“AI is one of the most important things humanity is working on,” Google CEO Sundar Pichai told a town hall audience earlier this year. “It is more profound than, I don’t know, electricity or fire.”
A few numbers support the hyperbole: Venture capital firms committed more than $12 billion to AI startups last year, double the amount they invested in 2016, according to KPMG. And 80% of large companies have launched AI projects. Behind the investments are some bullish economic forecasts: According to McKinsey Global Research, AI could contribute an additional $13 trillion to global GDP by 2030. That would put AI on par with the economic impact of steam in the 19th century and robots in manufacturing in the 20th.
As with many innovations, though, AI brings potential costs, particularly for workers. In the U.S., a McKinsey study estimates that by 2030 up to 73 million jobs could be eliminated through AI and automation. Even accounting for new job creation, as many as 54 million workers will need to be retrained for entirely new occupations.
While analysts crunch numbers and CEOs debate about how AI will reshuffle the modern labor force, employers investing in the new technology should consider an important truth: Work environments are already proving to be hazardous to employees’ health and well-being in the 21st century—and businesses should do more to create a healthier workplace (see sidebar for recommendations on how to do so).
There are many indicators that explain why health has been declining—and why it tends to be ignored. First, labor-market arrangements and workplace practices profoundly affect people’s health and thus contribute to the worldwide problem of soaring health care costs and rising inequalities in lifespans and other health indicators. As I’ve argued in my new book, Dying for a Paycheck, substantial epidemiological evidence demonstrates a direct and important connection between work environments and people’s physical and mental health and mortality.
Income inequality also has adverse health effects on the workplace. Average life expectancy in the U.S. has actually decreased over the last two years, in part because people at the top of the income distribution are enjoying longer life, while for those at the bottom, life expectancy is going down.
Second, while job satisfaction in the U.S. has rebounded since the depths of the 2008–09 recession, just 50% of workers report being satisfied with their jobs today. Eighty-five percent of employees say they are either not engaged or actively disengaged with their jobs, according to a Gallup worldwide survey, at a cost to the global economy of $7 trillion in lost productivity.
Worker well-being is under stress. Gallup has created an index based on five metrics that comprise employee well-being: purpose, social (having supportive relationships), financial (security and reduced stress), community (liking where you live and feeling safe), and physical (having good health).
In 2017, well-being scores in nearly half of U.S. states declined by significant margins, and for the first time since Gallup began tracking the metrics, no state saw a statistically significant improvement. This is the state of affairs as thousands of companies have started experimenting with AI and other forms of automation that promise even more workplace dislocations.
Here is the reality check you don’t hear often at technology conferences these days: Absent major public-policy innovations and interventions—which have been exceedingly, astonishingly rare in modern history—empirical evidence suggests AI will likely make workers’ situations worse before it makes them better, if it indeed it ever does.
The Macro View Of Labor Market Adjustment
The counter-argument to these dire warnings about the effects of automation and AI on people is that proper policy interventions can mitigate or even completely remediate many of the anticipated negative impacts.
In theory, this is true. Investments in education and job training can help people transition to new jobs and new occupations. Income maintenance policies, including unemployment insurance, guaranteed income and other social welfare interventions can mitigate the harmful effects of income reduction or the loss of income arising from job displacement.
According to a 2018 McKinsey study, 62% of U.S. corporate executives say they will need to retrain or replace a quarter of their current workers over the next five years because of automation. And some companies are stepping up to the task. PriceWaterhouseCoopers has a newly launched Digital Accelerator program, part of the company’s ambitious effort to “upskill” its 50,000-person U.S. workforce.
Therefore, the question isn’t whether policy interventions are possible in principle. The real issue is what are the prospects for such policies and programs actually being implemented?
With some exceptions, such as Germany, which has implemented shared working subsidies to avoid layoffs and has an active apprenticeship training effort, advanced industrialized countries have not historically done a great job with previous job dislocations coming either from recessions or from the transition from extractive industries and manufacturing to services. Now these countries face the task of helping with the transitions necessitated by AI and, for the most part, have less capacity to do so.
Education and training
If education and training are to be an antidote to the adverse effects on labor from artificial intelligence and automation, the global historical record on investment in human capital development is mixed at best. In the U.S., public disinvestment in higher education began more than three decades ago. In 2017, state spending on higher education remained well below what it had been prior to the onset of the recession a decade earlier.
As state support for higher education has declined, student tuition has soared and become an increasingly larger share of public university budgets. In 2008, tuition accounted for 35.8% of public higher education funding, but by 2013, tuition represented 47.8% of all higher education revenue in the U.S.
Not surprisingly, soaring tuition is associated with ever-declining proportions of students from lower income families, increasing numbers of foreign and out-of-state students who can and will pay more, rising student debt which in the aggregate now exceeds $1 trillion and burdens graduates with loan payments for years, and declining completion rates as more students must work while they are in school.
The story about other forms of training is much the same. Individual companies are reluctant to invest in training, particularly more general skills, because when people leave, their investment in training that individual goes with them. Training has been among the first things to get cut at most companies as soon as they face any economic stringency. One report noted that since 1985, the amount of funds for U.S. Labor Department training programs declined 20% in real terms, and that the U.S. ranks near the bottom in terms of labor market training as share of GDP.
While there is widespread recognition that reskilling workers for the new economy is essential, much of the needed investment relies on the voluntary, completely discretionary action of private employers, supplemented by government encouragement and some level of government spending. It is hard to see this approach reliably ensuring a successful transition in the face of the rapid automation of work and the enormous amount of job dislocation that is already occurring and will occur in the future.
Government budgets around the world are already under strain from aging populations and large accumulated deficits. It is hard to envision dramatic expansions in the social safety net given these realities.
AI In The Modern Workplace
As the Cornell economic historian Louis Hyman wrote recently, “the history of labor shows that technology does not usually drive social change. On the contrary, social change is typically driven by decisions we make about how to organize our world.” Strategic choice, driven by values as well as social pressures, matter. This statement is as true for artificial intelligence as it was for the creation of factory technology.
AI could be used to make workplaces and work better for people. To take one example, AI programs could monitor prescription drug use, people’s social media postings, and interactions with the health system to quickly identify work environments that are causing harm. AI could also help companies evaluate the effects of changes to improve employee well-being. Simply put, AI can be employed in the relentless pursuit of economic efficiency, even at the expense of people, or, alternatively, to help create more humane workplaces and societies.
In the end, whether companies create healthy workplaces where people thrive or toxic work environments that harm people’s physical and mental health depends on whether companies get serious about building a culture of health.
Healthy workplaces are ones that prevent stress in the first place, by having people work reasonable hours, in a supportive environment, where they have control over their work, sufficient time off to relax and can work free from bullying and harassment. Some companies are committed to these objectives, but not nearly enough of them.
For much of human history, progress along multiple dimensions was readily discernible. Literacy increased and more people stayed in school longer and presumably acquired more skills and knowledge. Medical science extended lifespans dramatically, first through progress against many infectious diseases and more recently by advancements in treating heart disease and cancer.
Productivity gains and global economic integration vastly reduced poverty and dramatically increased the comfort and standard of living for many people throughout the world. And technological advances have profoundly affected people’s ability to travel, communicate, feed themselves and numerous other aspects of everyday life.
But there are no guarantees that progress will continue easily and on the same trajectory into the future. There are numerous indications that work environments are getting worse along multiple dimensions ranging from work hours to work-family conflict to aspects of economic insecurity.
Work environments profoundly matter for people’s well-being. People spend a considerable proportion of their time at work. Social identities derive, in part, from people’s occupations and the status of their employer, assuming they have one. People’s lifestyles, economic security and their ability to provide for their families depend on their earnings, often derived from having a job. People often have and find friends at work, meet their romantic partners in the workplace and enjoy—or not—social support in and from their work environments. Workplaces matter for people’s physical and mental health and well-being.
Consequently, the leaders who oversee—who construct—workplaces and work arrangements have a profound responsibility to their people, whether or not they realize it and act accordingly. Robert Chapman, CEO of the manufacturing company Barry-Wehmiller, has said that according to the Mayo Clinic, a person’s supervisor is more important for their health than the family doctor.