Twitter's algorithmic feed was found to systematically amplify tweets from right-wing politicians and news outlets more than left-wing content across six out of seven countries studied.
Twitter conducted a comprehensive study examining tweets from elected officials in seven countries (UK, US, Canada, France, Germany, Spain and Japan) between April 1 and August 15, 2020, comparing their algorithmic 'Home' timeline with a chronological timeline. The study analyzed millions of tweets from politicians and hundreds of millions of tweets from news organizations. Results showed that in six out of seven countries (excluding Germany), the algorithmic feed provided statistically significant amplification favoring right-wing politicians over left-wing politicians. The most pronounced differences were in Canada (Liberals 43%, Conservatives 167%) and the UK (Labour 112%, Conservatives 176%). Right-leaning news organizations also received more algorithmic amplification than left-leaning sources. Twitter acknowledged these findings as 'problematic' and indicated that changes to the algorithm might be necessary. The study involved nearly 2 million daily active accounts committed to a reverse-chronological feed as a control group. Twitter's director of software engineering stated that further root cause analysis was required to determine what changes might be needed to reduce adverse impacts.
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