Artificial Intelligence Can Devise An Optimal Tax Policy To Reduce Inequality

A critical problem that economists have historically tried to solve is the problem of income inequality. In terms of government, fiscal policy, more specifically taxation is a most influential tool that can be used to address income inequality. It might be assumed that the government can raise the rate of taxation to high income earners to balance income levels around the economy. However, there exists a certain rate of taxation beyond which the incentives for tax avoidance start to become more prominent and discourage labour from working. This obviously to the government receiving less tax revenue, which counteracts the purpose.

The Laffer Curve - An economic model showing the relationship between the rate of taxation and the tax revenue.

Identifying the optimal level of taxation is quite complex. Human behaviour is highly unpredictable and gathering data can be time consuming. Despite decades of economic research being put into finding the optimal tax rate, it remains an open problem. But, scientists at the US business technology company, Salesforce, believe they may have found the key to solving the problem – Artificial Intelligence. The team has developed an AI system called the AI Economist, which uses reinforcement learning technology to identify the optimal level of taxation to make reduce inequality. The tool is still limited in terms of the variables considered while calculating the optimal tax rate, but it  is still believed to be in its early stages and might be a step in the right direction to effectively evaluate economic policies by making the decisions less political and more data-based.

In one of its early results, the AI Economist had generated a custom tax policy that was 16% more fair than those devised by state-of-the-art economists of the modern day. The AI system does so by running a specialised simulation involving four virtual workers, who collect wood and stone in a digitally generated environment. These workers then trade in these wood and stone or use them to build houses, in return for money. The workers are split into two low-skilled labour and two high-skilled labour, enabling specialisation to take place. The low-skilled labour immediately trade in their collected stone and wood for money, whereas the high-skilled workers use the resources to build houses and later sell the house. These workers are then taxed on their income at a rate that is devised by the AI-enabled program, using the reinforcement-learning algorithm. The objective of the AI economist is to deduce an optimal tax rate which maximises both productivity and equality.

Although simply emulating the behaviour of four workers may not be sufficient to create the most effective economic policies, the interactions between a small number of economic agents leads on to a large number of complex consequences. But there is a general consensus that in order for the AI Economist to be completely reliable, is it going to have to be able to model the behaviour of a larger number of workers.

Furthermore, the AI-policymaker is not only useful to device a general tax policy suitable for normal situations, there is also potential to alter the program to take into account extenuating circumstances, such as
pandemics. Adding additional variables to the algorithm such as social distancing and local business closures, can help provide suitable amendments to the fiscal policy to ensure that the normal productivity and equality levels are retained even during  unprecedented times such as the COVID-19 pandemic.