Censorship
This page is for notes about analyzing censorship of #OWS
Some activists believed #occupywallstreet was censored from TT. This is highly unlikely (see
http://www.quora.com/Twitter-Trends/What-is-the-basis-of-Twitters-current-Trending-Topics-algorithm). However, this belief caused some people to switch hashtags - thus further decreasing the chance of trending.
It should be noted that after switching hashtags, #takewallstreet (?) did trend worldwide on s17/18 although #occupywallst did not.
Yahoo mail censorship
It's also worth describing the series of events around Yahoo Mail's (again, most likely algorithmic/inadvertent) censorship of
http://occupywallstreet.org. (
http://gothamist.com/2011/09/20/adbusters_accuses_yahoo_of_censorin.php). <
<http://twitter.com/#!/YahooCare/status/116220596987232256>http://twitter.com/#!/YahooCare/status/116220596987232256>: "We apologize 4 blocking 'occupywallst.org' It was not intentional & caught by our spam filters. It is resolved, but may be a residual delay."More at <
<http://storyful.com/stories/1000008269>http://storyful.com/stories/1000008269>.
http://www.slate.com/id/2304404/
Gilad Lotan writes:
I agree that its probably unlikely that TTs were censored. That said, the Twitter TTs data seems quite odd.
I save all the Twitter trending topics across the 111 which they provide through their API every 5 minutes. According to this data, the hashtag first trended in Madrid the morning of Sept 17th (17 represents the number of total times the script found that hashtag trending for the given location since the initial time of 07:20:08).
#OccupyWallStreet Madrid 17 2011-09-17 07:20:08
#occupywallstreet San Francisco 108 2011-09-17 18:15:17
#OccupyWallStreet Portland 269 2011-09-17 18:55:15
#occupywallstreet Washington 10 2011-09-17 19:25:17
#occupywallstreet Chicago 78 2011-09-17 20:00:09
#OccupyWallStreet Boston 98 2011-09-17 20:20:09
#OccupyWallStreet Germany 122 2011-09-17 20:30:19
#occupywallstreet Italy 164 2011-09-17 21:05:20
#OccupyWallStreet Minneapolis 2 2011-09-17 21:45:13
#OccupyWallStreet Sweden 217 2011-09-17 22:50:21
#occupywallstreet Australia 8 2011-09-18 04:20:17
These are all locations around the world in which the hashtag trended. Oddly enough, NYC is not in there.
This doesn't mean that it was necessarily censored. Most likely, it had to compete with other, more popular topics, at that point in time. If I query for trending topics in NYC for that day, we see:
Reno New York 2011-09-17 01:00:14 260
#LifeWouldBeBetter New York 2011-09-17 00:00:13 211
#YouLookRealStupid New York 2011-09-17 05:55:14 204
Straw Dogs New York 2011-09-17 02:40:13 188
You Got Served New York 2011-09-17 04:00:13 166
The Lion King 3D New York 2011-09-17 00:00:13 155
#WorstPickUpLine New York 2011-09-17 00:00:13 146
How High New York 2011-09-17 06:15:13 132
Metta World Peace New York 2011-09-17 00:00:13 123
Big East New York 2011-09-17 05:20:13 107
#offthis New York 2011-09-17 10:05:13 103
#kindofabigdeal New York 2011-09-17 00:00:13 76
#WillGetYouSlapped New York 2011-09-17 17:25:13 73
Mayweather New York 2011-09-17 11:05:13 68
Next Friday New York 2011-09-17 06:10:13 68
#DrakeCriesWhen New York 2011-09-17 01:20:14 61
#takewallstreet New York 2011-09-17 18:50:13 55
Lion King 3D New York 2011-09-17 11:50:12 54
Clemson New York 2011-09-17 19:25:13 49
ACC New York 2011-09-17 18:40:13 41
.... and the list goes on...
Jillian York points to some existing work about censorship of TTs:
[[http://opennet.net/blog/2010/12/wikileaks-twitter-trending-topics-manual-interference-or-algorithms-usual]]
[[http://opennet.net/blog/2010/06/iranelection-censored-evaluating-twitters-trending-topics]]
ONI has a couple of other analyses, including one on the #flotilla hashtag, which a number of activists complained was being suppressed:
[[http://opennet.net/blog/2010/06/flotilla-censored]]
[[http://opennet.net/blog/2010/06/mexico-cananea-censored]]
Of course, Twitter has admitted to manipulating hashtags ([[https://support.twitter.com/entries/101125-about-trending-topics]]), particularly those related to Justin Bieber.
I also have it on good authority that they suppress hashtags with profanity
@mstem says: Short version: The trending service favors sharp upticks over total volume.
This year in review of trending topics could be useful if you ever do research on the topic:
http://yearinreview.whatthetrend.com/
@techsoc says:
I once talked to the guy in Twitter who oversees the whole trending topics operation. I asked him if there was manual review before going live--or if trending topics were fully algorithmic. He did say there was manual review at the final stage and they filter for things like profanity or against attempts to game or spam trending topics.
Plus, he explained that there were many complexities to picking out what's trending from the background. So, for example, they have to decide if something is just background (a common word) or just generally talked about a lot (wonder why "Bieber" is not always trending?) versus something that rises because, well, it's doing the thing we think of when we think of "trending"--rising out of the conversation.
So, yes, there are definitely human choices baked into the process.
Whether they consider "occupy" to be a no-no word compare to "take" is something we should try to ask them.
Gilad again:
Continuing our discussion on possible censorship of Twitter TTs, I've just published a bunch of data from the past four weeks of #OWS tweets, using both total volume of appearance (of hashtags/terms in tweets), and times/locations of these terms within the trending topics lists.
Whats interesting is how #OccupyWallStreet has never once trended in New York. Similarly, even with this week's terrible turn of events, #OccupyBoston has never trended in Boston. I'm pretty convinced that this is all algorithmic, and while some aspects can be reverse-engineered, there are way too many variables.
Two points to note: 1) people in Boston just aren't using hashtags as much as we'd think they'd be, and 2) Kim Kardashian's wedding is that much more enticing to tweet about, even for Bostonians...
What excites me about TTs data is how it reveals attention given to topics, and how they spreads among geographic regions. For example, there's a two day period where #OccupyWallStreet was heavily trending in the UAE and nowhere else. It was right after the Brooklyn bridge arrests, which might have generated an emotionally stronger response from users in the Emirates. Also, unsurprisingly, Spain was the first region on Twitter where #OWS trended...
[[http://blog.socialflow.com/post/7120244374/data-reveals-that-occupying-twitter-trending-topics-is-harder-than-it-looks]]
G
Implications of Algorithmic choice
zeynep writes:
1- Wow to the power of algorithmic choices! Choosing "velocity over popularity" is a very consequential decision. So even if this is "neutral" in the sense that once set in motion, it just runs without too much human interference (though Twitter does censor for profanity or attempts to game the system) this structure replicates something that television and traditional news coverage has long been criticized for--prioritizing the sensational over the structural.
Tools:
what tools are out there that people (now and in the future) might use to independently verify or disprove such claims. Wd be curious for tools that address this on FB or G+ as well. Or, a toolset that quickly enables someone who fears censorship on an SNS to verify.