Why won't Wikileaks trend on Twitter?

Twitter is "censoring" discussion of Wikileaks by preventing it becoming a trending topic, writes blogger Bubbloy. As he points out with some neat research, the popularity of Wikileaks discussion far outstrips that of the pop culture ephemera that does trend. But is there a clue in the fact Twitter's feature isn't called 'popular topics?' Perhaps the #Wikileaks tag has already experienced so much chatter that it's become algorithmically unlikely to retrend. If so, perhaps a page for trending topic 'graduates' — or at least a limit on the algorithmic relevance of last year's spikes — would be nice.


  1. Or could it be Jack Dorsey is still collaborating with the U.S. State Department? From LRB:

    On a balmy evening in April 2009 Barham Salih, then deputy prime minister of Iraq, sat in the garden of his Baghdad villa while a young internet entrepreneur called Jack Dorsey tried to persuade him that he needed to be on Twitter. Dorsey, the founder of Twitter, was in Baghdad at the invitation of the State Department.

  2. I’m turning off the conspiracy-theory centers of my brain for a second. If it is an algorithm that “expires” trends, isn’t that what keeps that list relatively “Bieber-Free?” And if so, isn’t that a good thing? Not like people are unaware of WikiLeaks.

    1. You should probably bother to read the article first, as the evidence in the article is quite compelling and deals specifically with this issue. Similar examples of highly popular but sustained info stick for long periods as in the example of the oil spill. Obviously it’s impossible to know for sure unless they provide us with the algorithm, but given the information in this article, it’s very suggestive.

      The fact that wikileaks is miles above anything else ever means that if it isn’t a conspiracy the algorithm is broken.

      1. @jacobian, you might want to read on a little further into the comments; there is a nice discussion about (as Rob so nicely puts above) the difference between a ‘trending’ topic and a ‘popular’ topic.
        Purely speculation on my part, but here goes….
        The two points that stand out (as mentioned by a commenter claiming to work on trends at Twitter), are that a trending topic must be ‘diverse’ and have an ‘organic increase’. It isn’t clear what specific definitions they’re using, but it is quite likely that tweets with #wikileaks by a user who is a follower of @wikileaks wouldn’t pass the diversity test. This would address the whole create-a-user-to-stop-a-trend issue — you could create a user named ‘Inception’, but that wouldn’t suppress #Inception trending, since those tweets aren’t close to the newly created user in the social graph.
        Looking at the stats in the linked post, it may be also that the volume spike is too steep to fit their definition of ‘organic’ (again, purely speculation on my part). It would be interesting to see a 3- or 10-day rolling average curve, instead of spot values.

        1. You have completely failed to read the article as well IsolatedGestalt. The data is compared to profiles which trended for much longer with data that was more monotonically decreasing. It’s pretty lazy to pretend like I haven’t understood the issue when you haven’t even looked at the data or the article.

          1. @jacobian, perhaps I was unclear, and — if so — my apologies. I was not suggesting that you hadn’t read or understood the _article_, but that it would be worth your while to continue reading the discussion as it unfolded in the comments. Whether or not you consider the extended discussion to invalidate the premise of the article, it certainly provides some more food for thought.

            Yes, the article discusses other profiles and their raw volume curves compared to the #wikileaks raw volume curve, from which data Bubbloy’s concerns are reasonably drawn. In the comments, however, the discussion continued into reasons that the given data may not be applicable in the way that you would expect. The primary bit that caught my eye had to do with tweet ‘diversity’, which may explain both the “username vs. hashtag” issue and the “popular != trending” issue.

            To expand on what I said above, and to rehash some of that other discussion, a topic XY could be considered ‘diverse’ if some significant number of tweets arise from a significant number of sources from whom XY is _not_ an expected discussion topic.

            To put it in local terms, there are a pretty good number of comments here on BB every day, but those are expected — we chose to come here, after all. By themselves, these raw numbers of comments don’t indicate much about BB’s cultural relevance. If, however, BB were to show up as a trivia question on a non-tech network game show, that one instance would carry much more weight in terms of cultural relevance.

            Similarly, a twitter user who follows @wikileaks would certainly be expected to use the #wikileaks tag, so their usage of the tag expresses much less about the “trend” than if that tag is used by someone who only follows their local veterinarian (not that those are mutually exclusive interests, of course). Because @wikileaks has so many followers, this kind of graph-based weighting could have what appears to be a very strong filtering effect, rendering a raw-volume-curve comparison pretty useless.

            I don’t have any connection to Twitter, so this is purely speculation, but it would explain the disconnect between the claim that “usernames can’t trend” and some provided examples to the contrary, in that the counterexamples are likely not close in the graph to the tweets using a tag with the same name. For example, you couldn’t suppress #happythoughts by creating @happythoughts, since your new user wouldn’t be closely connected to those tweeting #happythoughts.

            The overall point is that “Why won’t Wikileaks trend on Twitter?” can be answered — reasonably — without recourse to nefarious conspiracy. That isn’t to say that no such conspiracy exists, just that it isn’t required to explain the data.

  3. I think part of it may be the clustering of Wikileaks and Cablegate tweets in relatively dense pockets… Still, to me, it seems to indicate something is wrong with the algorithm.

  4. Sounds like Twitter is pulling a Google and showing only relative spikes in volume. This is what they used to subdue Bieber fever:


    “The new algorithm identifies topics that are immediately popular, rather than topics that have been popular for a while or on a daily basis, to help people discover the ‘most breaking’ breaking news from across the world. (We had previously built in this ‘emergent’ algorithm for all local trends, described below.) We think that trending topics which capture the hottest emerging trends and topics of discussion on Twitter are the most interesting.”

  5. The reason the #wikileaks tag didn’t trend on Twitter was very simple. Wikileaks themselves asked people to use the tag #cablegate for tweets about the leak. So that tag was widely used, and appeared in the list of trending topics.

    If anyone conspired to stop #wikileaks trending, it was Wikileaks.

  6. I understand that usernames don’t trend. As @wikileaks is the name used by Wikileaks, it doesn’t get on the list. This is why wikileaks recommends cablegate as the hashtag

    Infamy.infamy..they’ve all got it in for me.

    1. So you’re saying that you could immediately knock a trending topic off the list by registering it as a user-name? That doesn’t sound right. Too easy to manipulate. Plus, I checked a few of the trending topics, and there were indeed registered accounts by those names (granted, very, very small accounts with few or no followers).

  7. I think the “usernames don’t trend” has been misunderstood, it’s just username mentions that don’t trend, i.e. if people write @wikileaks it won’t be counted.

    But it would do Twitter very good to release their list of blocked words, I guess “#fb” and “RT” are on it?

  8. It was trending under the hashtag #cablegate a week or so ago. It’s not anymore because either 1) fewer people are talking about it relative to other noise that the twitterati care more about or 2) the Bieber effect when Twitter changed their trending algorithm to show only abnormal spikes — which would mean it’s a hot topic with a high level of discussion, but no spikes to indicate anything above what is “normal” for the topic.

    It’s not censorship. Yes, if anyone has a reason to fear a government conspiracy it probably is Wikileaks, but do you *REALLY* think there is some secret cabal in a back room saying “I know what will really get these infopirates! We’ll keep them from trending on Twitter!”

    1. Yes that is what is being suggested. Did Liberman call Twitter and decree that allowing Wikileaks to trend is treason? Were you on a media fast last week? ;>

    2. Yes this post suggests that someone like Liberman called Twitter and decreed that trending WikiLeaks is treason. The intimidation directed at Amazon and Paypal by government employees last week were unconstitutional actions against free-speech. If you call this conspiracy then I guess that’s what it is.

  9. The other week, there was some paranoia about censorship when the #demo2010 hashtag — used in the UK by the very well-attended and very heavily-tweeted student protests — stopped appearing on the trending list.

    The protestors shifted to a different tag (#dayx) which trended nicely for the duration.

    I’m not sure whether the “only spikes trend” thing would apply. It seems to at first, but I’d have expected #dayx to disappear from the list and it was around for a good chunk of time.

    Hey Twitter! You read BB, so get on here and set this straight before we assume you’re just a mouthpiece of the Man!

  10. @pridkett, you don’t have to believe in ‘conspiracy theory’ to realise that information management is one of the core jobs of state intelligence agencies. That’s one of the reasons why the Wikileaks leaks are so challenged to states (and not just the US) in the first place. We’ve already seen how the US state is using its influence over all the communications outlets that it is able to influence to conduct damage minimization (and to bring Wikileaks down in the longer term). Considering how important Twitter is in terms of determining what is ‘news’ and in political organisation more broadly, it would actually more more surprising if the states that had the ability and the will to affect it didn’t care and made no effort to influence Twitter.

  11. I’ve never heard the term “Cablegate” until this thread. Maybe it’s just not popular enough to trend up?

    Personally, I hate all new -gate names unless they’re funny or ironic. Like if there was a (large) scandal about the passing of legislation, we could call that “Billgate” and I’d be happy.

  12. #Twitter #is #stupid. #Now #maybe #if @Google #was #censoring #search #results #in #the @USA i cud c a rzn to gv a sht.

  13. Cablegate? Why? Did the leak happen at the Cablegate hotel? Fuck this “story+gate=good” laziness.

  14. I believe it is because the investment bankers who fund a money-loser like Twitter have made a few things very clear to those who run Twitter. And PayPal. and Amazon.

    and oh yeah, SWISS BANKS, know throughout for seizing the assets of those accused of crimes in thrid countries with minimal evidence.

    The US could not get Swiss banks to seize the assets of convicted murderous dictators this fast, so ask yourself what’s really going on here.

    Ask. Keep asking.

  15. Ignore those cables! Pay no attention to the man behind the curtain! Look! over there – It’s Lady Gaga pooping!!!

  16. I don’t think that Assange will be able to continue with his work, since they are searching him all over the world.

  17. So, by extension, the reason nobody notices except the supporters of a particular trend is that, really, none of the Biebernauts and Gaga Monsters care enough to call “conspiracy!” — at least, not compared to the Serious Business folks.

  18. ‘The overall point is that “Why won’t Wikileaks trend on Twitter?” can be answered — reasonably — without recourse to nefarious conspiracy’
    Yes it can. While I thoroughly believe that the US is manipulating corporate and or foreign government entities in other fields regarding this topic, despite the US’s statements to the contrary, this is an instance where Occums’s razor applies to the exclusion of conspiracy. Occum’s razor submits that the _least_ amount of speculation _possible_ be applied (note the operative words). In the instance of paypal or amazon, the very same concept of _least_ assumptions _possible_ applies to determine there was government (or pawn) interference. Any speculation towards the algorithm is required and evidential. If censorship and conspiracy were at work here there is no doubt that the @wikileaks account would be followed by the censors and it would have been evident to any one doing so that the #cablegate hash would have needed to be demoted as well as the #wikileaks hash, this would have been noticed when @wikileaks _told_ people to use it. The argument can of course be made that the gov. only censored the #wikileaks hash in an effort to be even MORE subversive and only prevent people who would otherwise not known about it to find it while leaving the people who already did know about it talking amongst themselves in a corner, but the entire trending algorithm argument (along with the quantifiable evidence that suggests it) would have to be thrown out by consequence and as such would be a violation of Occum’s razor. But then one would be in the same boat as the administration who took us to a superfluous war in Iraq, the ‘straw man’ boat, it sinks fast. Plus most the twitter users WOULD rather follow along as lady bieber and justin gaga ate each others poop while eric estrada or hulk hogan commented a play-by-play.

Comments are closed.