5 investments to close the gap between AI wealth and welfare

Amid the transformative promise of artificial intelligence, one significant question is “Will AI enhance welfare for all or generate wealth for just a few?”

Studying the rollout of other general-purpose technologies, such as the automobile and the internet, is helpful in the quest for an answer. 

In a working paper, MIT Sloan School of Management postdoctoral associate Isabella Loaiza and professor Roberto Rigobon outline the importance of investing in complementary infrastructure to ensure that everyone can reap the benefits of general-purpose technologies — that is, innovations with the power to transform economies, industries, and societies. 

Cars became a cornerstone of modern life once society invested in a series of system-level complements: roads, traffics laws, safety standards, licensure, insurance, and so on. The internet, meanwhile, has seen insufficient investment in governance, public data infrastructure, and access policies, the authors write. As a result, its gains haven’t been distributed equitably.  

“We’re at a point where this wave of AI is nascent, and we can still decide how to deploy it in society so that it brings welfare to many,” Loaiza said. “I would not like to see AI increase concentration of resources and higher inequality. I’d like to think that AI can close that gap, but that will require these system-level complements.”

Systems need to support technology

General-purpose technologies are innovations with three characteristics: They are widely used, each iteration improves upon previous versions, and the technology paves the way for additional innovation. 

AI can be defined as a general-purpose technology — and it will likely spur change far more quickly than previous technological developments, MIT Sloan principal research scientist Andrew McAfee has said.

To transform the wealth created by innovation into welfare, societies must make complementary investments, the authors argue in their paper, “From Wealth to Welfare: The Social and Economic Institutions to Complement AI.”

The key, though, is that these complements should be systemic rather than ad hoc, Loaiza said. 

“System-level complements are parts that work together to make something greater than the whole,” she said. “They need to be intertwined, not independent, and they need to have an established relationship.” 

5 complements to ensure AI equity

The paper defines five categories of system-level complements that AI requires in order to translate gains from technological progress into genuine improvements in workers’ welfare.

• Capital. General-purpose technology needs infrastructure that makes it accessible and usable. A mix of public and private investment is necessary, as in the case of public roads but privately owned gas stations. Who will fund AI infrastructure, and to what extent, remains to be seen.
• Human development. Technology scales when new developments make it safe and effective. For AI, this will take the form of emphasizing human capabilities that complement AI and transitioning work from a linear path (get a degree, get a job, and retire) to a cyclical model (build new skills, take on a different role, and feel a renewed sense of purpose). Support such as reskilling programs and unemployment insurance will be necessary to help workers navigate such transitions.
• Social and economic institutions. Governance ensures that technology’s gains can be shared. For example, the five-day workweek helped curtail worker exploitation. AI needs a combination of new labor standards and tax reforms. That would reduce companies’ financial incentives to prioritize machines that don’t need to be paid or offered benefits, and help societies “deter the erosion of [human] worker autonomy,” Loaiza and Rigobon write.
• Metrics. Companies and governments alike must measure technology’s impact to guide policy and influence investment. When it comes to AI, key dynamics to monitor include how labor and technology evolve in tandem, how AI’s impact varies across geographic regions (especially urban areas), and where AI contributes to financial or infrastructure risk. 
• Systems thinking. Actions taken locally can have a global impact. In that sense, AI should be viewed as a component in larger social, economic, and technical systems, not just a stand-alone tool. “AI does not act in isolation,” the researchers write, “and neither should our approach to governing it.”

AI Executive Academy

In person at MIT Sloan

Register Now

For fast ROI, invest in people

Building system-level complements is admittedly easier said than done. It often requires a vision, a long-term plan, and the budget to match. “It’s hard to coordinate for very large projects,” Loaiza said.

It’s also hard to conceptualize such efforts amid the short-term thinking of publicly traded firms filing quarterly earnings reports; governments facing two- or four-year election cycles; and citizens growing easily frustrated with, say, proposed data centers that will use copious amounts of water and energy in places where both resources are in short supply.

Loaiza said that each of the five system-level complements is equally important but that educating, training, and reskilling the workforce would likely have the biggest impact relative to the resources required. “We need to figure out a way to train people in a way that complements AI,” Loaiza said. “When we have people who can use the technology, then workers will have jobs to do.” 

In previous research, she and Rigobon defined skills that complement AI’s shortcomings, emphasizing human capabilities that demand empathy, presence, opinion, creativity, and hope. Their current research looks at these capabilities in the context of how university education is evolving and can contribute to equitable gains for students in different fields of study as they begin to enter the labor market.

“Artificial intelligence holds immense potential to reshape our societies, economies, and daily lives, but its trajectory is not predetermined,” the authors write. “Investing in complementary capital, human development, institutions, metrics, and systems thinking is key to aligning AI with human progress.”

Read “From Wealth to Welfare: Systemic Complementarities for the Age of AI”

Isabella Loaiza is a postdoctoral researcher at the MIT Sloan School of Management. As a computational social scientist, she studies ways to build an equitable and sustainable future of work. Her research explores AI’s impact on work, workers, and organizational talent practices, with a human-centered approach that emphasizes human-AI complementarities. Key research interests include automation and augmentation dynamics, skilling and reskilling, labor inequalities, and sustainable skills.

Roberto Rigobon, PhD ’97, is a professor of applied economics at MIT Sloan, a research associate of the National Bureau of Economic Research, a member of the Census Scientific Advisory Committee, and a visiting professor at IESA (Venezuela). He is co-faculty director of the MIT Sloan Sustainability Initiative and a co-founder and director of the Aggregate Confusion Project, which studies how to improve environmental, social, and governance measures.