At Rutgers University, two REU students worked on topics related to ACTION in Summer 2025. William Guo (U Penn) and Edward Xiong (MIT) joined Professor Jie Gao on a project that models learning in a social network. The project will impact downstream ACTION research, for instance in groups of agents tasked with network defense.
Truth Learning in a Social Network
Abstract: In the sequential learning problem, a network of agents make decisions, informed by a noisy private signal and the predictions of neighboring agents before them. We explore the properties of networks where agents make decisions in a uniformly random order. We characterize necessary conditions for such networks to achieve asymptotic truth learning and introduce various graph constructions that learn in different ways. We also develop an algorithm to transform arbitrary graphs into random-order learning networks using few edge/vertex modifications, with provable approximation guarantees. Finally, we analyze the robustness of learning networks, demonstrating that those achieving asymptotic truth learning under random orderings are resilient to a bounded number of adversarial modifications. Our findings reveal structural properties in networks that achieve random learning and offer algorithmic tools for engineering strong social networks.
In Summer 2026, the ACTION Institute will offer summer internship at additional campuses. Please reach out if you are interested in participating. Find more information on our Education and Outreach page.