In real life, taxation and public spending are two important policies related to socio-economic development and the gap between rich and poor. A reasonable government often needs to dynamically adjust these two policies according to the realities of the people. In reality, however, households are often assumed to be independent and that each household pursues his or her own interests. People's day-to-day activities can be boiled down to production, consumption, investment, and exchange. Each household is paid for its work (production), which is used to buy goods (exchange) to meet the needs of life (consumption), and may deposit surplus funds in the bank or for financial management (investment). For each family, how to distribute income and how to balance work and rest time to improve life happiness, has been a long-term topic. Some families are thrifty and save for a long time for a rainy day. Some families prefer instant pleasures and often can't make ends meet. Some people choose to work long hours and work hard but neglect their health; Some people prefer to work from 9 to 5 and spend a lot of time enjoying life. The aggregate economic activities of these households constitute a macro economic society, and their evaluation will be reflected in indicators such as economic growth, wealth distribution and social welfare. For the government, how can it promote steady and sustained economic growth?How to ensure social fairness and stability and narrow the gap between the rich and the poor?How can every family live a happy life and enjoy social benefits? These have been long-standing goals and responsibilities of the government.
In order to explore the complex game problem described above, we construct a TaxAl simulator to explore the conflicts and connections between the decisions of different actors in an ideal environment, and how an agent can make rational decisions in complex situations. TaxAl simulator has four roles: government, households, companies and financial institutions (as shown in the figure above). In practice, firms and financial institutions are determined by predefined hyperparameters and fixed rules as part of the environmental mechanism, while government and households change dynamically due to different decisions. The direct relationship between each role can be summarized as:
For a detailed description of the relationship, see GitHub Link: github link 。
The challenge consists of two tracks, focusing on decision making issues in the economy and society from different roles.
Track 1 adopts a optimal tax-government environment. In this environment, players will take on the role of a government, challenging how to promote domestic production and balance the distribution of wealth in a complex economic society. household characters will be randomly drawn from existing strategies to add randomness to the game. The government will have macro information about the economy and society, such as overall wage levels, average assets, average incomes and productivity of different wealth groups, and will issue policies to regulate tax rates and government spending based on this information in order to maximize gross domestic product (GDP) and balance the distribution of wealth (Gini Index).
Track 2 uses a optimal tax-households environment. In this environment, households, as part of the economy, need to adjust the proportion of their working hours and wealth investment according to their own understanding of the social information and economic situation to maximize their own personal interests. There are 4 household decision-makers in this environment, each independent of each other. The optimal benefit of each person is a function of personal consumption and working hours, and the higher the consumption, the shorter the working hours, and the higher the personal benefits. Therefore, an optimal household agent needs to achieve a balance between work and consumption through decision-making based on the observed information, combined with its own assets and productivity.
The competition will be held online on the Jidi multi-agent open-source platform (Jidi Platform). The Jidi Platform provides participating competitors with fair agent evaluation and real-time ranking functions online. Users only need to register an account on the Jidi Platform and sign up to participate in the challenge.
Each participating team consists of 1-3 members, including a team leader and two team members. The steps to participate are:
Institute of Automation, Chinese Academy of Sciences
Nanjing Artificial Intelligence Research of IA
Email: haifeng.zhang@ia.ac.cn
Institute of Automation, Chinese Academy of Sciences
Email: yan.song@ia.ac.cn
Institute of Automation, Chinese Academy of Sciences
Email: miqirui2021@ia.ac.cn