Twitter's machine learning algorithms amplify tweets from right-wing politicians over those on the left
Engineers and researchers don't know why
Twitter's algorithms are more likely to boost right-wing content than left-wing posts from politicians and news publications, according to a recent study.
A team of engineers working on Twitter's own ML Ethics, Transparency and Accountability (META) unit scraped millions of tweets of thousands of elected officials from seven different countries: US, Japan, UK, Canada, Germany, Spain, and France. They tracked how likely these posts made between 1 April 2020 and 15 August 2020 were to be placed in a higher rank in users' personal Twitter feeds using Twitter's algorithms.
Tweets posted by politicians from right-wing parties were amplified more than those from left-wing parties in all countries except Germany. The effect was strongest for Canadian and British politics. For example, content from UK Labour MPs was amplified 112 per cent as opposed to the 176 per cent amplification of Conservative MPs' content; and Canada's Liberal party politicos were amplified 43 per cent versus 167 per cent for the Canadian Conservative party.
The percentages were calculated by comparing how often a tweet ended up on a user's Twitter timeline when they had enabled "personalized relevance model", a system that shows the most relevant posts first, compared to "reverse chronological feed", where the latest tweets are displayed. Tweets that rank highly on people's timelines are more likely to be viewed and shared.
"Values over 0 per cent indicate that all parties enjoy an amplification effect by algorithmic personalization, in some cases exceeding 200 per cent, indicating that the party's tweets are exposed to an audience three times the size of the audience they reach,” according to a research paper published [PDF] by the team.
It's unclear why conservatives are favoured over liberals. There are a lot of different factors to consider and many algorithms operating in tandem. Tweets from right-wing politicians are more likely to be boosted over left-wing ones regardless of which party is in power. But party affiliation or political ideology isn't necessarily a strong indicator of whether a tweet would be more likely to be retweeted or not. Elected officials from the same party, for example, are not amplified equally.
"In this study, we identify what is happening: certain political content is amplified on the platform. Establishing why these observed patterns occur is a significantly more difficult question to answer as it is a product of the interactions between people and the platform," said Rumman Chowdhury, director of software engineering, and Luca Belli, a staff Machine Learning researcher.
- Facebook overpaid FTC fine by up to '$4.9bn' to protect Zuckerberg, lawsuits allege
- Banned: The 1,170 words you can't use with GitHub Copilot
- Biden order calls for net neutrality, antitrust action, ISP competition – and right to repair your own damn phone
- Now that Trump is useless to Zuckerberg, ex-president is exiled from Facebook for two years, possibly indefinitely
The same effect happened with news content too. Articles from right-leaning publications were more likely to be algorithmically amplified than left-leaning outlets. For example, Fox News was favored more than BuzzFeed. It's difficult to see why this might be the case, however, without seeing the data.
The META team is currently figuring out how to share its raw dataset to researchers and developers hoping to replicate the study without infringing on the privacy of users' accounts.
"For the past several months, META has been looking into methods to responsibly make available large datasets to support validation. We're finalizing a partnership to leverage privacy preserving technology to enable third-party researchers to reproduce this type of work, while also protecting and safeguarding the privacy of people who use Twitter," it said.
A scaled back version that doesn't contain all the details, however, can be made available to researchers upon request. The Register has asked Twitter for comment. ®