Hacker News Reader

Comment by antris | Hacker News Reader

>Twitter’s internal research is saying that certain tweets are organically more popular and engaging and more likely to be shown, not that the algorithm has an ideological bias

No one is claiming that there's a piece of code that detects a Twitter user being a right-wing politician and then showing it to more users.

Whether the bias is deliberate or not, it's still a bias. Right wing gets their message amplified more than the left.

View on HN

Perhaps I didn’t explain my opinion clearly, but I think we might have different definitions for what constitutes bias. You’re suggesting that a difference in outcomes means something is wrong or unfair with the algorithm. But what if it is that right wing users simply engage more, or that the right wing news outlets put out better content, or other such factor? I wouldn’t view the increased algorithmic amplification as problematic if there were such root causes behind it. Put another way, I am taking an equality lens (equality of opportunity) rather than an equity lens (equality of outcomes). If we take an equity lens, then you can start questioning why any number of things don’t get the same amplification - it doesn’t have to stop at a binary of “left versus right”.

That's not the definition of bias, in either the colloquial or statistics sense.

A platform promoting one side of the political spectrum over another is not bias? There's no deliberation needed to meet the definition of bias. Every person have unconscious biases, and any structural system, including algorithms that promote content on social media can have unconscious biases as well. Simply stating that it's not bias does not cut it for me.

OK, I admit it's not unequivocally not bias. It could be bias. I would just say there could be other explanations, such as a different ratio between non-politician activists to politicians on either side of the isle and how each slice of the pie tends to monopolize limited attention bandwidth.

If three political tweets by Lebron James are algorithmically boosted, and three political tweets by Ted Cruz are boosted, I wouldn't call that bias necessarily even though the metric being studied did.

Reply on news.ycombinator.com