clearly value work beyond the wage it pays them — indeed, people may refuse to leave their work even if they earned more income from not working. Yet can we actually attach a value to this non-monetary component? "What is Work For?" investigates this question by providing a way of understanding how people most value their work, classifying workers in four broad buckets: Intrinsically motivated workers comprise one third of the survey sample and find their job inherently meaningful for its own sake—through purpose, learning new skills, or a feeling of contributing to society. Transactional workers (26%) value their employment mostly for pay and benefits. Community-oriented workers (23%) prioritize workplace belonging and socializing with colleagues. And Status-driven workers (17%) find a sense of pride, social status, or recognition in work. As the figure below shows, these profiles are associated with sharp divergences in what people say they would do with their time if they were no longer working in their job and instead received a guaranteed basic income.
When asked about the guaranteed income they would require to voluntarily quit working, intrinsically motivated workers required a higher wage than transactional workers, and they were less likely to quit even at 120% of their salary. Meanwhile, when workers were asked how they would spend their time if they were no longer working but received a guaranteed income, the vast majority of Transactional workers say they would functionally retire and pursue leisure, hobbies, or travel. Meanwhile, about half of Intrinsic and Community-oriented workers would turn toward unpaid volunteering or family caregiving. And half of Status-driven workers would use a guaranteed income as a financial cushion to seek another job or start a business.
In line with prior literature, these results suggest that while money is certainly the major factor for why people work — often the most important one — other components like status, socialization, and purpose matter too. The more a person values these non-monetary pieces, the less willing they are to accept leaving their job in exchange for a substitute income. When designing policy that manages AI-driven disruption, it is important for societies to seriously consider these non-monetary factors.
2. Work and wellbeing in practice
How is human wellbeing affected by real-world work transitions, and what are the implications for AI futures? To understand this question, we thoroughly reviewed existing studies on populations who don't work or who don't need to work. These included the unemployed, retirees, lottery winners, and financially dependent spouses. Work, Wellbeing, and Choice: Empirical Lessons for AI Futures reviewed over a hundred studies from across the world, drawing from psychology, sociology, and economics.
Multiple studies find that the vast majority of lottery winners choose to keep working after receiving enough money to partly or fully replace their income. Remarkably, most lottery winners keep their jobs even after multi-million dollar payouts, more often using their wealth to buy flexibility or slightly reduced hours. Why might people value work beyond income? Consistent with the above survey, studies of real-world job loss show that financial strain is indeed the largest predictor of wellbeing, but non-financial benefits are also significant predictors – especially status, time structure, and collective purpose.
Of course none of this means it’s always bad to stop working. Rather, the evidence suggests a remarkable divergence in the effects on wellbeing from leaving full employment, depending on the context. Becoming unemployed (being available to work but unable to find work) is associated with serious negative effects on mental health, life satisfaction and wellbeing, with strong causal evidence. This includes long-term psychological "scarring" that persists after individuals regain employment. By contrast, retirees report high levels of life satisfaction and wellbeing. What explains this divergence? Retirees differ from the involuntarily unemployed in two different ways.
First, retirees normally choose to stop working. This matters even among retirees — those who retired involuntarily (e.g. because of being laid off) demonstrate far worse wellbeing. Similarly, in studies of working and stay-at-home mothers, what mattered for wellbeing was not whether the mother is working, but whether her work status is in line with what she prefers. Thus, someone’s wellbeing without work depends heavily on whether they chose not to work, reflecting the importance of agency.
A second factor is that there are very different societal expectations and norms for working-age people versus older people. The phenomenon of “retiring from unemployment” is a striking demonstration of the power of norms — when long-term unemployed people become old enough to be officially eligible for retirement, their self-reported life satisfaction often increases, and rates of depression appear to fall. These jumps in wellbeing occur even when income and daily routines stay the same. By reaching retirement age, the stigma and social expectations associated with unemployment seem to be lifted.
Evidence for the importance of societal norms extends beyond retirees. Across countries, the negative wellbeing effects of involuntary unemployment are worse in societies with stronger pro-work cultural norms, and more individualistic cultures. Men appear to be more affected than women on average, perhaps because of stronger social expectations for men to be the “breadwinner”. Conversely, unemployment seems to have a smaller effect when one’s peers (e.g. within a household, or within a region) are also out of work. These results indicate the importance of existing social norms in shaping work transitions, but also the possibility that norms can change (e.g. based on regional unemployment rates).
Are there activities that can provide the same non-financial benefits as traditional market-based employment? Hobbies, caregiving, community organizing, or volunteering can provide many of the same "latent" benefits. State-provisioned employment is another interesting example. At the turn of the century, oil wealth funded widespread state employment in most Gulf States, with nearly two-thirds of working-age adult citizens employed in the public sector (up to 90% in the UAE). Many reported being happy with this arrangement, perhaps thanks to job security and kinder working conditions. For example, 78% of Saudis reported preferring to work in the public sector, often even choosing to stay unemployed while waiting for public sector openings.
| Population | Key finding |
|---|---|
| Lottery winners | The majority choose to keep working, even after multi-million windfalls. |
| Unemployed | Involuntary job loss significantly and lastingly harms wellbeing. Psychological harms are lessened in countries with lower norms to work, or regions with higher unemployment. Wellbeing jumps up after crossing retirement age. |
| Retirees | High wellbeing as long as retirement is voluntary. Alternate activities such as volunteering can raise wellbeing. |
| Financially dependent spouses | Wellbeing is more determined by voluntariness than work status itself. Financial dependence can reduce agency, e.g. difficulties leaving unsatisfactory relationships. |
| Gulf State citizens | High historical levels of well-paid public sector employment (up to 90% in the UAE), with high measures of wellbeing. |
Overall, the existing empirical evidence on work and wellbeing points to a few major takeaways, relevant for managing any potential AI societal transition: (1) We must prioritize agency and choice. (2) We should account for entrenched social norms around work, but also the possibility of changes over time. (3) We may need to furnish alternatives to substitute for the non-financial benefits of work, such as volunteering or state-provisioned jobs. Purely material interventions that ensure a basic standard of living may be necessary, but are ultimately insufficient.
3. Work as an institutional bundle
What should we make of all these interactions between work and wellbeing? One answer is to think of jobs as an “institutional bundle” — serving economic, psychological, and social functions in a single package. AI-driven displacement could unbundle these components, leaving many people without an obvious replacement for the community, sense of purpose, and social standing that they derived from employment. In another forthcoming paper, we discuss the possible erosion of dignity as people are displaced from productive roles through which they develop identities and earn recognition.
The paper draws on decades of in-depth interviews and qualitative research on the meaning of work, from organizational psychology and beyond. The ethnographic research across industries consistently underscores the weight people place on competence, agency, and recognition, and how cultural norms shape what counts as worthwhile work. The same accounts show that work can just as easily leave people feeling devoid of agency, indignified, or devalued, eroding the same psychological needs work can provide at its best.
As mentioned previously, some alternate activities, especially civic engagement and formal volunteering, can provide similar structure, community, and social recognition as meaningful work. A key task for researchers and policymakers is to work out how these non-pecuniary values can be provided across different AGI scenarios, whether by protecting paid work, retraining, strengthening existing civic institutions, or designing new ones.
Transformative AGI scenarios could also disrupt the meaning and recognition people get from their work by decoupling output from effort. Psychologists have documented an "effort paradox": people often intrinsically value activities because they are demanding, using effort to judge if an activity is meaningful. In one experiment, people found their work less meaningful when using AI tools made them feel they had exerted less mental effort. So even if most jobs are spared and the nature of work changes to use AI tools, this could impact the meaning one gets from work and the skills they develop.
Social recognition depends on inferring human qualities from visible outcomes: a polished essay implies intelligence and discipline and a beautiful painting implies a craft developed from years of practice. Once an AI system can produce high-quality work in seconds, the output ceases to signal the presenter’s personal competence, especially if they do not fully understand the work they are presenting. Respect grounded in demonstrated competence then becomes harder to both earn and grant. When human contribution becomes invisible or irrelevant to results, effort loses both its internal meaning and its social legibility.
The challenge may already be visible in fields where AI has developed most. In their recent letter, A Severe Misalignment of AI in Mathematics, a group of influential mathematicians reacted to hundreds of AI-generated manuscripts delivering the correct outcome of proofs while bypassing the collaborative human process of collective understanding, arguing that solving problems has always been a means to build the conceptual intuition and capacity to pose new questions. As the declaration puts it, "The issues the mathematical community faces now are similar to issues that other scientific and creative professions are facing, and indicate issues that all of humanity might face: how to make sure that, as AI changes the way work is done, we do not lose sight of what that work was meant to achieve in the first place."
Managing the AGI transition therefore means taking seriously the risk of the potential loss of dignity from work, and building institutions and policies that maintain or expand the value that work has provided.
The future of work
When we meet someone new, one of the first things we are often asked is, “What do you do?”. For better or worse, our job titles are often taken to describe who we are, what we’re good at, what we care about, and what we’ve achieved.
Of course, plenty of jobs are exploitative, unstable, or tedious. Our work showed that many respondents are willing to take a pay cut just to quit. Indeed, one of the great achievements of developed economies over the past century is the steady decline in how long we must work to live comfortably, and a steady increase in the time we choose to spend in leisure. Yet our team’s research shows that work confers significant non-monetary value to a great many people. Furthermore, people's wellbeing depends substantially on their agency and ability to choose their work status, so that forced work displacement has very different outcomes than voluntary exits from work.
If advanced AI systems develop and diffuse widely across society, however, it would be naive to expect all workers will simply be able to retain employment in the ways that they previously had. As with all prior major technological disruptions, work is likely to change — some jobs will be displaced, others will change, while new types of work will emerge. Wage substitution can compensate people for lost income. Yet as we manage the AI economic transition, we should take seriously what else may be lost — meaning, purpose, recognition, agency — and consider better ways to keep hold of them.
This is the first in a series of pieces that will be published on wellbeing and work.