Salesforce's AI Economist aims to help leaders design tax policies

Ultimately, Salesforce's AI Research team says, the objective is to help leaders create policies that can lead to specific social objectives like helping the middle class.
Written by Stephanie Condon, Senior Writer

Salesforce Research on Wednesday unveiled the AI Economist, a new line of research leveraging reinforcement learning to help policymakers predict the outcomes of various tax policies. The ultimate objective, Salesforce's AI researchers say, is to help leaders create policies that achieve specific social objectives like helping the middle class. 

"Our model gives economists and policy-makers additional tools on which they can base their decisions," Lead Research Scientist Nikhil Naik said in a Salesforce blog post. "By simulating millions of years of economies and finding a variety of tax frameworks, the AI Economist can predict how people would actually respond to a tax, like whether it will incentivize them to work more or work less." 

The new AI framework uses a collection of AI agents to simulate how real people might react to different taxes. Each AI agent earns money by collecting and trading resources and building houses. They learn to maximize their utility (or happiness) by adjusting their movement, trading, and building behavior. While the simulation is running, the AI Economist learns to optimize taxes and subsidies to promote specific objectives.

The research is based on the premise that the right tax policy can strike an "optimal" balance between productivity and equality. While there are certainly policies that can increase both productivity and equality, a market economy does inherently create trade-offs between the two.

Using economic models to predict the outcomes of various tax policies can be difficult, the researchers noted in a blog post, because they fail to account for the dynamic nature of the economy -- policies are always changing, and workers gain new skills.  

"Economic theory cannot fully model the complexities of the real world," they wrote. "Instead, tax theory relies on simplifying assumptions that are hard to validate, for example, about the effect of taxes on how much people work. Moreover, real-world experimentation with taxes is almost impossible."

Using AI can overcome some of those challenges, they posit. At the same time, the team acknowledged that "AI-based economic simulations still have limitations." 

"They do not yet model human-behavioral factors and interactions between people, including social considerations, and they consider a relatively small economy," the blog post said. "However, these kinds of simulations provide a transparent and objective view on the economic consequences of different tax policies. Moreover, this simulation and data-driven approach can be used together with any social objective in order to automatically find a tax policy with strong performance."

Editorial standards