Carnegie Mellon University researchers found that roughly half of 200 million COVID-19 related tweets, including 62% of the most influential retweeters, were generated by AI bots spreading misinformation about reopening America during the pandemic.
Carnegie Mellon University researchers analyzed over 200 million tweets discussing COVID-19 and related issues since January 2020 and discovered that approximately 50% of accounts appeared to be bots, with 62% of the 1,000 most influential retweeters being bots. This represented a significantly higher level of bot activity than the typical 10-20% seen during contentious events like elections or natural disasters. The researchers identified bots using AI systems that analyzed accounts based on tweet frequency, follower counts, and apparent location changes. Among tweets specifically about 'reopening America,' 66% came from accounts that were possibly humans using bot assistants to amplify their messages, while 34% came from actual bots. The researchers found over 100 types of inaccurate COVID-19 stories being spread, with tweets about ending stay-at-home orders being the most frequent. The bot detection was based on indicators like tweeting more frequently than humanly possible, appearing in different countries within hours, using identical hashtags, and copy-pasted messaging. The researchers could not determine whether the bot activity originated from foreign nation-state actors or domestic sources within the US, but noted that the global nature of the pandemic created opportunities for countries and interest groups to advance political agendas through this misinformation campaign.
Domain classification, causal taxonomy, severity scores, and national security assessments were LLM-classified and may contain errors.
Using AI systems to conduct large-scale disinformation campaigns, malicious surveillance, or targeted and sophisticated automated censorship and propaganda, with the aim of manipulating political processes, public opinion, and behavior.
AI system
Due to a decision or action made by an AI system
Intentional
Due to an expected outcome from pursuing a goal
Post-deployment
Occurring after the AI model has been trained and deployed