Measuring happiness with Twitter and Facebook

We’ve talked a lot about what can be done with data on what information people are looking for online, for example on Google or on Wikipedia. However, we also have lots of data on what information people are distributing online, on Twitter and Facebook, for example. A number of studies have tried to use this data to better understand what affects our happiness, and how happiness spreads. These studies tend to base themselves on work which was previously done in linguistics. Now in this linguistics work, researchers created large lists of words and they gave them to other people to judge, for example, how positive the words are or how negative the words are.

We can now take this big list of words and compare the words in the list to words that people post on Twitter and on Facebook. In an early study, Golder and Macy used this approach to look at messages that people were posting on Twitter and see what that told us about our happiness. They found, as you might be able to imagine, that we really like the weekend. People were much happier on Saturdays and Sundays. You did start to see this happiness drop off on a Sunday evening, though, as we all prepare to go back to work. And equally, you could see it start to grow on a Friday. It seems, however, we really don’t like Tuesdays.

Another study used this same approach and looked at messages posted on Facebook. This study by Coviello and colleagues compared what we were posting on Facebook to data that we have on the weather, specifically how much it was raining. They found in this study that, if it rains, we tend to post less happy messages on Facebook. So the rain seems to be affecting our emotions. Perhaps most interestingly, though, this emotion seems to pass along our network.

So for example, if a friend of yours on Facebook is somewhere where it’s raining and this affects the emotional content of what they post on Facebook, according to this study, you’re then more likely to also post a sadder message on Facebook as a result, even though it’s not raining where you are. A further study, by Kramer and colleagues, took this a little bit further. Now Kramer and some of his colleagues actually work at Facebook, so what they could do was analyse the emotional content of these messages with an algorithm without looking at the messages and then automatically, for a subset of Facebook users, remove the messages which were most strongly positive or which were most strongly negative.

What they found was that Facebook users who saw fewer strongly positive messages were then more likely to post more negative messages and fewer positive messages. Equally, the other way around, users who’d seen fewer strongly negative messages would then post more positive messages and fewer negative messages. Now web companies manipulate their websites to try and affect our behaviour all of the time. This is part of their business. It’s just like salespeople try and affect our behaviour to encourage us to buy more products. However, this study caused a bit of an outcry, because on this occasion the researchers were trying to manipulate people’s behaviour in the name of science, without necessarily having got their consent to carry out this experiment.

On the other hand, however, the results of this experiment were made publicly available. You didn’t need a university subscription or something to see the results. They were available to everybody. Whereas normally, we don’t know what the web companies are doing, and we also don’t know what the outcomes of their investigations are. So what do you think? Where is the ethical line in this big data research? How can we make best use of these new data resources for society without causing anybody any harm?