Artificial Intelligence: Growth, Inequalities and Public Policies

Artificial intelligence (AI) is rapidly transforming economies, but its effects are still uncertain, and the subject of a vivid debate. Will AI drive innovation, productivity, and economic growth? Or will it lead to job losses, rising income inequality, and more concentrated markets? Public policy will play a key role in shaping AI’s impacts. Governments face however a major challenge: how can policies maximize the benefits of AI while minimizing its societal risks? Should they focus on boosting economic growth or reducing inequalities and risks? Striking the right balance is crucial, but we still lack clear answers on how to do this effectively. This project tackles that question by introducing an innovative approach that argues that instead of making policy decisions based on theory alone, new policies should be designed by using real-world data to learn from the effects of past policies. The project will analyze how previous AI-related policies have impacted businesses and society, helping to design smarter, evidence-based policies that would contribute to maximize social welfare in the future.

There is currently a vivid societal debate and nascent research about the economic effects of artificial intelligence (AI). Some expect to see positive impacts in terms of innovation, productivity and economic growth. Others point to possible economic harms such as technological unemployment, increasing market concentration and exacerbation of income inequalities. Public policies will have a crucial role to shape the direction and extent of these effects. However, it is still quite unclear what policies could do to foster the future economic benefits of AI, what they could do to avoid its possible harms, and how they could balance the positive and negative impacts. The present project will investigate the trade-off between growth and inequality effects of AI, and analyze how public policies for AI should be designed in order to take this trade-off into account. The key novel aspect of the project will be the application of the optimal policy learning (OPL) approach to the study of policies for AI. The main idea of this approach is that optimal policies should be designed by exploiting information about the effects of previous policies on the treated agents, in such a way that new policies will seek to maximize ex-ante the empirical welfare of the target population. In other words, the design of new policies is optimal because it is based upon, and it learns from, the effects of previous policies and the heterogenous characteristics of agents. To apply this approach to the study AI policies, the project will develop the following pillars: (1) a conceptual framework of firm heterogeneity and AI, which points out the main factors that affect firms’ adoption of AI and its effects; (2) empirical analysis of how AI and related policies affect different types of firms; (3) analysis of how different policies should be designed in order to maximize empirical welfare, and how these policies can address the trade-off between efficiency and equity effects of AI.