Slashdot Mirror
Twitter Buzz As an Election Predictor
Capt.Albatross writes "A study presented at the American Sociological Association's annual meeting suggests that simply comparing the frequency with which the candidates' names are mentioned in tweets can predict the result of elections almost as well as conventional polls, even without considering the sentiment (for or against the named candidate) of the messages. Furthermore, the correlation seems strongest in close elections. Additional commentary can be found at the Wall Street Journal and from Indiana University."
Google "buy twitter followers" and you will see a lot of companies dealing with this. Most of the time when you get random followers who just tweet the same format of things, its because they are bots.
Whichever candidate can afford hire the most companies to have bots repost what they are tweeting have the highest chance of winning. It all comes down to who has the most money for advertising, same as always.
Yes, perhaps a good predictor now, but only up to the moment when results of these polls are widely publicized... and some company offers to manipulate the trending words for a price?
That is only useful as long as no one bothered to notice this point consciously?
Get thee glass eyes, and, like a scurvy politician, seem to see things thou dost not.--King Lear
Mind you, this is cheap way of astro-turfing. But beyond the most superficial analysis, astro-turfing fails quickly. especially where reputation is considered. Create a couple of thousand of twitter bots ? Easy. Getting real people to follow them ? Hard. . .
My wife worked for an online "Marketing Sentiment" company, and she came up with the idea of trying to predict the American Idol winners using a very similar technique about 4-5 years ago (I forget exactly).
She found that it showed increased "excitement" in general terms, and generally there was a correlation with increased voting on American Idol, but it did not accurately predict the specific winners each week.
It sometimes seemed to work, though, so she (and others) messed with the algorithm quite a bit and tried to make it work. I.e. wasted a lot of time on it. Maybe someone with a more scientific approach would have succeeded, but the few times she had me look at the raw data (even with sentiment included) suggested to me a non-obvious relationship, and maybe not a predictive one.
Erich Boleyn
How's that Twitter thing working out for Anthony Weiner?
And...Twitter being the sole domain of the Narcissists, I don't think it is anywhere near a representative sample of the voting population.
When Fascism comes to America, it will call itself Anti-Fascism, and tell you to give up your guns.
Correlation does not imply causation. http://en.wikipedia.org/wiki/Correlation_does_not_imply_causation
...we can vote in elections via Twitter.
following events on Twitter is a blast. I did not watch the presidential debates, I just followed the twitter hash-tag. It is like watching a movie audience where you can see the film or even hear it, you just have to judge what is happening by the audience's reaction. It is much more revealing than you would imagine. For example, you can see who is going to win the debate 15 minutes into the debate. The dynamic is established that quickly. And there are surprises. For example, when "binders full of women" suddenly started showing up on my feed I thought, "Romney could not possibly have said anything so bird brained." but he did. Twitter is my all time favorite way to follow public events.
successful twitter campaigns will increase voter turnout in the voting polls (for both parties). the voting booths and locations will not be equipped well enough to handle the normal volume of voter turnout because most of the sheeple are used to staying home rather than actually taking the time to vote. they'll show up and thousands of democrats will find that they should have registered to vote but didn't and so now they cant, or else they will have registered but some republican somewhere in the chain of communication will have conveniently forgotten to process their paperwork. then there will be angry "mobs" all fuelled by twitter, so of course now we'll have to censor it, and so on and so forth.....
The pretext makes for good media fodder, so it's hard to judge the plausibility of the results without looking at the actual study. Couldn't find it in the Washington Post article or through a quick search. A few questions that come to mind..
- How many elections were studied? If it's a small number, then are the models overtrained? It's easy to come up with a model that connects two data sets if the data sets are known !
- Are "negative" tweets distinguished from positive ones in some way? If people dislike a candidate then they are probably as likely to post on them. e.g. try running this model on Donald Trump when he was talking about getting into the 2012 race. If you could distinguish "votes" and "antivotes" reasonably well then the idea should become more defensible.
Just great... Cue tens of thousands of paid campaign staff trying to boost their candidate's stats.
http://votecompass.com/images/au-2013/auspol-twitterverse.jpg
EMail: 0110001101100010010000000110001101110010 0110000101111010011011100110000101110010 0010111001100011011011110110
In the 2008 election, there was an enormous interest in Ron Paul. His google searches were higher than any other Republican nominee for most of 2007 and 2008. His Twitter interest was huge.. http://www.nethosting.com/buzz/blog/ron-paul-wins-gop-nomination-if-twitter-votes-counted/
But, things are not transparent like that in the US. The mainstream media controlled the 'buzz' around Ron Paul and continued to act like he had little chance. The google trends and Twitter followers were ignored. The Internet buzz was discounted.
Maybe this article will be accurate in the future as the Internet takes over (and more so the Internet Generation takes over).. But if you tried to predict the last election based on Twitter, you would either be thinking there was massive fraud or there somehow was a huge amount of the US population that never heard of the Internet.
--- We need more Ron Paul!
It seems that now that this has been observed, outcomes are bound to change.
"We shall grapple with the ineffable, and see if we may not eff it after all." - Douglas Adams
that a large enough population of people who vote will actually say who they vote for given the chance and that result reflects reality.
Not a single graph on any of those links...
God bless you, Washington Post.
They actually went to the trouble of including a damn chart, which shows just how weak the correlation actually is.
Umm, no offense to our friends at Gallup and such, but shouldn't we set the bar a little higher?
Correlation does not imply causation.
But in this case, correlation can lead to causation.
1. Send lots of tweets to simulate popularity.
2. Get publicity for being "in the lead".
3. Get more attention from journalists, debate organizers, and potential voters.
4. Get donations from special interests that want to back a "winner".
5. Profit!
...because the moment something like this is identified, it will be gamed.
-Styopa
Correlation does not imply causation.
Obligatory http://xkcd.com/552/
https://app.box.com/WitthoftResume Code: https://github.com/cellocgw
Ok, here's the thing. If Twitter use/mentions currently are a good predictor of election results, then that's cool. However, when you make it known, you invite a bunch of intentional skewers to the mix thus destroying the instrument.
in an inverted fashion from how Nate Silver surfaced way back in 08 that under-representing cell-phone-only voters in polls under- represented the young, it'd seem that using twitter as a representative sampling will under-represent the old - e.g. those who vote more than the young do? http://fivethirtyeight.blogs.nytimes.com/2012/09/19/obamas-lead-looks-stronger-in-polls-that-include-cellphones/?_r=0
You don't even need bots, just post a few pics of your boner.
"I'm not much interested in interoperability. I want substitutability. I want to be able to throw your software out."
Need any more proof your elections are rigged ?
End of Line.
Considering that 45% of Americans under age 35 turn out to vote and 70% of people over 65 vote, there is a high likelihood that a person tweeting about a candidate *isn't* going to vote for that person. And that the people that are going to vote have never seen Twitter before.