AI Applied to Tax Systems Can Help Discover Shelters, Support Equality 

By AI Trends Staff   

AI and machine learning represent the new frontier in tax administration. 

Since researchers at MIT created an algorithm that could flag a certain type of tax shelter in 2015, the concept of using AI in an effort to find those helping to shield their income from taxation has caught on.  

Denmark, which lost nearly US$325 million to tax evasion in 2018, implemented AI tools which have successfully identified 85 of every 100 cases of tax evasion, according to a recent account in The Science Times. France passed a law as part of the country’s 2020 budget that allows tax authorities to deploy algorithms to trawl through social media to detect signs of tax evasion, smuggling, and undeclared income.  

“The complexity of tax rules make it a challenge for any organization to stay compliant, much less reduce their tax liabilities. Therefore, artificial intelligence is well suited for tasks that require a deep analysis of the tax codes,” stated Luís Aires, independent VAT consultant and tax advisor, based in Lisbon, Portugal, writing recently in VATupdate. “Using years of previous tax documentation as a foundation for learning, the AI application can provide an in-depth understanding of the tax codes and stays on top of yearly changes. As a result, it’s easier for tax practitioners to identify key areas for possible savings.”  

Special methods of intelligent data analysis are needed to detect and prevent losses. Detection logic must recognize complex patterns over periods spanning second to months. The logic must also be easily customizable and ability to be maintained by specialists in a changing business environment. Ensuring compliance and finding fraud requires monitoring millions of daily transactions in real time. Proof of non-compliance that can stand up to audits is critical to tax enforcement. 

AI and Machine Learning Can Help Detect Money Laundering 

AI and machine learning can also be applied to uncover or detect money laundering. “Tax authorities use AI to predict risk for tax evasion, or to monitor and identify suspicious tenders or bids in public procurement,” Aries states. Some applications of AI and automated decision systems in society remain controversial, he notes. Questions persist on how to handle biased algorithms, on the ability to contest automated decisions, and accountability when machines make the decisions. Also, the right to privacy, the right to explanation, and the “right to be forgotten” remain topics of debate. “Nevertheless, due to the efficiency, apparent neutrality, stable performance, and cost savings associated with AI based processes, such tools are likely to be applied in more and more areas in the future,” he states. 

Government of India Using AI to Fight Tax Evasion 

The government of India has embarked on an effort to use a machine learning tool to fight tax evasion and identify bogus firms. The AI tool has been researched and developed by two US-based Indian researchers, according to an account in Financial Express. Dr. Aprajit Mahajan, Associate Professor, University of California, Berkeley, and postdoctoral scholar Dr. Shekhar Mittal, will scrutinize a vast dataset of Value Added Tax returns registered in Delhi between 2012 and 2017. 

The study, commissioned by the Delhi government, concluded that similar means were used by traders to evade the Goods and Services Tax (GST). “Future versions of machine learning will build on the GST data,” stated Jasmine Shah, vice-chairman of the Delhi Dialogue and Development Commission.  

The researchers said that this work is the first-ever systematic study on tax evasion in a country where there is weak tax compliance. “Our results indicate that by using our tool, the tax administration can prevent fraud up to $15-45 million,” the researchers wrote in a paper. “Anecdotal evidence suggests that such false paper trails are a common problem. Our work should have high policy relevance both within India and elsewhere,” the researchers stated.  

Salesforce Researchers Studying Whether AI Can Make Tax Policy More Fair 

Whether AI can be used to create a fair and equitable tax policy is the focus of research at the Salesforce Research team, which recently released a simulation tool called the AI Economist.  

According to an account in BrinkNews, the AI Economist uses a two-level reinforcement learning framework and is highly flexible, designed to optimize for equality, productivity or sustainability, which can be set by the user.  

The researchers compared the AI Economist with three other baseline tax methods: the free market with no taxation or redistribution; a progressive tax mirroring the 2018 United States federal tax schedule (i.e., marginal tax rates increase with income); and an analytical tax model proposed by economist Emmanuel Saez, which results in a regressive tax schedule in this case.  

In simulations, the AI Economist achieved a 16% gain in the tradeoff between equality and productivity compared to the next best framework, the Saez model. Compared to the free market, the AI Economist improves equality by 47%, with an 11% decrease in productivity, the researchers said.  

“We believe these initial results demonstrate the potential of applying a data and simulation-driven approach to quickly create equitable and effective economic policies,” stated Stephan Zheng, lead research scientist and senior manager at Salesforce.   

Reinforcement learning algorithms use smart trial-and-error strategies to optimize policy models for a specified goal. During this process, the learning algorithm continuously uses feedback it receives to improve the policy models. High-profile applications of RL enabled AI to compete and win against human players in popular games including Go, Dota 2 and Starcraft.   

Taxes were chosen as a focus for the model, since they are a near-universal part of society, used by local, state, and national governments. “But no one has truly determined now tax policy can be feasibly optimized in complex, dynamic economies,” Zheng stated. The number of contingencies to consider, he said is “near-infinite.” He credited Prof. David Parkes, head of the Economics and Computer Science Group at Harvard University, with assisting in the research.  

Read the source articles in  The Science Times, VATupdate, the Financial Express and in BrinkNews.