Twitter's algorithm is now open-source; possibly biased toward Elon Musk
To make the platform more transparent, Twitter has revealed parts of its source code on the internet. Of this, the tweet recommendation algorithm is getting a lot of attention for its alleged biases based on certain parameters like fame and political affiliation. The details have been brought to our notice by tech blogger Jane Manchun Wong and Zoe Schiffer of the Platformer.
What is the recommendation algorithm?
Roughly 500 million tweets are posted on Twitter daily. To provide users with a handful of top tweets, that will show up on their For You timeline, a mechanism is required. This is exactly what the recommendation algorithm does. It distills the data to show users only a few relevant ones. It relies on three parameters for the same.
How are the tweets chosen?
Twitter claims that it relies on three parameters to recommend tweets to users. Firstly, it sources tweets from multiple sources via a process called candidate sourcing. Next, it employs a machine learning model to rank the tweets. Finally, filters are applied so that users do not see content from those they have blocked, already seen tweets, or even NSFW content.
What is the issue?
According to Wong, in the US, Twitter's algorithm labels whether the tweet author is Elon Musk, besides labels like 'Democrat,' 'Republican,' and 'Power User.' Meanwhile, big names like Jack Dorsey, Barack Obama, Stephen Curry, and Katy Perry, are used as "testing accounts" for "getting random tweets for testing." This raises questions about tracking political affiliation, and whether some users are getting their accounts boosted.
Users' stats are being tracked
Schiffer has offered a different explanation for the way the algorithm is written. She claims that the above-mentioned tags do not affect the visibility of users but are simply used to track account metrics. This gives rise to another question. If tweets from democrats, republicans, and power users are tracked separately, is there not a chance of differential treatment?
Improved analytics platform for creators on the way
Twitter claims that it is working on ways to make the user experience more transparent. It will introduce an improved analytics platform for creators with more information on reach/engagement. There will also be more transparency regarding the safety labels applied to tweets and accounts.