 Well, thank you very much for having me here today and thank you all for coming. I know it's sunny outside So if that had happened in England, I can guarantee you no one would be here For me, it's really and now I must apply to psychologist and by training So the reason why I'm interested in big data and the open-source movement is really because it can tell us a lot about How people behave in their everyday lives and how that relates to something like more stable psychological traits that make people Uniquely themselves and so you can already tell that my talk is going to be a little different To the talks that you're probably going to hear over the course of the day But I still hope that you enjoy it and that you learn something new about how data can relate to something like psychology Okay, so now I want to start and with the news story that I'm pretty sure that many of you have come across in one way or another And so that was a new story that was originally published in the Swiss news outlet That's magazine and basically in a nutshell what the authors claim is that the Trump presidential campaign had used detailed psychological profiles of millions and millions of voters and in the US to target them with highly personalized Advertising and on Facebook and the goal was not only to kind of convince and Trump supporters to go out and vote for Trump But also to kind of discourage and Hillary voters to go out on election day and put in their vote And so I mean it's it's kind of questionable whether that story is true simply because we don't have any data So it's very not open source and in that sense And but it clearly put digital psychometrics and in the spotlight and it's kind of now known around the world And that's why I want to kind of talk about what is digital psychometrics? How does it work and what could be the benefits but also like the challenges that we experience by now being able to predict Someone's kind of highly intimate psychological characteristics from their online behavior Okay, so psychometrics really means that we're trying to assess someone's psychological traits something like their personality They're acute their life satisfaction And the way that we did that traditionally was that we use questionnaire measures So we would give people a questionnaire that says I talked to a lot of different people that parties and now tell us How much you agree with that statement and What we were really getting at is basically people self-perception So they told us about how they saw themselves and you can I mean it's quite obvious that that's a bit problematic for various reasons And I just want to kind of briefly talk about you the first one is that well people lie And so especially if there's a lot of stake at stake when you kind of apply to a job I doubt that many people would say that they strongly agree to the statement I make a mess of things so it's very very easy to kind of misrepresent on these questionnaires And the second problem is that it's very difficult to scale So kind of completing one of these tests takes between five minutes up to an hour So we can like as research is we can easily invite people to the laboratory have them complete the questionnaire Maybe do that for like 20 50 people But if you want to do that for millions and millions of people then it gets a bit problematic Now watch digital psychometrics and this is something that's only been around for yeah, maybe like the last five years So what we're trying to do is to kind of get away from giving people Questionnaires and asking them about their self-perception But we're trying to look at what people actually do out there in the real world So kind of which websites do they visit? What do they do in their social media profile? And what do they spend what kind of stuff do they buy with their credit cards? Or even just what is their phone recording in terms of geolocation and physical activity and We know from research and that we can actually successfully predict someone's psychological trades from something like personal websites from tweets Facebook likes status updates and profile pictures and In a way, it's not very surprising because everything that you do out there Basically tells us a lot about your personal preferences your habits what you do in that your everyday life And how that then translates into something like more stable psychological traits And I just want to give you an example So that's apply magic sauce is an API and that we developed at the psychometric center And it's based on the models that we developed name basically We came up with a name because people asked us so how do you how do you make these predictions? And we're like got a bit sick of it So we're like, oh you just take data you apply a bit of magic sauce and then you have the Psychological profiles, so that's how we came to that name and I just want to show you how it works So how can we get how it can you use it because it's it's free online everybody can use it So how can you get from something like Facebook likes to like a psychological profile? so basically what happens if you go and apply magic sauce calm and You can click on predict my profile, so that's open to everybody. It's absolutely free of charge And then you can donate different types of data, so you can donate your Facebook profile You can donate your Twitter data if you click here and connect with Facebook, and this is my profile now I've all I'm already logged in so I didn't have to put in my login details And so you can see it's connected to Facebook if you click on make a prediction Basically in a few seconds it scrapes your Facebook profile takes your likes takes your status updates and turns it into Psychological profile so you can do that from all the sources that you have or the individual ones and that are available in your Kind of profile, so you can see it predicts stuff like gender, so it's pretty close and for me and that one And age I mean psychological gender somewhere in the middle big five personality traits So this one in that case my Facebook suggests that I'm kind of liberal or autistic I'm a bit more introverted and a bit more laid back and the end-to-end the average So all of you can do that if you just click on apply magic sauce It gives you a description of what your profile means and in terms of your psychology also predicts intelligence a bit above average That's promising life satisfaction a bit lower Not so surprising working as an academic and it kind of basically predicts a whole bunch of other stuff So check it out and really that just takes a few seconds, so you don't have to sit down complete like Question is for half an hour, but really just log in with Facebook and then you get at least a rough idea of what your psychological profile looks like Okay, so the question is how does that work? So what's behind and our prediction? I am engine apply magic sauce, but also other prediction engines So there's IBM stats social that does and pretty much and similar prediction Predictions so it all started and with the my personality Facebook app So my personality was set up by one of my colleagues in Cambridge who didn't have anything to do during his summer in 2007 so he set up this app and that allowed people to take real psychometric tests So they could kind of complete a personality questionnaire IQ questionnaire life satisfaction And then they got immediate responses on there and they get immediate feedback on their responses So they completed a question and right there was feedback and on their personality or whatever they had completed Now when he did that you actually wanted to collect data of maybe a thousand people maybe some of his friends and But people loved the feedback so much that they shared it on Facebook and he ended up collecting a data set of seven million people Completely free of charge simply because people really like to learn about themselves and they like to know themselves And so what we have now is this database that has Psychometric test responses on something like personality values life satisfaction of around seven million Facebook users around the world Now what's amazing about this data set for people who are interested in data is that at the end of the like after people had received feedback They were asked whether they would voluntarily share their Facebook profile information with us and for research purposes And about 50% of people did so so for about three and a half million users We actually have their self perception in terms of psychological characteristics and we also have their Facebook profile data So we know what they liked on Facebook what they post on Facebook We have access to their profiles and pictures and basically their public profile And now this is them like this is the interesting part And I'm not going to go into into details here because I'm sure that pretty much all of you know more about that than I do but basically once you have this behavior on the one hand and you have someone psychological and Characteristics on the other hand you can start building models, right? So whether it's a simple linear aggression or you kind of bet gets more complex and into neural networks or something And basically you can kind of try and take someone's behavior And predict a psychological profile without them having to complete a questionnaire and Like if you want to know more about that the models and feel free to kind of come and see me afterwards But what I want to do is basically show you some of the relationships that we found in the data That really show and it's very obvious and how people of different personalities Different the way that they behave online See, okay, so this is and a word cloud and of the words and status updates So they're most strongly correlated with the personality trait of extroversion So extroverted people are those who kind of love to be social they love to hang out with others They're optimistic and enthusiastic and you can see that in the words, right? So extroverts and talk about parties amazing love weekend tonight And so they seem to be having and a pretty good time If you compare that to the introverted word cloud, which I think most of us will be Probably somewhere in the introverted and cloud here you can see that it's very different So introversion Associated with talking about computers enemy internet reading books And and this is like on a very basic level. This is just counting words, right? So you don't need any rocket science You don't need any complex machine learning algorithms to see that there's something in the way that people behave online That's related to their personality It gets even better So this one so high agreeableness is one of the personality traits that distinguishes between people who care a lot about their social Relationships or trusting or empathetic and kind of warm and welcoming So those people again, this is like it's kind of stereotypical, but it also this is what we find in the data So they talk about family wonderful amazing and beautiful and happy So they're just very grateful in general and now word of warning and the word cloud for the low agreeableness Which is basically people who are more Competitive and more critical looks very very different and this is nothing that we came up with This is just what you see in the data So those are the words that are most strongly current and I'm not going to go through all of them now The pleasure of that but you can see again that it's very obvious how different personalities relate to to online behavior You can do the same thing for likes so just as an example again introversion extroversion because that's Like introverts again talking about computers seems to be a big common theme here So final fantasy battle star Galactica and extroversion those are people who hang out and love to be around other people So they like to meet new people socializing parties. So just just are just Official Facebook pages. So it's not if one of your friends posts a cat video and you're like, yeah I like that. So those are just official pages. So in terms of like, I don't know BBC or CNN Okay, the latest development, which I think is extremely interesting is that you can do Predictions of someone's personality simply based on faces. So the profile pictures that we have available on the end on our in our My personality database and we're used to kind of make predictions from faces of someone's personality And this is basically just the AI Dreaming of faces of introverts and extroverts So those are the faces that the computer predicted to be most introverted and extroverted and again You can see very obvious differences that make a lot of sense. So the introverts have glasses Look a bit more shy and reserved with the extroverts Probably have their hair dyed because it's like much more blonde and then the picture of the introverted ones and also maybe wear contact lenses because they have like blue eyes where introverts have Brown eyes. So again, like whatever kind of digital footprint you take there's a pretty good chance that it's in some way related to something like psychology personality intelligence Okay, now the question is how accurate is it and I don't kind of I don't want to go into too much detail But the the the way that we usually assess the accuracy of a prediction model is we compare the computer-based prediction to the self-perception of the participant and when you kind of look across different personality traits different footprints the Accuracy is usually around a correlation of point five or point six Now the thing is that point five point six is that good is that bad? And it's really difficult to say so what my colleagues did and I think this is actually a very nice way of illustrating the power and Of computer algorithms is they compared it to the judgments of people in our environment So how good are our work colleagues our friends our family members our partners at judging our personality in Comparison to how good the computer is And what this so this is just a Examples for Facebook likes so the computer needs around 10 Facebook pages to be better than someone's work colleague It's about 65 pages to be better than someone's friend About 120 to be better than someone's family member and around 250 to be better than someone's spouse and Now the average number of likes at the time we conducted the study was 230 So the computer is basically better than everybody except kind of a little bit worse than our partner and spouse Which is for me? Basically insane because the computer only has to Facebook likes right and usually you spend a lot of time with your friends a lot of time With your family and your partner so they should technically have a very good idea And with who you are as a person yet the computer is pretty much has pretty much the same accuracy just by looking at your Facebook likes Okay, so now how can we use like these insights digital psychometrics and in action? So what can what are the applications and of being able to predict someone's psychological profile? And the main message is really that I think it helps us to kind of personalize services across the whole board So we can use it for targeted marketing and healthcare and education because the more we know and about the person About their personal needs and preferences the better we will be able and to tailor the services that we offer and to those preferences Okay, so I just want to go very briefly to the first two to give you an idea of how I see digital psychometrics working out and in practice So what we did or like the question that we were interested in it Can we use the knowledge of someone's psychological? Characteristics to increase the effectiveness of digital marketing So that's going back to the to the Trump story where they kind of try to target people based in their personality and Really kind of increase the persuasiveness of their messaging simply by taking these characteristics into account And we did an advertise what we didn't promote Trump and we promoted something that's quite far away An online beauty retailer in the UK and what we did is we simply came up and With a few extroverted messages and a few introverted messages. So you can see the extroverted messages There was always something going on people were kind of having a good time together and when they introverted ads It was like usually one person in a quiet setting Now what we did is we ran actual Facebook campaigns and over the course of seven days Where we targeted introverted extroverted users with these different types of messages So over the course of the seven days we attracted about six million impressions So that's the number of times the ad was shown in Facebook around 10,000 clicks and about 400 purchases on the beauty retailers Webside I just want to give you two metrics So basically what we saw is that when you match the message that you use to promote a product to people's Psychological characteristics. We see higher conversion rates so people purchase more often and they also generate more revenue For the company and this is really simple, right? So it's exactly the same product the only thing that we're changing is actually the way that we communicate and with people based on their Psychological profile now you can do that to kind of promote consumer products You could do the same thing in the context of political campaigns in the context of health care Okay, the second topic that I think is extremely interesting and extremely important these days and is health care and just one example is That we can now predict and whether someone is likely to To develop a depression simply by looking at their smartphone So if we take into account and their activity in terms of where they are how much physical activity there is and how much they sleep How much they interact with other people all stuff that's kind of passively collected on your phone while you carry it around Pretty much 24 hours and a day can tell us whether or not you're more likely than the average person to develop a depression And if you think about that the people who develop a Depression are usually not the ones who would go out and seek help that is actually really useful because we can actively help those people who are in a vulnerable situation and This was a new story that I think is extremely interesting and it very nicely for me kind of shows both the positives like potential but also the danger for abuse and which was like Facebook and was being accused of Doing research on the press team. So people like teens experiencing Anxiety or stress and then being basically being exploited for targeting now if you think about it This is actually an amazing chance. So if Facebook is doing research trying to identify people who are at risk Well, if that information was being used to kind of help those those teenagers who are probably not going to go and see Their parents because they're having a bad time Then on the one hand, that's an amazing opportunity that could actually be used to help those Teenages, but obviously it also has the whole other dark side. We're knowing it like having information On something like that could be abused to exploit weaknesses and people's characters But like for me, it's really when I read this news story I was like well obviously there is potential for abuse But don't kind of say that it's bad in general because it's a technology and if we can help those people the benefits that we could Would be like basically equalling out the potential dangers in my point of view Okay, so now look into the future of digital psychometrics and as all of you know, there's basically more and more data Collected on us every second. So I think the future is really kind of taking People's more stable psychological traits that we can get from something like Facebook profiles the websites that people visit But then combining it and with information about the context that the person is in So kind of really looking at the person at in real time. So what is that? Okay, we know that this person is extroverted But what is he doing right now is in an extroverted situation is it in a good mood? And what does that tell us about how we should communicate with that person? How may be? prone he is to kind of showing a certain type of behavior and the person that I think has very nicely expressed why this is and So important is one of my favorite comedians Louis CK and I'll just let him give you give you the spiel Just the constant nobody takes in life unless it comes through this Yes, like whenever I see a televised event That's a huge like the the Olympics opening ceremony or Times Square at at midnight on New Year's Eve Everybody they see sees of people all looking at it through their phone Like there's explosions and acrobats, but they're looking at it through a little three inch screen like I think if Jesus comes back and Starts telling everyone everything it'll just everybody's just going to be twittering and they won't Really, I am Christ and I have returned. Oh my god. Jesus is right in front of me Yes I think you very nicely summarizes the fact that most of our lives are now mediated by these digital devices So whether it's communicating with our friends or just capturing the moments that are most important to us Most of it is recorded by like a smartphone by wearable devices and this is only going to get more And I think it's on the one hand and this is actually also my my conclusion I mean on the one-handed it's it's extremely exciting that we know so much about people's lives now but also with that great power that we have comes a great responsibility and And I really mean that because I think for a lot of people it's Extremely scary because they don't know what's happening. They don't know how stuff works They don't know what's actually possible And I think it's true for all of us in the room But certainly also people in academia is that we get need to get a lot better openly communicating and transparently Communicating first of all what's possible and then also how the technologies that we develop are going to affect people's lives Because what they hear at the moment is stories like the news story about Trump or the news story about Facebook How those technologies are kind of ruining their lives making them prone to being manipulated and basically posing a huge threat to their lives Now I think what we need to do is basically show them the other side So show them that yes these technologies exist and we're working on them to kind of make them better But actually there's so much potential for kind of helping you make better decisions helping you kind of lead healthy and happy lives And we're not doing that behind the scenes But we actually openly communicating and to the public what's possible and what we're currently working on and I hope that we Can all do to do that together and I think obviously the open-source movement and is one of the movements where we can see that Like at least going in the right direction And this is is my vision for the future that it's kind of people like you and me working together academia working together with industry And making that change happen. Well, thank you very much