Recent AI systems have shown impressive capabilities of generating realistic social interactions. However, existing benchmarks on AI social intelligence are either 1) non-interactive, or 2) restricted to one-to-one conversations, falling short of evaluating the systems’ behaviors when deployed to human populations in the wild. In this project, we aim to bridge this gap by building a simulation of human-AI communities, based on an offline social platform, to benchmark AI systems’ prosocial behavior. The benchmark would challenge the AIs’ ability to understand, adapt, and promote socially desirable causes at a community scale, and would provide valuable guidance to future research on socially beneficial AI in business, politics, and other large-scale social settings.

I am participating in the SuperUROP to gain thorough research experience. I have done several UROPs in the past, mostly joining existing projects. The SuperUROP allows me to independently explore a research question from start to finish, with the guidance of expert mentors and peers. I am excited to gain deeper insight into conducting research, build on existing works in the field, and hopefully push the boundaries a bit further!