Groups of people often encounter difficulties when working together. In many situations, people prioritize their self-interest, which can undermine efforts to benefit the group as a whole, as seen in problems like traffic congestion, public-health failures and climate change. Even when individuals want to act for others, they may struggle to find the best course of action when confronted with social dilemmas where the interests of the individual and the group are not aligned.
The Addressing Collective Action Problems With Machine Intelligence project aims to explore how machine intelligence can help address these challenges in human collective action. While people currently use AI mainly for individual convenience and self-interest, collective-action theory suggests that machine intelligence optimized for public goods may require a different design concept than those intended for individual goods. The project seeks to identify the differences and clarify how social interactions and technology can work together to foster the emergence of social order. It focuses on how this process relates to collective action and the resolution of social dilemmas.
The Addressing Collective Action Problems project will develop and conduct controlled experiments in which many human participants interact with each other with and without the support of machine intelligence in specific social contexts, such as online messaging and driving coordination. The experimental findings will be verified through empirical studies on real-world AI applications. Ultimately, the project aims to demonstrate how machine intelligence can be used as a tool for social interventions that translate knowledge of computer science and AI into helpful practices for public goods.
The project is being led by Hirokazu Shirado at Carnegie Mellon University (Pittsburgh, US).