BotBuster Universe

Social media bots are automated fake users. These bots can be malicious, spreading mis/disinformation which results in polarization, hate and violence. They are also used by state-sponsored actors to push forth desired narratives. This research pillar studies the detection and analysis of these bots on social media.

Github Repo

Check out the GitHub Repo for a simple GUI for bot detection here!

(1) Finding Bots

• Stabalizing Bot Detectors:Bot detectors typically return the probability value of a user being a bot. What is a suitable threshold value to term the user as truly a bot?
• BotBuster:Mixture-of-Experts neural network architecture for bot detection on X (previously Twitter)
• BotBuster for Everyone: Bot detection for X Reddit and Instagram. This work uses random forests to construct a mixture-of-experts bot detector, building on Occazm's Razor that a simple bot detection algorithm will suffice.
• BotBuster-TelegramBot detection for Telegram. This work investigates the Disinformation Dozen who are known to spread disinformation about the coronavirus. We find that bots initiate disinformation ideas but humans spread them.
• CyborgsBot detection for Cyborgs, a type of half-human, half-bot. They are used a lot in strategic communications by activists and important people.

(2) Types of Bots

• General, News and Bridging Botsin the context of Digital Diplomacy in US-China relations. General bots are the common bot. News bots spread fresh news from journalists or aggregate news. Bridging bots serve to connect two groups of users.
• Repeater, Amplifier Botsin the context of Taiwan-China relations, Repeater bots constantly repeat the same passage incessantly. Amplifier bots share posts and act as loudspeakers.

(3) Bots v Humans

• Active, aggressive, but to no avail: Bots try to sow discord during the 2020 Singaporean elections but were not very successful
• Bot-based Emotions: Bots use more evocative emotions than humans during the Kashmir black day event
• Stance Flipping: Pro or Anti? A Social Influence Model of Online Stance Flipping. How do we know who would change their attitudes towards the coronavirus vaccine? How are they influenced by their past thoughts and social network?
• Bots and Influence Maneuvers: Covert coordination by bots through 4-chracter hashtags. And patterns of influence maneuvers.
• Cross-Straits Relations: Bots in China-Taiwan relations discourse

(4) Coordinated Bots

• CSI, Combined Synchronization Index: An index to evaluate coordination on social media. Humans have higher participation of synchronization with original posts than bots, but bot-human pairs have the highest synchronization index. Bots do have an effect on human users, and can manipulate their opinions.
• Online Coordination: Methods and Comparative Case Studies of Coordinated Groups across Four Events in the United States. How do bots and humans coordinate differently?
• Image Coordination Web: Image coordination patterns and strategies of China, Russia, Iran and Venezuela based on Twitter's Inauthentic User Takedown data.
• Synchronization in URL and Text: 2020 and 2023 French Protests.