More than 1,000 people descended on Baltimore Saturday to protest the death of 25-year-old Freddie Gray, a black man who died in police custody last week and whose name has been listed alongside Michael Brown’s and Walter Scott’s as evidence of mounting police brutally.
The protests got violent; six police cars were damaged and police in riot gear arrested 12 men. The story got prominent coverage on CNN and ABC, and in papers from the New York Times to USA Today. More than 150,000 people tweeted April 25 on the #FreddieGray hashtag.
And yet, amidst all that furor, #FreddieGray never trended nationally on Twitter: According to Trendinalia, a site that tracks historical Twitter trends, it never got above the #21 position. Which is no longer just a bragging right, you understand — it means the protest didn’t surface in the right handrails (and newly prominent trends tabs) of millions of Twitter users who may not otherwise have heard the news.
|April 23||April 24||April 25||April 26||April 27|
|#takeyourchild toworkday||#ArborDay||#WhereIWasWhen ZaynQuit||#FattenAMovie||#MondayMotivation|
|#ThoughtOf TheMorning||#SaySomething NiceAboutHillary||Kris Humphries||#LarryALove ToRemember||Konami|
|#LiamIsARoleModel||Blue Ivy||#RuinARomantic MomentIn4Words||That’s So Raven||Billy Corgan|
As is often in the case with this type of thing, an algorithm is to blame. See, as the data scientist Gilad Lotan explained in an analysis published Friday, the algorithmic system that determines Twitter’s “trending” topics favors things that spike quickly, rather than topics that build a large, sustained audience over time.
That makes sense, when you think about it: If trending topics were determined by volume alone, we’d never get anything that didn’t involve Zayn. (And it often feels like that, anyway.)
But even though the system makes sense, it often produces results that … don’t. Lotan initially made this discovery about the Twitter algorithm in 2011, when the hashtag #OccupyWallStreet failed to ever trend in its home base, New York. Meanwhile, momentary drivel like #KimKWedding and #ICantRespectYouIf did get surfaced.
“If we see a systematic rise in volume, but no clear spike, it is possible that the topic will never trend,” Lotan wrote at the time. That means three things, he said:
1. The longer a term stays in the trending topic list, the higher velocity [of tweets] required to keep it there.
2. It is much easier for a term never seen before to become a Twitter trend, and finally
3. It is extremely important to understand what else is happening in the region or network (if Kim Kardashian’s show is airing, you can forget about trending!).
Those are pretty dramatic implications, when you think about it: Had #Ferguson happened the same day Bruce Jenner came out, would it have taken off in quite the same way? At the time, after all, Facebook took a lot of flack for failing to put Ferguson in its “trending” tab, privileging silliness like the Ice Bucket Challenge. Speaking on a panel about algorithmic control earlier this month, the sociologist Zeynep Tufekci accused Facebook of “effectively censoring the movement.”
It rings false to accuse Twitter of the same algorithmic transgression; after all, Twitter is usually held up as the paradigm of the transparent, unfiltered social feed. But maybe it’s a good reminder that the biases built into algorithms — even the truly useful, valuable ones — can have a profound impact on what bubbles into the national consciousness and what does not.
Given that level of unchecked influence, we really need outside algorithmic audits like Lotan’s. At the same talk where Tufekci criticized Facebook, Lotan encouraged more data scientists to examine algorithmic systems — to try to “map the inputs to outputs,” the #FreddieGray tweets to the #SundayFunday trends.
“We can hold the system accountable then,” he said.
Liked that? Try these: