The following is a summary of Artificial Intelligence and its Implications for Income Distribution and Unemployment by Antonin Korinek and Joseph E. Stiglitz. This paper was presented at the Technology Policy Institute Conference on The Economics and Policy Implications of Artificial Intelligence, February 22, 2018.
In Artificial Intelligence and its Implications for Income Distribution and Unemployment, Anton Korinek and Joseph E. Stiglitz focus on a key economic challenge associated with increased use of AI: its effects on income distribution in the context of the future of work. They discuss four cases in which AI innovations are more likely to be substitutes rather than complements for human labor under varying market conditions, and the potential for Pareto improvement[1] in the economic outcomes of both workers and innovators.
The authors find that in a “first-best world,” in which all individuals have access to a perfect insurance market, those who benefit from the increased use of AI could and would compensate those who lose out, resulting in a Pareto improvement. Innovators whose AI applications increase manufacturing efficiency, for example, would profit, while displaced workers would be compensated by the innovators. The authors, being far from political neophytes, note that this scenario is unrealistic.
A Pareto improvement is also possible in a second-best world in which income redistribution is costless and viable. In that reality, the authors argue, AI would increase production capacity, which would shift the Pareto frontier outward and result in a Pareto improvement. That said, the authors acknowledge that though all parties could benefit, some may experience neither loss nor gain from a shift to a new Pareto frontier. In other words, not everyone would be better off, but nobody would be worse off.
The third and fourth scenarios are due to differences in market structures. In a scenario with a perfect market but costly income redistribution, a Pareto improvement is unlikely since the costs of implementing any redistribution scheme exceed the benefits brought by the innovations. In a scenario with imperfect markets, AI might have effects that do not increase efficiency, which would benefit neither innovators nor workers and would potentially shift the Pareto frontier inward, resulting in losses for one or both groups.
These outcomes depend heavily on the specific context surrounding an innovation. Differences between these four realities are likely to be subtle and may change depending on the nature of redistribution, institutional flexibility, and the market structure at any given time. They will also produce different results depending who comprises the groups of innovators and workers.
The authors argue that the surplus generated by the AI could be redistributed in a way that minimizes any harms caused by a resulting decrease in the demand for labor. They also argue that policies to counter changes in wages associated with increased use of AI – such as wage subsidies or earned income tax credits – could allow workers to retain some economic strength and bargaining power.
Taken together, these arguments reflect the authors’ belief that the increasingly rapid pace of AI innovation requires careful consideration of the economic impact of AI applications. They note only two market and societal contexts in which AI innovation might result in Pareto improvement, though only one is realistically possible. Though they call for policies that would support a move toward Pareto improvement – specifically to ensure redistributing income or subsidizing workers’ wages – the authors close by acknowledging the long and complex road ahead. If innovation continues at its current pace and the role of AI in the global economy and society evolves, it may replace or crowd out more and more “work.” Considering now policies that can mitigate the economic impact of this evolution may prepare workers, innovators, and policymakers alike for an uncertain future.
[1] A Pareto improvement scenario in this context is one in which no group, innovator or workers being replaced by AI are made worse off by its implementation.