David we are going to explore one of the most important facts about the present and future. Largescale Data Collection has and will continue to change politics in the developing and especially the developed world. Increasingly, private firms are assembling profiles of individuals. For example, every eligible voter in the u. S. With the death of information depth of information that would have made j. Edgar hoover weep with envy. The use of socalled data and the inferences made from it have a singular purpose. To craft and deliver messaging that will shape the future behavior of an individual, from buying a particular brand of toothpaste to discouraging a citizen from voting. From choosing one ridesharing service, to voting on decisions like the u. K. Brexit. And, the u. S. President ial election. These are assembled from what our guest rightfully calls are our Digital Footprints. Many of us to not even realize we are leaving these footprints behind. What we listen to one streaming services. What we watch on cable tv. Our guest tonight, dr. Michal kosinski, will introduce us to his work. Particularly, facebook likes and profile pictures. They will help us to begin to grasp how sussmans are being and could be used to shape our political and social reality. Join me in welcoming Michal Kosinski to the stage. [applause] particularly, facebook likes and profile pictures. They will help us to begin to grasp how sussmans are being and could be used to shape our political and social reality. Join me in welcoming Michal Kosinski to the stage. [applause] hi, david. Good evening, everyone. David thanks for helping us pure behind this curtain. Maybe we can begin by having you describe your work for us. What do you believe can be learned from our facebook likes and profile pictures. Again for having me here. I am a competition scientist. I am working mostly with data, in particular big data. Withad of spending Time Research subjects in my lap or running small experiments or maybe learning about people using surveys, i would look at the Digital Footprints that you so nicely introduced before that we are all leaving behind while using Digital Products and services. A is a great time to be computational scientist, a great time to be made. At the moment, because we all enormous amount of Digital Footprints. We are an estimated megabytes per day per person. If you wanted to put it on paper, filling it out, 0s and stick it out, to one day worth of data, the stack of paper would be like from here to the sun four times over. One, you wont exactly. Generating, we are all generating enormous amounts of information. Now, this information of course our trail of our behavior. Our thoughts, feelings, social interactions. Medications, purchases. Even the things we never intended to say, like not sure if you guys realize this, if you e a message on facebook, and then you decide, it is 2 00 a. M. And i should not send it, and you abandon the message, guess what. It is still being saved and analyzed. It is not just this one platform. In most cases, data is preserved even if you think you have deleted it. Take thisal is to data and learn something new human psychology or human behavior. One of the byproducts of doing this is i will produce models that take your jewel footprints and we will try to predict your future behavior. Maybe your psychological traits, politicalrsonality, views, and so on. What was shocking to me when i started working in this field, is how accurate those models are. Shockingne thought thing. They are also very difficult to interpret. I know a computer can predict your future behavior. Reveal or determine your psychological traits from your digital footprint. It is very difficult for a human scientist to understand how exactly the computer is doing it, which brings me to this black box problem. Which basically means, it might be human psychologist, scientists, would be replaced one day by ai. You basicallye, have those models we dont really actually understand very well how they do it. They are amazing at predicting your future behavior, your psychological traits, and so on. I worked with facebook likes quite a lot, not because facebook likes are the best type of digital foot print we are leaving behind. Facebook likes are not so revealing. Why . Because liking happens in a public space. When you likes of it on facebook, you probably realize now your friends will see what you have liked. You wouldnt like anything really embarrassing or maybe something really boring, something you want to hide from your friends. But now, when you use your web browser or you search for something on google, or you go and buy something, you have much less choice. You will search for things he would never like on facebook. You would visit websites you would never like on facebook. You wouldbuy stuff never like on facebook, like you would buy medicine that is very revealing of your health to read most of us do not like the medicine we are taking on facebook. If someone can get access to your data, your search data, records from your mobile phone, this digital footprint will be way more revealing than whatever i can do using facebook likes. Whatever findings i am coming up with, they are just conservative estimates of what can be done more revealing data. You can see the entire industry, the entire industries, not just one industry they are moving towards building their Business Models on top of the data we are producing. My favorite example is a credit card. How many of you guys have actually paid for the credit card recently . We have a few people that maybe didnt do their Research Online are so fancy, they pay for their credit card. Most of us, including me, we do not pay for credit cards. If you are not paying for something, and thick about it a credit card is an amazing, magical thing that allows you to pay for stuff without carrying cash. There is a complicated network behind it. Computers crunching data and someone. We are not and so on. We are not paying for it. Why . We are the products of the record company. You can go to the website of visa or mastercard, or any other credit card operator, and you will see they see themselves not as a financial company. He started as a financial company, helping to channel payments. Aw they see themselves as Consumer Insights company. By observing the things you are buying and when you are buying them. On theh you are spending individual level. They can learn a lot about you, but they can also see extract information on a broader level. When they see recently people in San Francisco started buying certain things or going to a certain restaurant. Valuable very information that can be sold. If you are not paying for something, you most likely are a product. Think about your web browser that you probably didnt pay for. Your facebook accounts. Your web search mechanism. One of the gazillions of apps you have on your phone. Think about how much data you are sharing with the others. Facebooks your use of originally as a graduate student at cambridge, i believe. I believe facebook likes republic. Anybody could see them. That make that kind of data sets available to you since it was just public on facebook . Michal yes, you are just pointing out, another reason why im using facebook likes. I was lucky to have a huge data set of volunteers that donated their facebook likes to me as well as political views. Personality and other scores. Basically other parts of their facebook profiles. Friend, he2007, my started this online personality questionnaire. Take a standard personality test, and then he would receive feedback on your scores. It went viral. We had more than 6 Million People that took the test. Half of them generously gave us access to their facebook profiles. When you finish your test, it would ask you, if you would be willing in return for us offering you this interesting thing, if you would be willing to give us access to your facebook profile. Which we would later use for scientific research. More than 6 Million People took the test. Around 3 million profiles, facebook profiles. At the beginning, in fact, people like to say, when i graduated from high school, i already planned to run this research 20 years later. It was not the case. In my case, i kind of stumbled , kind of got into this research by accident. I was developing traditional personality questionnaires. They are composed of, i am always on time at work. I like poetry. I dont care about abstract ideas. I had this data set of facebook likes were basically people who dont like or i abstract ideas, i dont like to read. Why would we even asked people these questions if we can just go to their facebook profile, look at their likes, and fill in the questionnaire for them. I started running those Simple Machine learning models that would take your facebook likes and try to predict what would be her personality score. This worked really well, which actually was pretty disappointing for me. I spent so much time developing questionnaires. Now a computer can do the same thing in a fraction of a second. We had other data in our data set. Ok, we can predict personality. I wonder if we can predict political views. Whether your parents were divorced or not. Each time we asked the question, the computer with think and said, of course we can commit it is amazing. Predict this. It is amazing. Rerun the models with independent data, thinking i must be doing something wrong to read even a computer can look at your facebook likes and predict with very high accuracy, close to perfect, whether you are gay or not. People dont like anything obviously gay on facebook. They do but it is a small fraction. It was based on the movies they watched, books they read. It looks very counterintuitive to me at the time you could do it. Spents i got older, i more time running the models, it is pretty obvious. Let me illustrate it for you guys, maybe let me try to offer you a short introduction to how those models work. It is actually pretty intuitive. If i told you there is an anonymous person and they like hello kitty. It is a brand, i am told. [laughter] you would probably be able to figure out, if you know what hello kitty is, that this person is most likely female, young. An iphone user can read you could probably go from there and about her. Inferences you would be very correct. 99 of people who like hello kitty are women. You dont need computer scientists to make inferences of this kind. Most of your likes, purchases on amazon, locations that you visited with your phone most of the search queries you put in google are not so strongly really laying about your intimate traits. It is not mean they are not revealing at all. They are revealing. To lady that you listen gaga 30 times yesterday. It is not only weird, it shows you something about your musical taste. It also probably it for sure little extent,ny your Sexual Orientation, political reviews. Virtually any other psychological trait we would like to predict. It is just that the amount of information there is really tiny. For a human being, this is useless. With a computer algorithm, it can get this tiny bit of information and aggregated over dozens of Digital Footprints you thousandsehind of Digital Footprints you are anving behind to arrive at accurate prediction of what your profile is. This is the paper i published in 2013. About the excited promises of this technology. Im still excited about the promises. It is used to improve our lives in many ways and we dont realize how. If you dont believe me, think about netflix or spotify. Facebook newsfeed. People spend two hours a day on average looking at it. Because the ai made an accurate prediction about what your character is. It adjusted the message to make it engaging. There are also downsides, as i am sure we will be talking about. It was basically the piper i published in 2013. It got quite some press coverage. Like,f the coverage was this is so cool. We can predict whether somebody is a republican from their facebook likes. Nice, shiny gadget can read said, no, you have to realize, there are tremendous consequences for the future of our society. So cool. Is just this is as far as we go. Interestingly, this is how the general public treated the results. Policymakers and companies took notice. Two weeks after the results were published, facebook changed privacy policies in such a way facebook likes were no longer public. 2013, likes republic for everyone to see. I didnt even have to be your friend on facebook to see everything you liked. Showed paper, our work by seeing what you like, i can also determine your Sexual Orientation. All those other intimate traits. I think it was a great thing facebook took notice and to preserve your privacy, switched that off. You also had u. S. Governments and eu governments that took notice. They started working on changing the legislation to protect their sum of theom shortcomings of this phenomena. David lets talk about some of the political uses of this work. I want to hear about how private firms are using Big Data Analytics akin to your work to shift Voting Results in one way or another by microtargeting messaging that is defined by its intended persuasion rather than accuracy. Is cambridgefirms analytical, part of a network of firms that were involved in both the u. K. Brexit vote and trunk campaign. Much of what we know about this story is due to recent investigative journalism. Especially by the guardian newspaper. I thought i could provide the audience with a quick review of the story. Cambridge analytical is a u. S. Firm. Robert mercer. Until august, 2016, had steve bannon as the vice president. It is one of the most successful Quantitative Hedge Fund managers. A major owner of breitbart news. Executiveon left his positions when he became manager of the trunk campaign. Of course, he is now the chief strategist to president trump. They employ data mining as well as government records to develop a dossier on every u. S. Voter, which was first used by the Ted Cruz Campaign and later the Trump Campaign to mark a target microtarget their message campaign. Has been a central consultant for this kind of thing with the various u. K. Organizations that pushed for the brexit vote. Appears tonalytica be the owner. Time magazine reported yesterday congressional investigators are looking at cambridge analytical in the context of their exploration of russian activities. Included russian elements. Short, quite a tangled web. You tell us about how your work relates to this whole thing . How we should think about the claims about the effectiveness of their work for the trunk campaign . Michal those are very good questions. There was a lot there. First of all, we dont really they were. Fective interestingly, when you listen, saying howd by amazingly efficient they were. When they realized governments were getting interested, some things they have done became not entirely legal, they suddenly changed theirs feel. Now they say, it did not work at all. We are just making stuff up. Which obviously means they are lying now or they were lying then. Sure, can tell you for first of all, we have a lot of that we produce and acted amy showing such approaches work really well. Only theee, it is not Trump Campaign, but all the politicians employing messages like this in their campaigns. Barack obama was the first politician to do it on a massive scale. Outrage,emember any especially on the left side of the spectrum. Hillary clinton as well. She not only spent three times more money than donald trump doing personalized targeting on social media, also hired away smarter people in my opinion. Yes, she lost, but she didnt usingecause trump was some kind of magical methods. Was caused by Something Else. Can this winsk me the election, the answer is, yes and no. Life when youf are running a political campaign. Like tv spots and writing articles, putting ads in the papers. Everyone is using it, it is not giving anyone and unfair advantage. The only advantage here is barack obama was the first one to use it on a massive scale so it must have given him an unfair advantage. We as humans, we like to focus on the negative. It is great we focus on the negative. This is clearly a great psychological trait because it allows us to be successful as a species. But, lets put aside focusing on the negative and risky. Politiciansbout being able to personalize their message. Their outcomes people seem not to notice. You one on one, that is what i can do. Mean use algorithms to help to talk to you oneonone about things most relevant to you. They ran those of the rhythms to try to understand your character, your interest, your dreams. Make the message more interesting and relevant. All, has oneof outcome. Messages became more important. Say, yes we i could can. Everymoney showing it on screen and be successful. And i couldnt do much else. On a to settle down message that was a common denominator. Not particularly aimed at anyone. Now, i can talk to you about things that are relevant to you. Which meansthings that are releo them. The Political Program became more important. It is not only that because this in turn has mattered for important outcomes. If i make a message more relevant to you, you become more engaged in politics. That is great for democracy when we have the voters more engaged in the messaging they are getting and we have more people engaged in politics and it is great. I believe. Second of all, it also makes politicians think about, ok, what is important for you david . In the past i could say, yes we can. Now i must think what is important to you and and thinking about it would perhaps make me update my belief about what political my political agenda should be. And specifically when it comes to minorities. Message onoadcast my tv come i focus on the majority interest. That is the logic of broadcasting. If i speak with people oneonone, i now can develop interest naturally as a politician into what they have to say and what they are interested in. That is the first change, more engagement and more importance of the message. Slogan. Tional the second change, we saw in the recent election, politicians like donald trump but also bernie sanders, bringing in people into the political process that traditionally were not so interest