Instagram's hashtag system blocked related hashtags for Trump-associated content while allowing negative and misleading hashtags to appear for Biden-related searches, creating unequal treatment of the two presidential candidates during the 2020 election campaign.
Instagram's hashtag recommendation system exhibited differential treatment of content related to the 2020 U.S. presidential candidates Donald Trump and Joe Biden. The Tech Transparency Project discovered that Instagram blocked related hashtags for all 10 popular Trump-associated hashtags examined, including #donaldtrump, #trump, and #trump2020, preventing users from seeing potentially negative content. Meanwhile, all 10 Biden-related hashtags displayed related hashtags that often included negative content such as #creepyjoebiden, #joebidenpedophile, and #joebidenisaracist. The issue was first noticed in June 2020 and persisted until at least July 16, 2020. Instagram acknowledged this as a 'technical error' or 'bug' after BuzzFeed contacted them for comment. The differential treatment affected hashtags with millions of posts - #Trump had 14.6 million posts, #Trump2020 had 2.5 million posts, and #MAGA had 5.6 million posts, while #JoeBiden had 390,000 posts. Instagram disabled the related hashtags feature entirely while investigating the issue. The platform has an estimated 112 million users in the United States, with 67% of adults aged 18-29 using the service.
Domain classification, causal taxonomy, severity scores, and national security assessments were LLM-classified and may contain errors.
Unequal treatment of individuals or groups by AI, often based on race, gender, or other sensitive characteristics, resulting in unfair outcomes and unfair representation of those groups.
AI system
Due to a decision or action made by an AI system
Unintentional
Due to an unexpected outcome from pursuing a goal
Post-deployment
Occurring after the AI model has been trained and deployed