Dr David Stillwell, University of Cambridge
How well does your best friend, or even your mother, know you? And how do you think their judgement of you would compare with a computer? New research out this week in the journal PNAS reveals that the computer is a better predictor of your personality than either of them. Cambridge scientist David Stilwell has created a program that can use the things you like on Facebook to make detailed personality judgements. Interestingly, as few as 10 likes provide a more accurate picture of you than a work-colleague can. Chris Smith reports...
David - Well, this is the culmination of about 8 years research. So in July 2007, we put a test on the Facebook social network and got people to take the personality test and then get feedback on the results. As part of that, they could opt in to share their Facebook likes. Over the past 2 or 3 years, weíve been interested in how accurately we can predict someoneís personality, just from their Facebook likes.
Chris - In other words, by trawling across a lot of data about what people self select to say that they have a liking for, then you're able to draw conclusions about what sorts of things in general will be true of that particular person.
David - Exactly. So, there are about 100 billion things that people can like on Facebook. So, theyíve got a massive choice to choose from. On average, people like about 220 different things.
Chris - Do you have to be a very prolific user of Facebook for you to get an accurate appraisal?
David - Well, given an average number of Facebook likes, the computer is actually as accurate as someoneís spouse at understanding their personality. So thatís more accurate than their family, more accurate than their friends, and they're far more accurate than their work colleagues.
Chris - When you say you know them better than their spouse, what sorts of things can you say you know?
David - So, this is based on the self report questionnaire that people did. So, we measured 5 traits in the self report and these are the so-called big five personality traits. Itís quite broad factors such as how extroverted you are, how open to new experiences, or how sort of agreeable so that means interested in social welfare versus quite competitive.
Chris - Are there wider implications and uses of these?
David - So, the immediate implications are for personalizing peopleís online experience. So, deciding which adverts you view and hopefully, making those adverts more relevant and more interesting to you. So, thatís not just selling you different things, but also, selling me the same things in different ways. So, letís say I'm interested in charities. So, if I view a charity website, if I'm a very conscientious person, so interested in sort of numbers and statistics, then the charity could give me more information about what the charity is doing. If I'm a more emotional person then the charity could tell me stories and show me pictures. So, the same thing is being sold, but in different ways.
Chris - What's really new about this though, David, because is this not pretty much what credit card companies have been doing by watching what I buy, where I buy it and when I buy it for decades to build up a profile, knowing more about my purchasing habits than I do?
David - Well, I mean, what's different about predicting personality as opposed to just correlating people who buy this also by that. What's different about personality is as I say, you can sell the same thing in different ways. You can also get an understanding of what is this person that I'm selling to. So I mean for me, itís more of a throwback to the past. In the past, you used to go to a shop and the shop keeper would treat you like an individual person and would sell to you as an individual, knowing your psychology. When the internet came along, that kind of went away. In the name of efficiency, we gave everyone the same experience. But now, we can go back to kind of the possibility of treating people like individuals rather than just numbers.
Chris - Lastly, how does this go down with consumers? Do people think this is a good idea or people concerned?
David - Well, I personally find it quite creepy, that itís possible to make these predictions. I think fromÖ
Chris - You invented it.
David - Itís not just my predictions, but even the other predictions that are being made. I think the reason I find them creepy is that I donít understand why I'm seeing the things I do online. So, Facebook is showing me adverts, Google is showing adverts, lots of companies are showing me things, and I donít understand what data they're using behind the scenes to make those predictions. That doesnít have to be the case. A company could say, they show me an advert and you could say, ďYou're seeing this because you're an extroverted person or because you're well-organised."