Subscribe to my newsletter
Confluence discusses technology, science and society, and prompts you to think critically about your world. Dispatched fortnightly.
16 On bias in Facebook trending topics
I quit Facebook years ago and have almost never been interested in discussing its issues, downfalls or triumphs, but the recent row about the social network being biased in choosing its trending topics — specifically regarding accusations that republican news was muffled — has made me want to talk about this, albeit briefly. First of all, I think little of politicians calling for a congressional hearing. They often call for whatever moves to strengthen their stance at that point in time. This does not mean I support one part or other in this feud, but I have been fervently hoping that people see the bigger picture.
Facebook recently revealed in an internal document the steps they take to choose their trending news stories. While I appreciate that they spoke of using a thousand news sources etc., only one point struck me: there is human involvement at one level — human editors cross-reference if a story flagged by their algorithm is being covered on ten select news sources. And where there is human involvement, there is bias. The bigger picture to me, here, is that this is precisely the kind of job you want being handled by a robot. A quintessentially emotionless, unbiased, by-the-book task following rules built on popularity and geography and not personal preference. That is not to say the “human editors“ at Facebook tweaked the system on purpose, but, as humans, we do tend to let bias creep in.
In any case, even if Artificial Intelligence will not be our future, it will certainly be a part of it, and now is the right time to frame laws and insist on their use (as well as clearly disallow it) in specific scenarios. The issue here is one that concerns much more than the elections everyone seems to be worried about.