As the use of social media data in research has grown, there has been a degree of uncertainty over just how reliable it is.  After all, there is much to suggest that we’re far from ourselves on social media, and rather than projecting reality, we project how we’d like things to be.

Whilst that hypothesis has a lot going for it, it may not be entirely accurate.  A recent paper authored by researchers from Cambridge and Stanford suggests that a computer program can take our social information and make a better guess about our personalities and moods than those close to us.

The data was taken from the Facebook activity of 86,220 volunteers.  Their likes were analyzed to try and paint a picture of their personality along the traditional big five personality traits so commonly used in studies such as this.

This automated analysis was then compared to how their friends judged their personality.

It turned out that the computer was reasonably accurate after as few as 60 likes.  Give it 300 likes or so however and it becomes remarkably accurate, and much more so than most of the friends used in the study.

What was particularly interesting was that the computer was generally much better at detecting things such as depression than people who actually knew them (or thought they did).

Suffice to say, this kind of analytical capability is not confined to social research, and we’re seeing it creeping into other fields too.  For instance a Stanford based start-up is looking to apply this kind of thinking to workplace management.

It’s perhaps noteworthy that the algorithm feasted itself on data from likes rather than anything from the status updates themselves.  It’s fairly well known that people tend to curate their status updates quite rigorously in order to paint a positive and happy existence.

Indeed, studies have gone as far as to suggest that your happiness in relation to others can be linked to the amount of time you spend on Facebook.  The like history of a person however may be slightly less managed, and therefore give a more accurate impression of the individual.

It’s an interesting finding.  I wrote recently about a vending machine that’s capable of analyzing your health records before allowing you to buy a chocolate bar, and the growth in AI assistants has been quite remarkable over the past few years.  With the likes of Watson gathering pace also, it seems inevitable that algorithms will play an ever increasing role in our lives.

If this study is anything to go by, they may one day soon be able to keep an eye out for our mental wellbeing too, although it should be said that the machine in this instance was the best of a bad bunch, with everyone being pretty poor at predicting.

So a glimpse of the future perhaps, but still some way to go before it becomes reality.

Original post