Personality & Big Data: Impacting mental healthcare more than it does advertising?
Running into dr David Stillwell while discussing product personality research for a large car brand at The Psychometrics Center of The University of Cambridge, I took the opportunity to propose an interview about Personality & Big Data, David’s field of research. Having set up an improvised stage in a deserted computer lab, dr Stillwell shared his findings and views about the impact Personality & Big Data research can have on:
- helping better understand depression
- helping big companies create long term trust with their customers
- helping us find the house we really want
- help people match to products
Time-sequenced questions and transcript of the interview below.
- 00M15 What findings in your research didn’t and did you expect?
- 00M57 What findings in the field of personality and big data will be applied shortly?
- 01M30 Could you give us an example of how individuals might benefit from personality and big data?
- 02M41 What constitutes the paradox between choosing and using product?
- 02M09 Where can we see big data produce major improvements in the field of psychological research?
- 03M08 What should we take into account when analysing psychological findings based on big data?
- 04M05 What value remains for conventional testing when the accuracy of for example FB-likes generated predictions are so high?
- 05M00 Would that mean that we are not the same person in the offline as in the online?
- 05M42 Increasingly inferring personality from social data, how can that play out in the construction a new big data-driven framework for human behavior?
- 06M58 Being employed by a business school in Cambridge, how can a business school benefit from your type of research?
- 08M24 Can you elaborate on the shift in data collection from transactional sources to social ones?
- 09M40 How can people themselves benefit from the personality information that is available about them?
- 10M49 What measures do businesses take to increase trust with customers whose information they use?
- 11M54 Why haven’t businesses implemented these measures?
- 12M28 How can Personality and Big Data play out in the process of home buying?
My name is David Stillwell, I am a lecturer in Big Data and Quantitative Social Science.
00M15 What findings in your research didn’t and did you expect?
I was surprised how accurate it was. We then did a follow-up study and we found that it was even more accurate than someone’s friends, even someone’s family, predicting a person’s personality. On the other hand I wasn’t that surprised because if you do talk to someone for an hour and you tell them, you know, your favourite things, what you like to do as a hobby, where you like to go and eat, things like that, then actually you are giving away quite a lot of information. So Facebook makes it quite fun to share your Facebook likes and to share things with your friends but some of their information is quite personal.
00M57 What findings in the field of personality and big data will be applied shortly?
So the big thing from my research is that now you can predict people’s psychometric results just by existing data. So it’s very easy for companies, and for individuals themselves, to just share data, press a button and to have predictions made about them and then companies can apply that in areas where they were never able to use psychology before.
01M30 Could you give us an example of how individuals might benefit from personality and big data?
So, previously there was no way to get someone to fill in a personality questionnaire just to get, you know, personalized adverts. But now that can be done automatically in milliseconds. So probably advertising is where this would be use first just because, you know, Google, that’s where I, you know, Facebook, they all get their money from advertising and it’s such big business. But hopefully it will be used in more kind of personal way as well in future. So perhaps the next thing will be dating and then in future anywhere where you have got a person and product and you are trying to match them together.
02M09 Where can we see big data produce major improvements in the field of psychological research?
The interesting thing for research is to get information that previously we had to ask people over and over again, you know. For example, how happy are you. You know, depression. People aren’t depressed all the time, it sort of goes, comes and goes in spikes and so actually in order to understand depression either you have to ask people every ten minutes, you know, are you depressed, are you feeling depressed, are you feeling depressed or if we can automatically infer that kind of information then in future we might get better understanding of what’s happening on a minute-to-minute basis. Is it, you know, you have met this friend and therefore the depression has gone away and you had a fun conversation. You know, what is it that’s actually going on in people’s lives, minute-to-minute that affects their psychology.
03M08 What should we take into account when analysing psychological findings based on big data?
So the disadvantage with big data in general is that it tends to be correlational. So, we watch what millions of people do, we then find correlation between behaviour and psychology and then we assume that that’s telling us something useful. I think the best way to use big data is as, sort of, in combination with more traditional approaches that is still used in psychology.
So experimental approaches where you are putting people into groups, and one group is control group and one group got some experimental effect and then you will see what effect that has on people. So then you have got the big data is telling you, well, here is what millions of people are doing day-to-day and then the experiment is telling you that this is the cause and result, this actually means something psychologically, this affects what people do.
04M05 What value remains for conventional testing when the accuracy of for example FB-likes generated predictions are so high?
That’s a really good question. So one interesting finding we have from people who’ve, get their personality predicted from our, their Facebook data is they then say, well this doesn’t match my personality and the answer to that is often that, okay, that may not be how you act in real life but this is the way you are presenting yourself online to your friends. So that has meaning in that online context but that doesn’t necessarily match with what you are doing in your real life.
So, the questionnaire approach could then allow you to compare the two, so what am I like online, what am I like…among my friends and then you can see if there are differences there. Maybe you are acting in different ways in different situations which you wouldn’t be able to understand necessarily if you were just using the online big data approach.
05M00 Would that mean that we are not the same person in the offline as in the online?
Well with a majority of people, they probably act in the same way in both situations, but it’s still interesting to know that this algorithm is saying that you are, you know, an extraverted person and maybe you think that you are an introverted person, because decisions are going to be made on that basis in future, so advertising will be shown to you, you know, dating website will be making predictions about you based on just your online data, so knowing what your online persona is and could give you insight into why are these decisions being made for me.
05M42 Increasingly inferring personality from social data, how can that play out in the construction a new big data-driven framework for human behavior?
Most personality questionnaires at the moment are theory-driven. So a very expert and a very prestigious researcher sits down and says I think these are the personality traits that are important for understanding behaviour and then they come up with, you know, the Jungian types or whatever system they come up with. But it is theory-driven and someone looking at the personality literature and saying, you know, I think that this is what’s important.
What the big data could tell us is, if we can analyze it the right way, maybe there are some small but very important things that affect people’s behaviour that we have just never thought of before. So, one example: people know that love is quite important in their lives, but love is not measured in personality questionnaires. So maybe that kind of information could come from the big data to let us understand that you know it is quite a specific concept, but it’s actually very useful for predicting what people do in the real world.
06M58 Being employed by a business school in Cambridge, how can a business school benefit from your type of research?
The business school here knows that companies they are working with have more and more data about customers, about staff, about policy in general, so, you know, large groups of citizens and they also know that they are teaching business leaders, they are teaching government employees how to make decisions that affects all of us.
So then what those people need now is an understanding of, you know. Can they make some of these decisions based on the big data? How can they leverage the existing big data that is there to help them make better decisions? What kind of analysis could they be doing that they haven’t even dreamed of being able to do from the data? Such that then they can make business decisions driven by data rather than their hunches or, you know. Even the best expert is going to have a limited understanding of the full range of customers that might exist. I mean we are all limited by our experiences, but if you have got big data on, you know, all of the customers that you have got then algorithms can give you that kind of overview that an individual person finds very difficult.
08M24 Can you elaborate on the shift in data collection from transactional sources to social ones?
So I think both forms of data are useful. The interesting thing about social data is what people want to tell each other. and I think that’s why it’s so great at predicting personality, because it’s a bit like a bumper sticker. You are trying to make a statement about yourself and that’s why you are trying to share your personality with other people and you are saying this is important to me. Transactional data is what people are actually doing; so what they are searching for, what they are buying.
So it can tell you kind of more almost secretive things that people don’t necessarily tell each other so often or also more sort of boring things that they don’t bother to tell each other, because why would I tell my friend that I just bought washing up liquid. So, I mean, I think both forms of data are useful and both forms of data will continue to be useful. It depends on what type of questions you are asking, are you asking questions about what individuals think about themselves or ask, are you asking questions what are individuals actually doing and can that tell us more about themselves than maybe they even realize.
09M40 How can people themselves benefit from the personality information that is available about them?
So one thing that we are encouraging companies to do, and also individuals to do, is to have an understanding of what predictions are being made about them based on the data that exists. So that’s why we have got the Applied Magic Sauce website where you can go to the sign log in and see what predictions we have already made about you based on your Facebook data.
But we are also encouraging companies to, when they are personalizing your search result or when they are showing you advertising, it would be really good if those companies were actually saying, you know, we think you are this type of person and that’s why you are seeing this advertising or you have done this in the past and that’s why you’ll see this, these search results. So more transparency and more understanding of what predictions are being made can, number one, make individuals feel more comfortable of what’s happening and number two, also give them more understanding of what’s happening.
10M49What measures do businesses take to increase trust with customers whose information they use?
So the trouble is right now is that businesses kind of lurch from crisis to crisis, because, something comes out in the media or Facebook you are doing this with my data and then that becomes a media story and Facebook trying to explain it, and then something else comes out, you know, next week and Facebook has to explain it again.
So their poor PR people are, you know, being worked over time. But it doesn’t have to be that way if they are more, kind of, sharing in the first place then I think people would realize in the vast majority of cases, although big data can be used in a very advanced way, often it’s used in a very simple way. It’s just, I have searched for holiday so show me even more holidays you know, I bought shoes, so show me even more shoes even though I probably don’t need anymore.
So if companies were showing this information then I think the companies would benefit in the long term because users would trust them more to be doing the things that they are doing with the data.
11M54 Why haven’t businesses implemented these measures?
I think probably the reason they haven’t is companies may be concerned that by telling you why or how they are targeting then their advertisers may see that as a commercial secret, you know I’m targeting people between 25 and 29 and they are men and they are doing this and this, and they might be trying to hide that but I don’t think that’s the way to go forward in the long term.
12M28 How can Personality and Big Data play out in the process of home buying?
Sure, so I mean, to me it makes a lot of sense that someone who buys a house in the first place is trying to buy a house that matches their personality. So obviously practical considerations do come in, I would like to buy a mansion but I can’t but given the options that you have then you make a choice that this is the best house for me and that’s probably driven by psychology.
So given then that I’m selling a house in future, then the house that I’m selling probably I have, number one, bought a house that matches me and number two, I have done some effort to customize the house to match my personality. So the house that I’m selling is probably a house that someone like me would be more interested in buying. So it is, a bit like dating, you know, people find other people like them, you also spend a long time in your house so people. You know it would make sense that they would be searching for houses like them as well.
Interview Peter van der Bel
founder & curator of The Centre for Applied Product Personality Research