Reinforcement learning in the field of tax law

J. Molenaar & Y. Kaya

June 24, 2019

Artificial intelligence has entered our lives and fundamentally changed them. Moreover, research in AI still shows continuous improvements and more capabilities every day. A recent example of AI's accomplishments was Google’s DeepMind, which was able to learn by itself how to walk through the use of Reinforcement Learning. Until now, the current literature has primarily focused on making sure the AI-agents will be able to technically become a part of the community and making sure no humans are harmed along the way. AI-agents however, would have to deal with other 'laws' as well, such as contract law and tax law. Therefore, In this paper, we'll look at the current state of technology to explore whether an AI-agent can learn how to obey tax legislation based on the reinforcement learning method. In order to draw conclusions, we'll discuss in great detail the characteristics and conditions of reinforcement learning, use cases thus far and present a model for a TAX-AI agent based on Reinforcement Learning.

Blockchain: gambling tax - automated and transparent payment to the tax authority with the use of smart contracts

M. van Schaik & R. Michgels

June 24, 2019

In this paper, the authors have taken a look into the possibilities of automated and transparant payments of gambling tax to the tax authorities with the use of smart contracts. This paper was written for the completion of the Tax & Technology 2 course at the University of Tilburg in June 2019. The main purpose of this study was to see the possibilities of taxation within the developing gambling markets, where more often gambling platforms use blockchains, for example the Ethereum blockchain. First the current situation of gambling taxation and online gambling is explained, after that the use of blockchain for gambling tax payments is researched. The main result is that is it possible to make payments with the use of a smart contract, but some law changes are necessary before this system can work. There are some challenges which should be considered as well, for example the current obligation to file tax returns and usage of the taxpayer identification number as this number should not be stored on the blockchain.

Binary translation of legislation and case law - A Deep Learning dive into the possibilities of Natural Language Processing for legislation and case law

S. de Jong & S. Derksen

June 25, 2019

The purpose of the authors is to discuss the application of machine learning in the field of tax law, and specifically the possibilities of applying Natural Language Processing (NLP) techniques such as text and argumentation mining on the existing data from legislation and case law. The law and rulings of judges are currently written in human language. The authors investigate to what extent it is possible and desirable to come up with a version that can be understood by a computer. This paper will provide a literature review on the possibilities and the limitations. The combination of the tax and technology approach makes that this paper is an interdisciplinary contribution to existing academic literature. The authors recognize the usefulness of machine learning applications in law, but do not expect these to be operational on short notice. However, this field of study is rapidly expanding and creating various applications for different problems. Based on literature review, the authors believe that decision tree analysis on legislation combined with the text mining on case law will create possibilities to make written human language quite understandable by computers.