Addressing real-world socio-economic challenges requires designing and testing economic policies. However, this is difficult in practice, due to the lack of appropriate economic data at the micro level and the limited opportunity to experiment. But Salesforce, a leading CRM company and a world benchmark in artificial intelligence (AI) thanks to its Einstein technology, has developed a solution that makes it possible to create a virtual testing ground available to economists. It is a framework for testing variables through laboratory simulations.
Publicizing this application of AI is the main objective of The AI Economist, a machine learning framework that allows scenario simulations to see the social impact of political and economic decisions. This information, based on artificial intelligence and machine learning, could help governments make decisions with less risk.
The model aims at an optimal balance between equality and productivity, which is one of the main differentiators of the model with respect to other AI and machine learning applications. The AI Economist is not interested in defeating an enemy, but constantly looking for a beneficial scenario for all parties: companies, citizens and the State.
Promising results
The results have been promising so far. For example, in some of its simulations The AI Economist achieves 16% gains compared to the next best model, that of the prestigious economist and UC Berkeley professor Emmanuel Saez. The AI Economist also improves equality by 47% compared to the free market with only an 11% decrease in productivity. One of the most interesting aspects of The AI Economist is that it implements tax lists qualitatively. It imposes higher tax rates for the highest income and lower rates for the middle income.
Human simulations
Additionally, The AI Economist is also effective in simulations with human participants, achieving competitive equality-productivity tradeoffs and a significantly higher income-weighted average social welfare. In this sense, these results suggest that future AI-powered fiscal models could successfully improve social outcomes in the same way as in this virtual real-world scenario.
Salesforce’s vision for AI Economist is to enable an objective study of the impact of economic policies on real-world economies, at a level of complexity that traditional economic research cannot easily address. The intersection of machine learning and economics presents a wide range of interesting and useful research directions and results in new advances that will have a significant positive social impact.
Understand a changing world
A government may prefer income redistribution to improve equality. However, higher taxes can discourage work and can especially affect highly skilled workers. A ‘smart’ fiscal policy optimizes this balance between equality and productivity. Taxes can improve equality by transferring wealth. At the same time, a higher tax burden can also lower productivity because it can demotivate workers. The AI Economist pursues a fiscal policy that optimizes this balance.
AI-based economic simulations, however, still have limitations. They do not yet model human behavioral factors and interactions between people, and consider a relatively small economy. However, these types of simulations provide a transparent and objective view of the economic consequences of different fiscal policies. Furthermore, this data-driven approach can be used in conjunction with any social objective to automatically find a successful tax policy. Future simulations could improve the fidelity of economic agents using real-world data, while large-scale engineering and technological advances could increase the scope of economic simulations.
