Decoding the Twitter algorithm: recommendation of news content
With the recent news of the purchase of Twitter by Elon Musk, and the rapid changes going on in the business as a result, uncertainty about the future of the platform gives users, advertisers, and organizations plenty of cause for concern. Whether we like the platform or not, it is hard to ignore the role it has played–and likely will continue to play–in people’s lives, public opinion, and the democratic process.
Social media platforms have been criticized for exposing users to mainly belief-confirming information, generating filter bubbles, and sometimes promoting hate speech, conspiracy theories, and fake news. This results from selection algorithms whose primary goal is to help us sift through the endless stream of new information available online, typically by selecting popular content we are likely to find interesting and relevant.
The specific way in which these algorithms work is a well-kept secret by social media platforms, including Twitter, who are looking to protect their intellectual property. The recommend systems that use these algorithms are the key to the success of platforms such as TikTok, Instagram, Facebook, and Twitter. This secrecy can be very problematic considering the effect they have on how we perceive the world.
In a recent study conducted with Marcel Garz, we investigated how Twitter’s algorithmic content selection affects German newspapers, by looking at the change from a chronological timeline to an algorithm-recommended feed. From our research, we found three key elements of how the algorithm works.
1. Rich-get-richer effect:
We found that the Twitter algorithm amplifies popular news topics and popular newspapers. Thus, the more popular a news topic or newspaper is, the more Twitter’s algorithm will promote it to users. This is what we call the rich-get-richer effect, where those accounts that have a bigger presence on the platform will get the most exposure, therefore amplifying their presence even further.
For example, national outlets, mainstreams newspaper, and tabloids with a higher number of followers or that are followed by the user’s contacts would be suggested to the user’s feed. In contrast, the algorithm will show less content from local outlets and niche newspapers.
What this means is that it’s harder for fresh voices to break through. Small groups of large accounts easily control the narrative on a topic, and the same perspectives are dominant within social circles.
2. Favoring sensationalist content:
In our work, we also found that the algorithm will privilege shorter tweets with emotionally loaded words and exclamation marks. In effect, long Tweets on #UpdateOfRelativityTheory stand no chance against short emotional tweets like #Suspicious!
These kinds of content choices are typically associated with sensationalist journalism, which is favored over quality news stories in the ranking of content provided by Twitter’s algorithm. Tabloids and low-quality journalism are more likely to be disseminated on the platform, and because of the psychology of how we react to emotional content, these posts are likely to get even more dissemination via retweets by users.
What this means is that those news outlets that are looking to polarize and stir emotions are being favored by Twitter over those who present a nuanced take on a topic. Quality information and journalistic integrity are coming second to emotional reactions and sensationalism.
3. Keeping our attention and engagement:
In our study, we found that the Twitter algorithm will give an advantage to tweets that drive user engagement through retweets and likes. For example, the algorithm will prefer a question or a quote to increase the number of retweets.
Ultimately, the role of the Twitter algorithm is to promote content that will benefit their interest in keeping users engaged and spending time on the platform. These interests don’t always align to provide quality journalistic content, balanced information, and in-depth discussions.
Why this matters?
The way these algorithms work is relevant because people assume naturally that the content at the top of their feed is the most important and most accurate, and they completely ignore content that gets pushed down or entirely removed by algorithms. And while news content selection and curation are not new or exclusive to digital platforms–just think of the editorial slant of some traditional media outlets–when the choices are being made by proprietary algorithms it becomes harder to tell how the decisions are being made and whether they are showing an accurate picture of the world.
Considering the public debate about increasing polarization and fake news, the European Parliament passed into law the European Digital Services Act (DSA) in October 2022, which requires platforms such as Twitter to offer transparent accounts of the functioning of their algorithms to regulators, academics, and society.
It remains to be seen how long the algorithms will continue to favor the loudest and boldest of social media content. In the meantime, the key to exposure does not seem to be high-quality, fact-checked content but quick, quirky, and captivating news flashes.