Modern financial markets represent fertile soil for AI systems. As of October 2019, at least two thirds of the UK financial services companies use AI, with its growing adoption in trading, risk management and pricing. Should the regulator trust the AI technology that has been independently developed by competing firms? Will the financial system suffer from AI that is not properly designed and not thoroughly tested? Will the money of retail investors be safe? How would AI bots react to different regulations put in place by regulators? What is the best way of regulating financial markets populated by AI?
Assessing the trustworthiness of AI bots and how to safely regulate them in financial markets are challenging research questions, as there is no experimental platform and counterfactuals are not easily feasible.
This project proposes to build incentive-aware digital twins for finance, that is, a realistic simulation platform that will allow to experiment with financial markets “in the lab”. We envision a methodology built on two main pillars of symbolic AI. Firstly, agent-based models can be adopted to calculate the empirical payoff matrix (e.g., returns) for each combination of strategies (e.g., AI trading algorithms) adopted by the agents (e.g., traders). The output of this step will be an empirical game. Secondly, equilibria of this game will be computed and analysed to understand where the incentives of AI-powered market participants lead the financial system.
The platform will be useful to regulators to understand how to define market rules in presence of profit-maximising AI algorithms and evaluate the systemic trustworthiness of AI. The research will focus on modelling as well as computational challenges to guarantee the scalability and adaptability of incentive-aware digital twins for finance.