How AI improves fiscal administration and invoicing
The use of AI in fiscal administration and invoicing is giving some encouraging results. Finally, a system has been created which is able to simulate thousands of businesses in parallel and generate optimised solutions, while minimising the social impact.
These solutions can also be used in everyday life, or to assist governments, institutions and economists in planning more equitable fiscal policies, which take into account the needs of everyone. The final objective of this experiment is to achieve financial parity.
Reinforcement Learning: how AI assists economists
Although attempts to try and balance taxes and welfare can be particularly complex, thanks to Reinforcement Learning (the working principle on which AI for tax administration and invoicing is based), experts have the chance to test solutions within a short period of time and find the right balance between the various financial instruments.
Artificial intelligence simulates real situations, beginning with actual data from past economic scenarios. This is used to identify optimal taxation rates to permit a specific socio-economic objective to be reached.
Simulation using AI is opening new frontiers to economists all over the world. It is now possible to assess the likely repercussions of changes and (even minimal) adjustments to various economic policies, without causing a global crisis.
Reinforcement Learning is a powerful instrument, destined to facilitate the task of governments to provide citizens with a compromise between financial parity and productivity. Via this system, inequalities can be gradually reduced and wealth redistributed more equitably.
How simulation works in detail
The program, designed to provide feasible answers as to how to obtain the perfect economic balance, is an authentic simulation of human behaviour. Inside, it contains a series of artificial intelligence agents, which emulate the hypothetical choices real people would make.
Each of these agents earns money from the exchange of resources or the construction of houses, with the final objective being to maximise their own income (or contentment). In order to reach that result, the agents continue to vary their behaviour independently during the sales and construction phases. By doing this, they ‘learn’ to look for the best solution to obtain the greatest benefits.
These choices are analysed and applied on a larger scale, enabling AI for tax administration and invoicing to optimise taxation and welfare, with the aim of promoting global objectives. The complex models generated are able to predict people’s reactions to the introduction of a new tax and allow experts to take into account numerous additional variables (such as the environmental aspect).
Data from experiments with AI
According to the first experiments using AI for fiscal administration and invoicing, it was observed that AI can bring an improvement of 16% in the balance between economic equity and productivity. This is an encouraging result, indicating that this system is more advantageous than the Saez taxation formula (a widely used system), the US federal tax rate system or the free market.
Artificial intelligence has also proposed personalised income-based taxation schemes. The most well-received of these incorporates higher tax rates for medium to high income earners so as to provide greater welfare support for those on lower incomes.
The behaviour of some artificial intelligence agents has also proved interesting, as they have tried to optimise their own income in a climate of economic instability. Such actions also reflect real life situations and often lead to marked economic disparity. The opportunity to predict such reactions gives institutions a way of intervening through the use of preventative measures, developed on an ad hoc basis to contain these behaviours.
Translated by Joanne Beckwith
