HawkEye

Twitter's Birdwatch is a new community-driven misinformation detection platform where users provide notes to label tweet accuracy, and rate the 'helpfulness' of other users' notes. This work investigates the robustness of Birdwatch against adversaries injecting fake ratings and shows that the current Birdwatch system is vulnerable to adversarial attacks. To overcome this vulnerability, we develop HawkEye, a cold-start-aware graph-based recursive algorithm and show that HawkEye is more robust against adversarial manipulation and outperforms Birdwatch in identifying accurate and misleading tweets. Code and data are available at https://github.com/srijankr/hawkeye.