Oral evidence: Social media data and real time analytics, HC 245Wednesday 18 June 2014

Oral evidence, 18 June 2014.

Witnesses

Professor John Preston, Professor of Education, University of East London, Professor Mick Yates, Visiting Professor, Consumer Data Research Centre, University of Leeds, and Dr Ella McPherson, Research Fellow, University of Cambridge, gave evidence.

Chair

Good morning, and thank you very much for coming. I think that all three of you were listening to at least part of the earlier exchanges, so you have the gist of our direction of travel. First, I ask you to introduce yourselves for the record.

Professor Yates

I am a professor at Leeds University, and I work on the ESRC-funded consumer data centre. In a previous life I was at Dunnhumby, which you may or may not know is the company behind the Tesco club card.

Dr McPherson

I am Ella McPherson. I am an ESRC fellow at the University of Cambridge in the department of sociology.

Professor Preston

I am John Preston. I am professor of education at the University of East London. My area of specialism is disaster education and public response in disasters.

Chair

We are talking this morning about social media, and one of the areas that we are particularly interested in is how Governments and governance could be improved by smart use of social media. Is it possible to see some gains?

Professor Yates

First, there is a distinction that was not made by the first group between social media and big data, in the sense that social media, generally speaking, is unstructured—the photograph that you put on Facebook, the colours of your shirts and so forth—whereas most big data historically is structured datasets, such as till records, census records and so on. You have an opportunity to take the structured and the unstructured. In a commercial sense, if you knew what colour shirts you liked, you could make a better promotional offer, based on your Facebook pictures. In the social science sense, you can do the same thing: attitudes. What are people’s attitudes to different kinds of food? Where do they go? How do they visit and what methods of transportation do they use? That will come out of social media, but it would not necessarily come out of structured datasets. If you put those two things together, you have a rich way of thinking about consumer behaviour, which goes well beyond business.

Dr McPherson

Our perspective on this is that social media data is a complement to other sources of data that the Government are already using. We think that it should be used in addition to, not as a replacement for, other methods of inquiry such as interviews and surveys.

Professor Preston

In terms of my research, which is on how social media could be used in a disaster or emergency, I would agree that it is not a panacea for Government, but you have to be careful in the ways you use it. It does not work, for example, in warning and informing the public of a crisis, where old media such as radio and television, are much more effective. It does not work in terms of helping people to organise themselves in terms of a large-scale event like the evacuation of a city; there we found that social media could disrupt the evacuation and cause congestion. Where it does work is in terms of helping policy makers in Government understand how a situation unfolds, to look at how populations are responding during a disaster or emergency and during the recovery, to put resources where they are best used. You have to be careful how you use social media in an emergency situation, but it is of use.

Chair

Leaving aside disasters, those are interesting comments on how other situations unfold, evolve and develop. Are there examples, outside the ones that Professor Preston gave, where social media could be a useful tool in planning Government responses?

Dr McPherson

One area where social media has proven to be very useful is in understanding what is happening in closed societies or countries under conflict. You were talking before about issues to do with data manipulation. That is always a problem, but in scenarios where previously it had been difficult to get information out of situations, social media is proving to be a way to access it and to get some idea of what is going on in those scenarios.

Professor Yates

You could also use social media to understand how ideas flow through systems; people swap ideas, change ideas, improve ideas. As a way of gauging innovation in a system, you can look at what is happening and what people are saying. In theory, you could also use it to stimulate innovation if you were part of the process of communication itself.

Chair

I ask this question as somebody who is over 65, who has a Twitter account and Facebook account and who is “LinkedIn.” Are there problems interpreting the data about some demographic groups within the population?

Professor Preston

Certain demographics use social media a lot less, and even if people count themselves as social media users, we distinguish between heavy users of social media and those who might have put down only one or two tweets in their lives. It is not only an age thing; it is also socio-economic status and ethnicity in terms of the users of Twitter or other social media, so you have to be careful when you interpret the results. That is why it does not work very effectively as a method—

Chair

Surely that is true of any data source.

Professor Preston

Not necessarily. If you look at the cohort studies data, for example, that the Economic and Social Research Council fund, it is designed to be representative of the population. The millennium cohort study, for example, is designed in such a way as to be representative, whereas Twitter is based on users who want to tweet, or not.

Dr McPherson

That means that it is important to understand who the user base is, and to have research into that. It is always important, as I said, to combine it with other sources of data.

Professor Yates

It is blindingly obvious, but people use social media deliberately to share, and they choose what to share. That tends to be a younger person’s thing, but it also applies to the 65s. That is the difference with some of the more structured datasets, and it is where social media is completely different. People have taken the decision to share information about themselves, positively, with the rest of the world.

Stephen Mosley

I am interested in whether the UK Government can use social media data and real-time analysis to react to threats and to things that are happening within the UK. Do you think that the Government are properly equipped to use social media information and real-time analysis to react to emergency events?

Professor Preston

In our analysis, we looked at three cities in terms of local government: London, Birmingham and Carlisle. They each had different strategies in their use and interpretation of social media. Birmingham was very much ahead of the curve in using Twitter, Facebook and YouTube; London, being the seat of Government, was quite top-down; and Carlisle said that they did not want to use social media that much—it was very much face-to-face contact and radio communications that they wanted to use. It depends on the city’s orientation as much as anything. Birmingham has changed recently. The resilience team has changed, so it depends upon officials who have an interest in doing this for it to be a success. Policy makers find that they lack the tools, or the confidence to use the tools, to do this in real time; for example, during an emergency. There are tools that enable you to analyse sentiment—how the population is feeling emotionally in a crisis—and that can be incredibly useful. Are people fearful? Are they anxious and what are they anxious about? In real time, policy makers would not really want to use that tool because they would be scared of the implications; they would be scared of making the wrong decision on the basis of social media data.

Chair

Before we move on, may I test that a little further? Have you undertaken any research around communities close to COMAH sites that are used to emergency planning processes?

Professor Preston

In a more recent project that I am doing on population response to infrastructure failure, populations close to COMAH sites are used to using social media and looking at social media in terms of what is going on, whereas populations that are not near COMAH sites are not as used to it.

Chair

They are more aware of the likely effect.

Professor Preston

They are more aware of the effect and of the alerting protocols.

Dr McPherson

May I add something about its use in crisis or emergency situations? I am concerned about using social media to establish events—to establish what has gone on. The point that we made in our written submission was that using data sources requires verification. You need to do a technical verification of social media information that relies on particular technologies, and that requires knowledge and understanding of how social media data can be manipulated. It takes a human element, because at the end of the day establishing the truth is a subjective exercise. Moving from social media data telling you what is going on in an event to reacting in real time I don’t think is possible, because you need to put verification in the middle and that takes time.

Professor Preston

There are two verification methods for social media in real time. One of them is to use the users themselves. In the Australian floods, people were using the hashtag “mythbusters.” “Mythbusters” is a television programme about busting myths, conspiracies and so on, and people were using the hashtag to say what was real and what was not. That became a trending hashtag, with people using it to ask whether the flood was really happening in an area. People actually made use of that hashtag. Another one is the use of real-world data, using social media in conjunction with CCTV data or other kinds of situation awareness that the emergency services might have. But I agree that it takes a while to establish what is true and what is not. One of the things that we fear in emergencies is that people, or even terrorists or subversive bodies, might set up slop-bucket accounts to tweet that something is happening when it is not occurring, in order to make the situation worse.

Stephen Mosley

That was the area I wanted to move on to next. Is there any evidence, in the situation that you just highlighted, of third parties trying to encourage other people to do things?

Professor Preston

One of the biggest examples is the Syrian Electronic Army, as they classify themselves. No one knows who they really are, but they tweeted that the White House had been destroyed and no one knew where Barack Obama was, which caused billions to be wiped off the stock market in the United States. People are using it for disruptive activities at the moment.

Dr McPherson

There is an absolutely huge amount of evidence emerging about this in Syria, but in other cases as well. I was looking recently at a video of someone being shot by a water cannon. It was being circulated in Latin America, with the title “This is happening in Colombia,” but there was the same video in Mexico. The problem that this raises is that of verification. Even if they are ultimately disproven, it requires a lot of resources for groups to debunk them. It takes time, and it takes people away from other work that they could be doing, so it is quite disruptive.

Stephen Mosley

Looking at the UK, have there been any cases of that? I am on social media quite a bit, and see these stories circulating. I guess there is no checking, so how do you debunk these myths?

Professor Preston

Twitter can be self-correcting. In our project, we had an analysis of a plane crash in Cork in Ireland. Social media got there first, in terms of how many casualties there were, before other media. It tends to be self-correcting so, first, people were tweeting that there were 20 casualties, then someone said that there were no casualties, but both of those were debunked. They iterated to the right—unusually, before the old media got there and the BBC broadcast how many casualties there were. People act almost like scientists sometimes. It is not just rumours; people say, “Is this correct or not? I know someone who works at the airport.”

Dr McPherson

There was some great research done by The Guardian in conjunction with the LSE about the rumours circulating during the London riots. They showed timelines of tweets supporting the rumour and then debunking it, and how that worked, so it is self-correcting. You asked how you might verify. This is an emerging area, and I have been looking at it in terms of human rights organisations as well as journalists, and guidelines have been in development in recent months. The fundamental approach that verification professionals take to this information is to assume that it is false and your job is to prove that it is true, and if you can’t prove that it’s true you do not use it. I remember some quote that it is less about the snazzy technology than it is about good old gumshoe detective work. You might try first to find the source—the person who posted it—and speak to them directly and ask them details about it, just as you would with any other bit of information. It then needs cross-referencing with other databases. Technology can help; for example, you can use satellite images to verify whether large-scale bombing has taken place. You can also look at weather databases and ask the person, “What is the sun doing right now where you are?” and cross-reference to see whether they are actually there, but it is a complicated process.

Professor Preston

It is also important to remember that old media can be hacked. The emergency alert system in the United States has been hacked several times in the last few years, broadcasting erroneous messages at state level. It is not just new media that can be subverted.

Stephen Mosley

Lastly, do you think that social media platforms should be obligated to share information with Governments when there are concerns about security issues?

Professor Preston

In disasters and emergencies, they are willing to do so readily anyway. Most users have an idea that they are sharing data, and they would like it to be used for altruistic purposes. If there is a disaster or an emergency, people are willing to do so. With the Boston marathon bombing, for example, people were willing to share their data and images with the authorities. People are less willing to do so if they think that the security services are spying on them for other reasons.

Stephen Mosley

Or if social media is being used to encourage that trouble.

Professor Preston

Yes, absolutely.

Jim Dowd

I want to look briefly at how real-time analysis can be used in non-emergency circumstances. First, I want to test what you said, Professor Preston, about Twitter being self-correcting. Surely it is not self-correcting at all. It undermines the authenticity and the believability of Twitter the more conflicting messages you get. If everybody is tweeting—unless they are doing it maliciously, as Stephen suggested—“I believe this to be true,” what are people going to do? I think that they will lose confidence in the whole system.

Professor Preston

People do not behave like that. They look for corroboration. If you saw a tweet saying that there was a fire at King’s Cross station, people would say, “Well you’re saying this, but what evidence is there”? And then they would look for someone who had tweeted—

Jim Dowd

In other words, they do not trust it.

Professor Preston

Not necessarily initially; there might be a bit of overshooting, but eventually it comes to be corrected over a period of time. That happens more rapidly as social media progresses.

Professor Yates

In effect, you could argue that they do trust it, because they are looking for sources of corroboration from the very same system, and they assume that the system will provide the correct answer, so generally speaking—

Jim Dowd

Are you suggesting that that is a false assumption?

Professor Yates

That is the way that people actually use it.

Dr McPherson

My evidence has shown that it is not at the system level that trust is taking place; it is looking at every individual user and deciding whether or not to trust that user. You look at the information but you also look at the source, and you do not trust the information unless you trust the source.

Professor Preston

As a small example of that, we did an exercise with the Olympic Delivery Authority when the Westfield shopping centre opened. We were looking at different kinds of data. Someone was tweeting that one of the gates was shut and you could not get in that way, and people were saying, “Maybe someone is trying to get in more quickly,” but once they had tweeted a photograph, there was some corroboration and people started to believe that the gate was actually shut.

Jim Dowd

I accept that; over time, if the same message is repeated, it becomes the norm.

Professor Preston

No information is correct at first shot. If we heard that there was an emergency in London or elsewhere, the broadcast media would probably get it wrong.

Jim Dowd

A lot of social media users are the least likely to believe the Government or any public authority.

Professor Preston

Maybe, yes.

Jim Dowd

Can we go back to non-emergency uses of real-time analysis? Are there any circumstances where it will be of benefit, other than in emergencies?

Professor Yates

It is not a real-time analysis of social media, but the US amber alert system on missing children is quite a good system. You can get real-time alerts sent to people’s phones through tweets or whatever, in whatever way people have signed up to the system, and they provide information to the highway patrol. It is a small slice of use, but it is a clever use and it is well accepted in the US. It is a way of communicating as well. It is not just, “I’ve got a disaster. Let me understand the disaster,” but what you would like to communicate to people. To what extent are the Government prepared to open their communication channels to talk to people? I go back to the question of trust. I am not sure that we know each other, so I am not sure that I trust you, but maybe after an hour I might, because you have been saying sensible things that corroborate—

Jim Dowd

But I am a Member of Parliament.

Professor Yates

Sincere apologies. My general point is that we as human beings trust each other over time by understanding what is being said and the context for what is being said. In fact, social media provides you with lots of different personalities. On Twitter, for example, I am reasonably well regarded—believe it or not—for photography, but I am probably not well regarded on politics. People know the context of my personality in that social media, and that is how people use it. Government could do the same thing. What is the personality that you want to project, in certain circumstances and situations and in certain media, to communicate with the public at large? It is a positive opportunity rather than a negative one, if that makes sense.

Professor Preston

Another example, away from disasters, is looking at flu or the spread of infections. You can use social media to work out where outbreaks are and when there will be a demand for antibiotics and so on, and use that to predict where the demand will be on GPs and pharmaceutical services.

Jim Dowd

Can it be used to replace or supplement the traditional census?

Professor Yates

You can supplement it but not replace it, for all the reasons that everybody has given so far. The census will provide a broad look at everything, in a structured and detailed way, and you cannot replace that with social media, as to a certain extent it self-selects. Social media can create user attitudes—even colours, if you like; different things that are not available in traditional census mechanisms—and it can be extremely valuable. It is one of the things that we will eventually do in the centre at Leeds.

Jim Dowd

You are clear that it could not replace the census.

Professor Yates

It could not replace it, no.

Jim Dowd

The civil service has traditionally not been the most responsive or adaptive of institutions—for good reason, I hasten to add; I do not say this as a criticism. How amendable do you think it is? Do you have any evidence that they will be able to utilise this either at a national or local level?

Professor Yates

Can I address that by stepping back? The biggest barrier to the use of big data and social media analytics in general is the understanding and management of what it is capable of. We have a skills shortage, and I would like to come back to that later, but we also have an understanding shortage, even in industry, about how you can use analytics and data to make different kinds of decisions. In that sense, the civil service is no different from industry at large. The system needs an education job at principal level, and on bringing in the right people to do that. It is an “experiment and see” mentality, perhaps, but it is the same problem that, frankly, you will find in most big companies.

Pamela Nash

We had evidence from the University of Cambridge touching on some of the themes that you discussed with Mr Dowd, saying that real-time analysis is incompatible with verification. That is common sense, and we did not need to ask the University of Cambridge for that. Does that rule out its use by the Government for civil policy purposes? How do we get around that?

Dr McPherson

It does not rule it out at all, but what needs to be in place is a rigorous verification system. There is a new publication called “Verification Handbook,” which speaks to a lot of experts in this area about the techniques that they use. First and foremost, it says that whatever verification system you are going to use should not be put in place when you have an emergency and want to use the information. It should be systematically thought out ahead of time. It is something on which participants can be trained. It has ethical aspects as well, which should be included in that training. I don’t think it rules it out; it just means that verification needs to occur, and that verification should be based on a plan.

Pamela Nash

Can you tell us more about what you mean by that? These are situations where I imagine that it would not be easy to plan ahead unless, as in another inquiry we have done, you have good horizon scanners where you could use Twitter or other social media to analyse an unfolding situation. How do you plan ahead to use that information?

Dr McPherson

There is a methodological level of planning, which is understanding how verification works, and it can be done with social media; but there is also how this sort of data would fit into existing Government bodies and their existing sources of data. You can imagine that the police have all kinds of sources of information, including on-the-ground contacts and networks. You would then get social media data, and try to verify it. You would plug it into your existing on-the-ground networks, with whom you can have quick telephone or face-to-face contact. These will be people that you trust in that community, and you can ask them to verify it for you. It is a system of triangulation that, at its basis, should be used for any source of information, but I want to highlight that it needs to occur in this situation because social media is particularly vulnerable to certain types of manipulation.

Professor Yates

There is also a distinction between individual verification and mass verification. With mass verification, you have the same techniques as used with other data sources. You are looking at a vast quantity of information, and trying to figure out the truth from it. That is a bit different from identifying whether we said a certain thing ourselves at a certain time or a certain moment. From a broader sense, social media can be extremely useful because it is very fast. You can look at the way that some companies use data. I am not sure if you are familiar with how Google do translation, but they favour quantity of data over accuracy, and that is how they get better results. They have literally digitised every document that has been translated and figured out the context. If I hold up a candle and say, “This is light,” does it mean that it is illuminating or that it is not heavy? It will depend on the context in which I hold up the candle. The software deals with that by crunching vast amounts of data to get to a very fast response, which is why we all get instant translation on our phones. The same kind of techniques can be applied to social media.

Professor Preston

There are other uses outside verification. Whether the data is verifiable or not is one issue, but in an emergency situation people go through different stages. We found that at the first stage they use social media to seek information. They then go through an emotional stage and then an opinion-sharing stage. If you know that as a policy maker, you can tailor your types of message to those different stages. Which stage are the public in at the moment? If they are seeking and sharing information, you want to put information out there, but if they are responding emotionally then you want to put something out about being concerned or being compassionate about the victims of the emergency. If they are sharing opinions, you want to deal with that side of things. Another facet, away from verification, is that some people on our project came up with a tool to detect outliers on social media during an emergency. If people are saying things very different from anyone else, it might be interesting for you to pick up on it, even if it is not verifiable. They might have access to information that others do not, or they might be an outlying group with an interest in the crisis.

Dr McPherson

What you said reminds me that whether or not it is true does not have a bearing on whether it has an effect. It does have a bearing, but it may not have a bearing on whether or not it has an effect because it depends on what other people who are looking on social media believe. Even if you eventually debunk some kind of rumour, if it is circulating and people are believing it and reacting to it, that is another area where the Government might be concerned about responding.

Pamela Nash

In terms of the processes that you have described, Professor, and the nuances that you mentioned about the different stages of people using Twitter, are the UK Government doing that already? If not, how do we get ready to make full use of it?

Professor Preston

In my conversations with civil servants, they are certainly interested in dealing with it, and emergency managers are interested too.

Pamela Nash

That was a very political answer.

Professor Preston

Yes. They are interested in doing it, but they do not feel that a tool is available that Government could necessarily trust. In terms of sentiment analysis or analysis of emotions during any sort of political situation, they would like a tool that is kitemarked—one that Government could trust—but it is not there at the moment, so that makes them wary of doing it, whereas at the local level, you see people experimenting with all sorts of different things in local authorities and local government.

Pamela Nash

Is there any incentive for anyone to produce that tool?

Professor Preston

There is a commercial imperative for someone to produce a verifiable tool that Government could use and trust. In the States, the Federal Emergency Management Agency is very big on engaging the private sector in producing these sorts of things, with Government contracts and so on.

Mr Heath

Professor Preston, you mentioned observations in a flu epidemic, which reminded me that when I was a Minister at DEFRA we were dealing with Chalara fraxinea, ash dieback, and one of the things that we had then was an app and an ashtag hashtag, which we were rather proud of.

Chair

How much did you pay for that?

Mr Heath

I wonder whether there is greater scope for observational science, using citizens as observers, in collecting information about natural phenomena such as epidemics. If so, won’t we need better protocols for interpreting the information that comes in? Obviously it is partial and some of it is incorrect, but it nevertheless has some value.

Professor Preston

There is great scope for doing that. In my example about Birmingham and how Birmingham Resilience organised things, they depended on people at a local level to say whether events were happening and to tweet and use hashtags and so on. There is huge scope for that. Although the data may have observational errors, most people are minded enough, in terms of helping the scientific effort, not to skew the data. I cannot think of any example of a scientific effort that has involved Twitter, like spotting bird species and so on, where people have deliberately skewed the data.

Chair

Does that conform to the Google rule about collecting volume? Every now and then, the pinpoint data might be inaccurate but the general pattern is correct. For example, David’s constituency was badly affected by the recent floods, and an individual Twitter message might have been inaccurate, in the same way as an individual physical measurement might not have told the full picture, but the collective pattern does.

Professor Preston

That is right. You can get sampling errors.

Mr Heath

I want to talk about the collection of personal information, and the disparity between the preparedness of the individual to give personal information to companies, whether they are social media sites or private or industrial, and their lack of preparedness to do the same for the Government. This may all be down to George Orwell, or there may be something else.

Professor Yates

It goes back to what we said before. In social media, people make a conscious choice to share specific information for a specific purpose—some of it silly, maybe, but some of it for a useful purpose. They have already decided to give that, and they know that it is going to be used in some way, shape or form, by other people on the system or by the company itself. That is a bit different from being compelled by law to give information about your tax return, your expenses or whatever. There is a different sense for the human being about how you deal with that information. In one sense you are doing it freely, but in another sense you are being compelled. If you can get people to do these things freely, that would be incredibly helpful. There would then be no sense that they were being imposed upon.

Mr Heath

Quite so. We freely give information to our doctor about our symptoms and our personal details, but there was marked reluctance by some for that information to be shared for purely benign research purposes in the NHS database.

Professor Yates

I completely agree, but that is a failure of education. It goes back to the point of to what degree we are educating the population at large about the value of using data to take better decisions on behalf of all of us, in many different ways. We do not do that very much. Even in my business school, we teach analytics almost as a statistical activity, whereas we should be teaching it as a contextual activity about how to use the insight to get to a different place. Although it might be hard to convince the population that it is something they should truly understand, we have to start that process. Had we had that education in place, I contend that getting informed consent would have been a lot easier.

Mr Heath

That is helpful. Thank you.

Professor Preston

There are platforms available where people can share information for disasters. There is one called Ushahidi, where people are very willing to share their information, but it depends on the level of government you are talking about. People might be wary of sharing information with central Government, but if it is about helping in an emergency at city or local authority level, people might be very willing to share information.

Mr Heath

We talked with the previous panel—I think you were all listening—about the application of our own data protection laws, but also the potential extension of the EU legislation. Are we going to inhibit our Government’s ability to use data effectively by perhaps increasing the protections available under data protection laws?

Professor Yates

That is a difficult one. I have to state up front that I am more in favour of the US style of using data to innovate rather than some of the rights-based approaches, although obviously there is a balance between the two. I was at a conference recently where a gentleman called Steven Finlay, in a new book, talked about rules on how to deal with personal data. He said that it depends on the data itself. How immutable is the data? Your DNA cannot be changed and it should be completely protected. To what extent does it benefit you as an individual or society at large? To what degree should we protect it? How will the data impact me as an individual? If it is going to impact me a great deal, I really should be protected. There must be some sort of balancing act with the needs of the individual. Just to paint it as, “I must protect the individual’s information” is too broad; to what extent does the individual really need it to be protected, and can we protect it while also allowing companies and Governments to use the information in a sensible way for the benefit of people at large? I am not sure that any legislation deals with that.

Mr Heath

That suggests that you think that personal information is in common ownership rather than the ownership of the person to whom it refers.

Professor Yates

No, I tend to think of it as personal. I am sorry if I suggested otherwise. Your DNA is very much your own information. Can it be used by doctors to help you? Of course it can, so to what degree does it belong to the doctor or you? I am not really sure, frankly.

Professor Preston

Doing research is a legal and ethical minefield, and data protection legislation is obviously a central part of that. Anything that could make innovation or academic research easier in this area would be welcome.

Mr Heath

It would probably do the reverse.

Professor Preston

That is right, yes.

Professor Yates

I have an observation on that which might be helpful. I was thinking in preparation for this session that, in the legal world, there is the Creative Commons approach to you releasing your IP on the web, with certain restrictions on it. The idea of the kitemark was suggested, but I wonder whether, from an individual point of view, we could have some way of saying, “I am prepared to share this,” with some kind of licence, “but I am not prepared to share that,” with another kind of licence, so that the individual could positively take a decision about which data they want to share and in what context. That would be very helpful.

Chair

We have talked about licensing regimes. Taking the simple example of the club card, the trouble is that if you want the points you have to sign up to the terms and conditions; if you want to be on Twitter, you have to sign up to the terms and conditions. Companies, under current legislation in this country and elsewhere, have all the aces.

Professor Yates

They do, in a way. That is absolutely true.

Stephen Metcalfe

Following on from that point, I want to look at users’ expectations of how their data will be used. We talked earlier about end-user licence agreements. To show how helpful they were, Jim told us that there were 110 pages. Are they designed to be as accessible as they appear, so that you can hide all sorts of stuff in them? Is there a way of making it easier for people to know the key issues that they are signing up to—the bit that you would not expect to be in there, as opposed to what you would expect?

Professor Yates

I think that there is. It is almost like what are the bullet points of the agreements. I completely agree. If you take the example of a loyalty programme, whatever the terms and conditions and the detail, people are making a conscious exchange. They are allowing the company to use their information on what they have bought or the planes that they have flown on in return for something—some kind of discount, some kind of service improvement or whatever. That is why people sign up for those programmes. People understand the idea of exchanging their data for a benefit. If those simple things could be explained in those 110-page agreements—“What is the benefit of Twitter sharing my information?”, “Why would Facebook share my information?”—it would be helpful.

Professor Preston

Socially, people treat social media a bit like they treat the pub. They feel that if they go into a pub and have a private conversation, it does not belong to the pub; it is their conversation. They interpret Twitter or Facebook in the same way—as a place to have a conversation. People need to know what they are signing up to, but because they are US companies, the advice would be exhaustive and legalistic, and their lawyers will have advised them that the terms and conditions should cover any situation that could potentially occur.

Dr McPherson

Even if you know and understand the terms and conditions, there is a difference between the theory of what you are signing up to and the practice of how you use it. In my own Twitter use, for example, I think of my followers when I tweet; I am not thinking that it could be used to draw conclusions about other things, or that it might end up travelling the world or in a newspaper. I read an interesting piece recently that advocated considering Twitter users as what in journalistic ethics is referred to as “inexperienced sources,” who would not necessarily know what the personal implications are for them of their information going broadly public, and emphasising caution.

Professor Yates

Another side to this is that when you log on to a website and you sign in with Facebook or Twitter, not everybody realises that you are allowing the other site that is using that login to access your Facebook or Twitter information. You get a thing that says “sharing your profile,” but that is through the API, so you can access profile information from Facebook or Twitter without paying Facebook or Twitter, just by saying, “I sign up for this.” It is not a barrier to understanding that individual. Usually, there isn’t informed consent. If you sign up and share your Facebook profile and photographs or whatever, they then become part of the other person’s system. In the mobile context, that is called a token. That token is long lasting. It is there for a long time; it does not just disappear.

Stephen Metcalfe

As is often the case, I suppose it is getting people to catch up with the speed at which the technology is developing and to get social attitudes to change. As politicians, we would want to put everything that we ever write, say or tweet in the context of how we would feel if it was on the front page of the local or national newspaper. If you do that, you protect yourself. With the pub analogy, if you are having a private conversation but someone is standing next to you and you are aware that they might be listening, you probably would not think that the pub is recording it for security purposes. We have to help people understand how the world around them is changing, but do we need to do that on their behalf, from a Government point of view, to protect them—to create systems that avoid them making these mistakes, or do we need to educate people about the freedoms and choices that they have now?

Professor Yates

Education, education, education. To me, that is the most important thing. You talked quite a bit about data scientists in the first session. Unfortunately, data science in most countries tends to get defined as a technical activity. As it was originally defined by IBM, it was a combination of analytics and artistry, so that you understand the business or social context of what you are doing. There is a great data science course in America at the Illinois Institute of Technology. We have one at the University of Bedfordshire, but even my institution does not have a course where we train all these people across the board on how to use analytics and insight to take public decisions. Education has to be the answer rather than legislation.

Jim Dowd

Following on directly from that, you said earlier, Professor Yates, that you wanted to come back to skills and training in these matters. I asked the first group about this. It is a new science or discipline—whether it is science or guesswork, I am not quite sure. How are we adapting to it? How is the provision of courses for professionals in this area?

Professor Yates

The data capability strategy that was mentioned before is a good document. It is robust and it points things in the right direction, although largely from a technical point of view; it does not necessarily address the social, business or contextual understanding that we also need. If you look at the data, you get the insight, but the visualisation and impact of the data is more important than the data itself. How do we train people in that? We do not have enough programmes that do that. We have some in the DTCs—the doctoral training centres; there are 21 in the UK. The White Rose in Yorkshire is doing a big data programme to address some of these issues. We need more of those programmes. That one is aimed at doctoral students, not at students in grammar school, high school or comprehensive school, so we probably need that kind of programme too. More education at different levels needs to be addressed. Don’t get me wrong: this country is probably one of the leaders in some of this analytics work, largely because of what companies like Dunnhumby have done in the retail space. We know how to do the work; the shortage tends to be finding people who can put it into context, not for doing the analytics itself. That is where we need to focus.

Professor Preston

In terms of skills, it is a very interdisciplinary area. The benefit of doing our project on social media in emergencies was that it brought people together. I am from education, and there were economists, physicists, mathematicians, computer scientists and people involved in language processing. It is partly a reorganisation issue—how we get people from different disciplines to work on this complex interdisciplinary problem. It is not necessarily a matter of new resources; it is a matter of how you organise the resources that you already have. As my colleagues said, we already have great potential in the United Kingdom in terms of social and physical sciences.

Jim Dowd

One of the broader problems that the Government have in highly technical and advanced areas—in the field of IT, for example—is that they cannot compete in attracting skilled labour into their service because it is so commercially valuable. Is this not likely to replicate itself in the case of data analysis and big data? It is of such enormous value out there that the Government could not possibly compete on labour rates and so on.

Professor Preston

There is a possibility of strategic partnerships, working with universities and other innovators and with private sector innovators to make these kinds of products, or whatever you want to do. It is not necessarily a matter of the Government funding it all themselves. I don’t think that works. If you look at the Department of Homeland Security and FEMA, their approach is very much to have science centres in the universities to innovate in this area. That is one approach.

Jim Dowd

I am not sure about the record of the Department of Homeland Security in building their new headquarters. There is not much faith in what they are up to.

Professor Yates

The UK has a great reputation on open data; it is one of the world’s leaders, so the Government have a good reputation in this field. It is not a bad reputation; it is a question of building on it with the kind of partnerships that John is talking about. That would be helpful.

Chair

Professor Preston, on your observation about the multidisciplinary nature of the skills needed, do the research councils get it now? Is there a joined-up approach within the research councils to help fund projects that fit into this category?

Professor Preston

Yes. Our project was jointly funded by the Engineering and Physical Sciences Research Council and the Economic and Social Research Council. It brought together those two bits of expertise in one project. Those sorts of activities help. The research councils are getting it in terms of big data.

Professor Yates

The project that we had was funded by the ESRC. We would not have been awarded a bid unless we had put a multidisciplinary bid together inside our own institution. It was almost forcing us to think in an interdisciplinary way, which was a good thing.

Stephen Mosley

Talking about skills in the UK, do we have the infrastructure in place to take advantage of the market?

Professor Preston

We are trying an approach of concentrating on a few universities or providers to deliver those skills. Whether that approach will work in practice, I am not sure.

Stephen Mosley

I didn’t mean the skills. Do we have the physical infrastructure—the data centres and the like?

Professor Preston

I could not answer that, sorry.

Professor Yates

Generally speaking, we are pretty good on that, although I would like higher-speed internet in my village.

Stephen Mosley

Get your postcode on the record while you’re at it.

Professor Yates

Sorry. That was a flippant comment. To be serious, generally speaking, we are pretty good at building decent infrastructure, and some of the initiatives that the Government have taken in this space have been quite good. Commercial interests demand that you do not just have British infrastructure, you have global infrastructure; systems like Hadoop and so on are cloud-based, so you do not need the physical infrastructure in a particular jurisdiction or country.

Stephen Mosley

With the increase in regulation from the UK or Europe, do you think that this sector could be a flash in the pan that will evaporate when regulations choke the volume of data that is being collected?

Professor Yates

No, very simply. It is an unstoppable train, given the amount of data that we are collecting on everything. Soon we will have an internet of things, where our fridge will talk to our car and remind us to go to the supermarket. It is an unstoppable train.

Professor Preston

Companies like Google operate almost above the regulatory environment in Europe. They will say that if they cannot do it in Europe, they will take academics to the US and work with them there.

Jim Dowd

What if Google runs the world? What will happen then?

Professor Yates

Perhaps they have already taken over the world.

Stephen Mosley

Do you see the threat from regulation to be for home-grown UK and European-based companies rather than global companies? Would you draw that distinction?

Professor Preston

It could threaten what UK and European companies are doing if we are not careful, whereas the US has a looser regulatory environment for companies in this area.

Professor Yates

That is generally true; it is a looser environment that is more focused on driving innovation than on protecting the rights of the individual. That is an obvious distinction. At the same time, we should be quite pleased with what the UK does in this field. We are good at analytics and social science, we have some really good universities and we are quite good at being interdisciplinary. As I said, a lot of these technologies are cloud-based, so they do not need physically to be in the next city. They are being used in different ways, and the more that we can encourage innovation to use those systems, the better. We are going to be fine.

Chair

Do you agree with that, Dr McPherson? At least one Cambridge academic who speaks a lot on this sector expresses great concern that we will one day produce a less benign Government, and that all our data will be misused by the state. But you have academic freedom here.

Dr McPherson

Personally, I am concerned with the ethical aspects of using this data, with a lot of topics that were brought up by the previous panel about the inability to be truly anonymous and the inability to future-proof. I know that there are a lot of cases where we are talking about using the data for innovation and so on, and for very positive reasons, but I am thinking of the more vulnerable people—I am thinking of my human rights work—those for whom exposing their identity has tremendous repercussions. That is the area that concerns me.

Chair

You accept Professor Yates’s comment that it is an unstoppable train.

Dr McPherson

Yes, for sure.

Professor Yates

By the way, I agree with the ethical concerns. Don’t get me wrong; I was just addressing innovation issues, not the ethical concerns.

Dr McPherson

I changed the subject.

Chair

We’ll leave it there; it is a fascinating point to stop at. Thank you very much for your attendance this morning. Oral evidence: Social media data and real time analytics, HC 245 33