Rethinking Online Content Ecosystems through the Lens of Computational Economics

Location: BBB 3901
Zoom: https://umich.zoom.us/j/99952998439

Abstract

Human-created online contents not only are important to our life but also supplied tremendous data that enable the recent success of generative AIs (GenAIs). In turn, GenAI is also revolutionizing the way contents are now created. Such human-AI mutual dependence is fundamentally transforming existing content ecosystems. If not addressed properly, it may distort human creators’ incentives and even drive humans out of the ecosystem. In this talk, I will present our recent works on understanding content creation competitions through the lens of multi-agent systems and computational economics, and how to design creator reward mechanisms to incentivize desirable distribution of created contents. Our designed mechanism not only enjoys nice theoretical properties, but also achieved promising efficiency during live tests with ~10 millions of users on Instagram Reels.

Bio

Haifeng Xu is an assistant professor in computer science and data science at the University of Chicago, and directs the Strategic Intelligence for Machine Agents (SIGMA) research lab which focuses on AI agents and data-driven decision making, particularly in non-cooperative multi-agent setups. His works have been recognized by a few major awards, including an AI2050 Early Career Fellow, IJCAI Early Career Spotlight, a WWW best paper and an AAMAS best student paper.