Join 3,497 readers in helping fund MetaFilter (Hide)


The mood of the nation
April 18, 2012 5:52 AM   Subscribe

By analysing millions of tweets, scientists from Bristol University claim to be able to predict the "mood of the nation" (the nation being the UK) - so much that they say they could have predicted last summer's riots in England. Right now, the angriest country in the UK is Wales, and the saddest is Northern Irleand. The Scots, obviously, are the most fearless. And there's not much joy in anyone's hearts.

Full paper here: Effects of the Recession on Public Mood in the UK

Previously, for the US, by time of day.
posted by ComfySofa (35 comments total) 5 users marked this as a favorite

 
Interesting. I wonder what that giant fear spike on the 8th of April is all about.
posted by fight or flight at 5:58 AM on April 18, 2012


For Scots, rioting is something we do to keep warm...
posted by aeshnid at 6:02 AM on April 18, 2012 [3 favorites]


Hmm, so their data for the whole of Wales is based on the tweets of city-dwellers in an urban cluster some thirty miles across, down in the extreme south of the country. Okay.
posted by Prince Lazy I at 6:04 AM on April 18, 2012


I guessing Easter Bank Holiday travel for the fear spike on 8th April
posted by Gilgongo at 6:05 AM on April 18, 2012


Related: Twitter mood predicts the stock market.
posted by AceRock at 6:12 AM on April 18, 2012


I'm guessing my birthday caused the April 8th fear spike. Only a few more until my unliving minions rise!
posted by kyrademon at 6:14 AM on April 18, 2012 [2 favorites]


To be clear, the researchers didn't say that they could have actually predicted the riots. They went out of their way to say:


"While we leave the interpretation of our findings to social and political scientists, we observed how the period preceding the royal wedding seems to be marked by a lowered incidence of anger and fear, which starts rising soon after that.

(not a criticism of the OP, but of the science journalism in the linked article)
posted by graphnerd at 6:14 AM on April 18, 2012 [2 favorites]


I'm guessing my birthday caused the April 8th fear spike. Only a few more until my unliving minions rise!

Well, that explains why my Creme Egg was filled with spiders.
posted by fight or flight at 6:17 AM on April 18, 2012 [8 favorites]


8th was Easter Sunday - so were there lots of tweets like

"I'm afraid I've eaten far too much chocolate today" ?

It would be interesting to know what terms they use to detect the different mood states. I hope it's better than this, but I'm 'afraid' that it might not be.
posted by DanCall at 6:19 AM on April 18, 2012 [1 favorite]


This certainly is measuring the mood of the general population if we accept that Twitter users are the same as the general population and not, say, richer, younger and affected by different issues (ACTA probably is more of a concern for this group, for example). The paper doesn't address this at all.
posted by jaduncan at 6:23 AM on April 18, 2012 [1 favorite]


Is it right that emotion is measured here simply by the use (or not) of particular words? There must be more to it than that, non? Or else, I'm afraid I'm a bit worried that the results to do with levels of fear, say, could be a bit shocking.

Or that we've just found out that people often say 'Happy Christmas'. Oh:

these effects could be filtered away by a more sophisticated tool designed to ignore expressions such as ‘Happy New Year’

So, could be but... weren't?
posted by robself at 6:25 AM on April 18, 2012


Oh, shazbot. (I hope that was in the list of words)
posted by robself at 6:25 AM on April 18, 2012


Is it bad that I tried to click and drag the lines of the graph? I mean, I just want folk to be happier...
posted by Jehan at 6:29 AM on April 18, 2012 [3 favorites]


No, Jehan. It's not bad.
posted by GallonOfAlan at 6:38 AM on April 18, 2012 [1 favorite]


"Could have predicted"? Call me when they actually predict something.
posted by nosila at 6:44 AM on April 18, 2012 [1 favorite]


I'd like to see this emotional timeline plotted against the "most read" news stories on BBC News, Google News, etc. I'd love to be able to click on each point on their graphs and see a random sample of tweets that were counted as contributing to that tally, to get a feel for how plausible their data are.

Hmm, so their data for the whole of Wales is based on the tweets of city-dwellers in an urban cluster some thirty miles across, down in the extreme south of the country. Okay.

In their defence, that's probably 80% of the population (and, I'd bet, an even higher proportion of the twitter-using population), and was chosen on the basis that Cardiff, Newport and Swansea are amongst "the 54 most populous urban centres in the UK".

On the whole though, I can't get myself too excited about this. It's an interesting idea, but it seems to have all sorts of problems, especially the use of just counting single words to gauge emotional content (I agree that "Happy Christmas" and "I'm afraid I ate too much chocolate this easter" are great examples), and the massive selection bias introduced by sampling (a) only twitter users in (b) large urban areas.

Still, with all the research going into natural language processing and social network analysis, it's a pointer toward what various governments are almost certainly building up to, if not doing already. Being able to monitor the moods and concerns of your citizens in near real-time is an enormously powerful tool, and I can't imagine that they're passing the opportunity by.
posted by metaBugs at 6:47 AM on April 18, 2012 [1 favorite]


Prince Lazy I: Hmm, so their data for the whole of Wales is based on the tweets of city-dwellers in an urban cluster some thirty miles across, down in the extreme south of the country. Okay.

Given how often you see conclusions drawn about England (or even Britain or the UK) based on observations regarding London, it's almost to be expected.
posted by Dysk at 6:52 AM on April 18, 2012 [1 favorite]


The mood in my street is currently existentialist resignation with a side of excessive drinking.
posted by Decani at 6:56 AM on April 18, 2012 [3 favorites]


Finally, I've figured out a justification for the $1B purchase of Instagram. Facebook is going to run an hedge fund and use advanced behavioral image processing techniques to predict the stock market based on analysis from all the photos. Pictures of dancing and smiling will signal an upswing in the market, while pictures of sitting alone in a dark room will herald a downward trend.
posted by demiurge at 7:09 AM on April 18, 2012 [1 favorite]


"We have found a way to predict things in retrospect!"

Yay.
posted by Sys Rq at 7:17 AM on April 18, 2012 [1 favorite]


Prediction is hard. Especially about the future. Have they done that?
posted by DU at 7:42 AM on April 18, 2012 [2 favorites]


Does twitter make it easy to download these tweets, or do you have to scrape them yourself?
posted by Coventry at 8:09 AM on April 18, 2012


The last time England was in a funk, we got the Sex Pistols, The Clash, The Police, and Motorhead. In the happy 90's, the world got BoyZone, the Spice Girls, and Robbie Williams.

I think this is very good news.
posted by Renoroc at 8:29 AM on April 18, 2012 [2 favorites]


The mood in my street is currently existentialist resignation with a side of excessive drinking.

I would like to move there.
posted by yerfatma at 8:57 AM on April 18, 2012


The last time England was in a funk, we got the Sex Pistols, The Clash, The Police, and Motorhead. In the happy 90's, the world got BoyZone, the Spice Girls, and Robbie Williams.

I think this is very good news.


This is exactly the optimistic stance I (as a 19-year-old art student) took when Bush got elected. Turns out, no.
posted by Sys Rq at 9:00 AM on April 18, 2012 [6 favorites]


"Who cares if everything sucks as long as we have artists making art that reflects the suckiness back to us competently."?
posted by DU at 9:09 AM on April 18, 2012


(Besides: Duh.)
posted by Sys Rq at 9:25 AM on April 18, 2012 [1 favorite]


Renoroc: “The last time England was in a funk, we got the Sex Pistols, The Clash, The Police, and Motorhead. In the happy 90's, the world got BoyZone, the Spice Girls, and Robbie Williams. I think this is very good news.”

Like Sys Rq, I used to think this way, too. I stopped thinking this way around the time I realized that I was giving Margaret Thatcher credit for the work of Lemmy Kilmister.
posted by koeselitz at 9:41 AM on April 18, 2012 [2 favorites]


The last time England was in a funk, we got the Sex Pistols, The Clash, The Police, and Motorhead. In the happy 90's, the world got BoyZone, the Spice Girls, and Robbie Williams.

You wouldn't happen to be a Trotskyite, by any chance?

Because despite the crisis of capitalism, I'm pretty sure the workers revolution is not imminent.
posted by PeterMcDermott at 10:06 AM on April 18, 2012 [1 favorite]


That "twitter mood predicts the stock market" paper is interesting, but what it actually says is that they were able to use twitter mood to improve standard prediction models... and that the only "mood" that had an effect was "calm" (not positive/negative or happy/sure/vital/kind/alert), which was reflected 3-4 days later. In other words, they found a way to add "irrational panic" to the standard prediction model lol.
posted by subdee at 10:42 AM on April 18, 2012 [1 favorite]


I'd put more stock into this work if it didn't feel like it was cribbed from a Shakespearian comedy relief scene.
posted by Kid Charlemagne at 11:05 AM on April 18, 2012


The last time England was in a funk, we got the Sex Pistols, The Clash, The Police, and Motorhead. In the happy 90's, the world got BoyZone, the Spice Girls, and Robbie Williams.

As a fan of the Smiths, Cure, Bunnymen and much other 80s goodness, should I be praying for Thatcher 2.0? Besides, the 90s also yielded TIP records, Jump up, and various other forms of electronic awesome, so your argument is invalid.
posted by anigbrowl at 11:37 AM on April 18, 2012 [1 favorite]


No matter: the UK (with or without Scotland) is in slow decline. As for Scots being the most fearless, so are lions as they're being stalked by hunters with scopes. Lots of posturing going on over there. That's what happens when a region is in it's last throes of dominance.
posted by Vibrissae at 1:01 PM on April 18, 2012


Sentiment analysis of tweets is something that a few researchers are currently working on. Mike Thelwall has done (the following are all .doc links, sorry) a lot of work in this area, for example. It's not something that computers are particularly awesome at, although they're OK at it -- the reason is stated by others above: that a lot of people do say things like "I'm afraid I ate too much chocolate" and the machine will see "afraid" and then flag it as having a negative slant, when really it's neutral at best.

Qualitative analysis by humans is much better at doing this kind of thing, and the idea is that by using a huge corpus of human-read data that is appropriately coded (and that on its own is already enough of a problem; getting two or even three people to agree on the sentiment of a tweet is hard work and requires a lot of patience - I speak from personal experience) we can teach a computer to appropriately code the sentiment of things like tweets. So: People are working on it! It will happen in our lifetime! Probably!

On the topic of "cool things that Twitter can do/predict/help us understand" I found the "Twitter predicts box office earnings" (PDF) article to be really fascinating, and the "tweets predict citations" article was great as well (which a colleague and I called "citwaitations" - citing peer-reviewed work on Twitter, which is surprisingly common for academics). And there's been a lot of work on Twitter for monitoring public health and epidemics, but now I'm just talking about how awesome Twitter can be for prediction and I've gotten away from the sentiment analysis angle, which is what this post is all about.

The point is that Twitter is valuable for more than just hearing what people eat for lunch, a phenomenon we called the "cheese sandwich problem." I'm not sure why we had to specify that the sandwich was full of cheese. Perhaps because cheese is awesome? Are Twitter users more likely to be vegetarians but less likely to be vegans? I think I just thought up a new study...

Full disclosure: I have done research in this area, but have only linked to one thing that I had any part in here - the slides from our talk on citwaitions, where I was second author. I didn't come up with the term; my co-author did, and as you'll see from the slides, I argued against it vehemently, which I now regret.
posted by k8lin at 1:25 PM on April 18, 2012 [3 favorites]


I would like to move there.
posted by yerfatma at 4:57 PM on April 18


There are flats and houses for sale on my street right now. For this area of London the prices are not unreasonable. The neighbourhood could only be improved by an influx of even more resigned existentialist piss-artists.
posted by Decani at 5:09 PM on April 18, 2012


« Older Mara Wilson, child star of Mrs. Doubtfire, Miracle...  |  World Population Density Visua... Newer »


This thread has been archived and is closed to new comments