# Trust AI Regulation? Discerning users are vital to build trust and effective AI regulation

**Authors:** Zainab Alalawi, Paolo Bova, Theodor Cimpeanu, Alessandro Di Stefano, Manh Hong Duong, Elias Fernandez Domingos, The Anh Han, Marcus Krellner, Bianca Ogbo, Simon T. Powers, Filippo Zimmaro  
**arXiv:** [2403.09510](https://arxiv.org/abs/2403.09510)  
**Date:** 2024-03-14  
**Summary page:** https://www.summarizepaper.com/en/paper/2403.09510v1/

## Key points

- Regulation is crucial for incentivizing AI creators to develop trustworthy systems and for users to trust these systems
- There is a debate on the form and implementation of AI regulations
- Authors propose using evolutionary game theory as a tool to model dilemmas faced by users, AI creators, and regulators
- Effective regulation requires regulators to be incentivized to regulate efficiently
- Mechanisms for incentivization include governments recognizing and rewarding regulators who excel in their duties, as well as users influencing trust decisions based on regulator effectiveness
- Evaluating diverse regulatory frameworks through an evolutionary game theoretic lens is essential for establishing a regulatory environment conducive to fostering trust in AI systems

## Summary

In their paper titled "Trust AI Regulation? Discerning users are vital to build trust and effective AI," Zainab Alalawi, Paolo Bova, Theodor Cimpeanu, Alessandro Di Stefano, Manh Hong Duong, Elias Fernandez Domingos, The Anh Han, Marcus Krellner, Bianca Ogbo, Simon T. Powers, and Filippo Zimmaro delve into the necessity of regulation in the realm of artificial intelligence (AI). They acknowledge the consensus that regulation is crucial for incentivizing AI creators to develop trustworthy systems and for users to place their trust in these systems. However, there exists a significant debate regarding the form and implementation of such regulations. The authors propose leveraging evolutionary game theory as a quantitative tool to model the dilemmas faced by users, AI creators, and regulators. By doing so, they aim to provide insights into the potential impacts of different regulatory regimes on the development of trustworthy AI and user trust. Their analysis reveals that effective regulation hinges on regulators being incentivized to regulate efficiently. Two mechanisms are highlighted as key drivers of this incentivization process. Firstly,governments can play a role by recognizing and rewarding regulators who excel in their duties. This approach fosters trustworthy development and user trust if the AI system's risk level remains manageable for users. Alternatively,user scan influence trust decisions based on regulator effectiveness.This strategy leads to effective regulationand subsequently promotes the cultivation of trustworthy AIand user trust – provided that regulatory implementation costs remain reasonable. The findings underscore the significance of evaluating diverse regulatory frameworks through an evolutionary game theoretic lens.By considering these dynamics comprehensively,stakeholders can work towards establishing a regulatory environment conducive to fostering trust in AI systems while ensuring their reliability and safety for end-users.