Oral evidence: Social media data and real time analytics, HC 245Monday 23 June 2014
Oral evidence, 23 June 2014.
Witnesses
Professor Liesbet van Zoonen, Professor of Communication and Media Studies, Loughborough University, Professor David Robertson, Head of School of Informatics, University of Edinburgh, representing the UK Computing Research Committee, Dr Mathieu d’Aquin, Research Fellow, the Open University, and Emma Carr, Acting Director, Big Brother Watch, gave evidence.
Chair
Good afternoon. Thank you for coming in this afternoon. All of you have been sitting here listening, so you understand the format. It would be helpful if you could introduce yourselves for the record.
Professor van Zoonen
I am Professor Liesbet van Zoonen. I am professor of media and communication at Loughborough University. I lead a big research project about public taboos and desires around identity management and the sharing of personal data. It is very much focused on what people want and do not want. Dr d’Aquin: I am Dr Mathieu d’Aquin. I am a researcher in data technologies at the knowledge media institute of the Open University. I am currently focusing a lot on the data infrastructure for the Milton Keynes future cities project.
Professor Robertson
I am Dave Robertson. I am head of the school of informatics at the University of Edinburgh, but I am really here representing the UK Computing Research Committee, which is an expert panel of the British Computer Society.
Emma Carr
I am Emma Carr, the acting director of Big Brother Watch, which is a civil liberties and privacy campaign group based in the UK.
Chair
Thank you very much. We touched on this directly and indirectly with the previous panel, but who in your opinion is most likely to misuse data—the Government or the private sector? Do the Government need to approach social media data usage differently from how a private company might?
Professor van Zoonen
From our research, we can say that the UK citizens’ approach to this is that they tend to believe that the Government are a bigger risk than private corporations, although that is changing very rapidly due to practices by Tesco, for instance, that are becoming more well known. If you look at what people think about this, they still think the Government are the biggest risk.
Chair
That is Government per se. Does it break down into different Departments having a greater degree of trust?
Professor van Zoonen
Yes. It also breaks down into local government having more trust than national Government. The further away people are from national Government, the bigger the distrust of national Government and the trust in local government is. Of course, those different levels of Government are not included very often in the discussions we have about big data. We tend to presume that it is a global, national discussion, rather than also a local one.
Chair
A while ago I saw some data that suggested that one of the Government Departments fairly high up the list in trust—relatively—was the Treasury, which seemed to me to be an absurdity, but there are some odd human reactions to Government Departments.
Professor van Zoonen
Yes. It also depends a bit on which people you ask and what their political views are. There are definitely groups of people who have a high degree of trust in the legal institutions, the police and the secret service, and groups of people who do not have that. In that sense, there is not an overall picture. Dr d’Aquin: Unlike Professor van Zoonen, I cannot talk about people’s perception, but our research shows—besides what we might not know about Government capacity—that the organisations that will have the best ability to misuse personal data are certainly private companies, especially large-scale private companies not located in the UK, including the big social media platforms in the US, as well as companies that are not necessarily directly associated with social media but that collect personal data as a side effect of their activities. Of course, Google comes to mind. Some of our research has shown that, looking at the web activities—the web use—of average users, the biggest amount of disclosure of personal data will go to systems such as Google Analytics. That also includes mobile phone providers, internet providers and all the companies that have the capacity to collect very large amounts of personal data, which could, if they wanted, be largely misused.
Professor Robertson
It is really difficult to identify where the threats will come, but there is a special responsibility for a Government in all of this. The one thing a Government can do that is really hard for others to do is set an example. Open data is a good example of that. Government can set an example by backing initiatives of that kind. That is the limit of it—you do that and see what follows and so on and so forth. To some extent, you can have regulation as well, but that kind of light-touch example setting has already been pretty effective in many ways. Maybe you could go further, but it has done a lot. The other way round is to apportion blame. I will leave that to people who are more closely involved, but I expect it is really hard. It is really brittle as well. It is very difficult to know at exactly what point you lose trust in some of these systems because they are so pervasive. It is difficult to know where the rubber hits the road with the system and where it will have an influence on people—or not, as the case may be.
Emma Carr
Depending on what kind of data you are talking about and what the use of those data is going to be, the polling seems to indicate that the public have a very similar outlook on the public and the private sector in terms of how safe their data are with them. However, generally, people hold the Government to a much higher standard when it comes to data protection. That may be because people know that private companies are usually using their data to make money—for advertising purposes. When we see any indication that the Government are making money out of our personal information, especially when it is being sold to, say, insurance companies—Care.data comes to mind—people seem to get extremely upset about data being used within the public sector generally.
Chair
I will come on to Care.data. Professor Robertson, the UKCRC said that the Care.data initiative was “out of step with public opinion.” Could you explain that?
Professor Robertson
The reason for choosing that example was that it is a good example of what I was mentioning before—how brittle these systems can be. We had no particular reason to believe that there was any notion of malice or impropriety in that particular initiative. As far as I am aware, the people running it were doing the best job they could possibly do, but it got slightly out of step. By out of step, I mean that the people whose data it was were not fully aware of what was going on. Because they were not fully aware in the right sort of way—in their culture or their understanding of what is going on in the scientific community; you can frame it how you like—it was badly rolled out. Because of that, there is a scientific consequence, which is that research that could have gone on quite legitimately and successfully, had that been done just a little bit differently, will probably be held up.
Chair
Are there particular examples you can draw our attention to in terms of that initiative?
Professor Robertson
Emma Carr is nodding and may be able to give better feedback than I can. The simplest thing is that people simply did not know that this was going on until it leaked out in the press. In Scotland we have not had this kind of impact because we did not take that step, but when we talk to people about this sort of thing quite often their view is, “Oh, aren’t researchers using my data for something useful that will help to make people better? Wouldn’t that be good?” That is very different from suddenly finding out that people are going to do something you were not told about that may be partly commercial and partly about harvesting data and is now going to become institutionalised. It just plays badly, regardless of what the motives are. That is the problem—it is the public perception of what you do. It makes the problem very difficult. It is the reason you have Caldicott guardians for medical data. You have to have people whose job it is to understand how to reconcile those dangers.
Chair
Isn’t that a rather difficult area? There are circumstances where exactly the same practice could be used for a commercial benefit. Let’s take the use of cameras as an example. There has just been an understandable hue and cry about terrible cases of abuse in a care setting. I heard one person advocating for the families of the victims say that there ought to be cameras in every care setting. That was for a perfectly good moral reason, but, equally, cameras can be used to understand customer behaviour and so on. You cannot generalise, can you?
Professor Robertson
That is the problem—you cannot generalise. This is moving very fast. Some form of data analysis is embedding itself pretty deeply in domains; health care is one, but you can choose pretty much any other domain you like. If you look at what industry predicts, a big proportion of the data available to it will be social data—data derived from the living population. If that is true, it will embed itself very deeply indeed. It is really hard to reconcile some of these issues without understanding domains very closely. Care.data could not have been predicted by a computer scientist working alone. I suspect it probably could not have been predicted by a sociologist or a social anthropologist—or even a medic—working alone. Somehow the mix has to work. You can see signs of these kinds of interdisciplinary groupings building up, but it is an avalanche. It is probably impossible for us to move so fast that we could make it risk free.
Chair
Do you want to add to that?
Emma Carr
Big Brother Watch did some polling at the beginning of the year. We found that 69% of people had not been informed of the right to opt out of Care.data. This is entirely the point. Whether we are talking about Government use of data or private sector use of data, it is all about consent. You cannot just presume informed consent because, as in the case of Care.data, a leaflet has been sent out. Especially when it comes to something like medical data, where if people do not have trust that their data will be kept secure it may have hugely detrimental effects both on their own health and on the public sector as a whole, you have to ensure that there is a higher standard than for just tweeting or Facebooking about your day. If you stop doing that as a result of not trusting Facebook or Twitter, it does not really have the same consequences.
Stephen Metcalfe
Talking about trust, where has this lack of trusting the Government with data of any sort come from? Where has it grown up from? It strikes me that inherently the Government are trusted less than almost anyone else.
Professor van Zoonen
There are a couple of reasons you could think of. One is the particular context of the UK and the reason why data are collected here, which is very much set in a security framework. That has historical reasons and has to do with the alliance with the United States. If you compare the purposes of data collection for the UK Government with the way in which that discourse is framed in continental Europe, for instance, you find that in continental Europe the whole discourse about data collection from citizens is about providing service to citizens—making sure that a Government can offer better services to its citizens. The discourse in the UK is a safety discourse. That is a helpful discourse for citizens in the immediate aftermath of a crisis, when people buy into that and think it is really necessary, but after half a year people think, “Right, that is not that good an idea.” The transaction is a little bit abstract. You give your data and get security, but the longer that takes, the more abstract it becomes. In the Indian situation, for instance, and in Europe, you give your data but you get a better service. That is a really different way of presenting the policy.
Stephen Metcalfe
So in the UK the citizen does not see the link between sharing data and better services.
Professor van Zoonen
No, not yet, because the whole discussion here is always about security. Look at the surveillance cameras, the Patriot Act and its aftermath and data protection—it is all in the context of protecting your citizens, not of offering better services. That is there, but not as the first thing. Dr d’Aquin: I agree with that view. There are quite a number of different elements. I am not exactly sure we can say that citizens do not trust the Government, but what private companies—especially the big private companies—give back in exchange is more direct, sensible and visible. There is the notion that if you want to collect information from users in a commercial system, or even in an academic system, you have to provide incentives—you have to have something in return. From a commercial perspective it is much easier to make that something in return visible to the user. There is another aspect. Our research experience suggests that a lot of the lack of trust comes from fear of interpretation. If data are collected by Government, users need to have an understanding of what is going to be done with them. While it is still very obscure in most cases why personal data are being collected all round and what your social media platform will do with them, the obscurity is much bigger when it comes to Government. As has been mentioned, there is no clear indication of what the exact use and benefit globally of this particular data will be. That is where the fear of interpretation comes from. In some of our research, we have seen that, if you put users in front of the personal data their employers collect, reactions are often very pragmatic—“Of course,” “That is fine,” “That is normal,” “That is part of operational services”—but on certain very specific aspects the user will want to have a right to respond to their personal data. For example, a user spending a lot of time on the vacancy pages of their own organisation will feel that they need to provide context to that particular information so that it can be interpreted properly. The lack of trust also comes from this—the inability of the user or the citizen to provide the context for interpreting the data they make available. It is a lack of control, really.
Chair
In these two answers there are some interesting contrasts, aren’t there? On the one hand, I suspect that just about everyone in this room has a piece of plastic provided by some banking organisation or another where we have given a lot of data. The trust vehicle there is that, if they go wrong, they commit to making good their mistake financially, broadly speaking. On the other hand, I suspect we all have a Government piece of paper—an Oyster card—in our pocket. The payback there is the convenience of not having those wretched queues at the ticket office, isn’t it? So there is a point where the public give in to some of their worries about the role of Government in exchange for a service—the point you made, Professor van Zoonen.
Professor van Zoonen
Going back to the Care.data discussion, if you look at the way that is brought to the public—at least, in the website information—it is all about helping the NHS improve its services. That is interesting, because there is no direct link to how that helps the individual patient; it is all about helping the institution. The other side of that—what you get for the transaction—is underplayed, so to speak. Again, that is an example of an institutional interest—a Government interest—taking precedence over the individual citizen or patient. It is a matter of framing your data-sharing policy, in a way.
Stephen Metcalfe
Do you want to comment on the original question of where this lack of trust comes from?
Professor Robertson
This is a personal view, but I do not believe that Government is necessarily trusted less in all respects than some companies. I do not think there is a tremendous amount of trust in some of the large companies dealing with data—it is just the way that it is articulated. You have to choose where your pedigree examples are from. In Edinburgh, there is probably not a lot of trust in the local authority building tram systems because it did not go tremendously well for us, even though we now have a tram system—which is good, sort of. It is that kind of thing. There is still a monolithic feel to Government IT, which does not square well with the picture that was painted for you earlier of very lightweight, streamlined services, relevant to the user and so on; it just does not feel that way. I feel for people in Government, because that is quite a difficult one to shift. You do not have the luxury of being able to say, “Oh well. We’ll start with a clean sheet of paper.” There is a whole lot of baggage to do with public service and Government that you would not have if you were a start-up company, so it is very difficult. On the other hand, that is the way the world appears to have gone. From apps and all the way through, people are much more accustomed to being able to chew things up in small bits and understand vicariously the notion of a service that is delivered over the internet. That was not true five years ago. I should think it is tough for Government; that is part of the reason.
Emma Carr
There has been far more in the media about data loss in the public sector than about data loss in the private sector. Part of that is Big Brother Watch’s fault, I am afraid, for putting that into the public domain. Quite often you hear about USB sticks with child protection information being left on buses or mobile phones going missing and the relevant security not being there to protect all of that information.
Stephen Metcalfe
Because that never happens in the private sector.
Emma Carr
Apart from one or two high-profile cyber-security attacks involving people’s credit card details or whatever, it is on a much smaller scale. Obviously it happens, but—
Chair
How do you know it is on a lower scale? You can’t possibly know that.
Emma Carr
How do I know what is on a smaller scale?
Chair
That loss of data in the private sector is on a lower scale.
Emma Carr
I think it is probably on less of a large scale than in the public sector.
Chair
How do you know that?
Emma Carr
That is entirely the point. We find out because it is usually to do with credit card details. People generally have to find out about that because there is a financial aspect. When it is in the public sector, we generally find out because we have put in a freedom of information request, which we cannot do with the private sector, unfortunately. Again, that is more to do with having a much higher standard for information relating to people’s security or health, which people care about much more. Coming back to Care.data, when you have the public sector asking for larger datasets and wanting to do more with them, it raises the question whether the safeguards are there to ensure that those data will be kept to a higher standard than people would necessarily expect from the private sector.
Stephen Metcalfe
So it is not necessarily the individual coming to the conclusion that they should not trust the Government when the data are passed over, but someone telling them they should be outraged that the Government want to use those data to provide a service. Do you think that is why the Government are now pulling away from sharing more data? Are they fearful of media and public reaction because people are being told through the television, the radio and the newspapers, “You need to be outraged”?
Emma Carr
I disagree that the Government are pulling away. I am on several advisory boards that are helping the Cabinet Office and NHS England to try to go forward with data-sharing practices. We can all agree that it would be beneficial to the public sector to have that extra level of data and to be able to analyse it to help public services—it is just about going about that in the right way. Care.data highlighted exactly the wrong way to go about it—by not telling people what you are going to do with the data, who will have access to them and how they will be shared.
Stephen Metcalfe
Do you not think that the damage has now been done—that now no one will ever want to share those data?
Emma Carr
Exactly. That is why I am on the Care.data advisory board. We are working extremely hard to ensure that those mistakes are put right and the public can have confidence, when it does go ahead, that they have been taken away. We need to ensure that at all levels of data sharing, whether it is in the NHS or at local authority level, people can see from A to B exactly what is going to happen with their data. Unfortunately, there have been too many occurrences where that just has not been demonstrated.
Stephen Metcalfe
On that one area of Care.data, were you proposing that people should positively sign up to allow their data to be used, rather than negatively pull out?
Emma Carr
We would always advocate an opt-in rather than an opt-out as best practice, but we understand that that is not always practicable. However, we suggest that you do not tell people by means of a leaflet that you intend significantly to introduce different means of sharing medical data.
Stephen Metcalfe
So somewhere in the middle there is a way to be found. Do you think that the Government Departments that are potentially in possession of large quantities of data and might have opportunities to share them are not skilled enough to know how to use them and how to prepare the public for sharing them?
Emma Carr
We have been involved in workshops in the Cabinet Office about using big data to solve fraud, error and debt. Interestingly, when we ask what data they want to have and currently cannot have access to and what they want to do with it, often they do not have the answer to that question and take quite a long time to respond. Often in our experience at Big Brother Watch we find that they know big data is available—it is talked about and is seen as something of a new, modern silver bullet to aid public services—but there is something of a question mark against what they then want to do with it, perhaps because they do not have the necessary knowledge inside that Government Department.
Stephen Metcalfe
Going back to the public’s view on how all of these data are stored or used, is the concern about their collection and storage the fact that they might end up on a memory stick on a bus or a stolen laptop, or is it not about where they are stored but how they are then used? Where does the main concern lie? Has any research been done into that?
Professor van Zoonen
It is also about how and where the data are stored. In general, if there is a central storage system, there is much greater suspicion about what is going on than if storage is decentralised. In Estonia, for instance, they have a national identity card that offers a lot of different commercial and Government services. The trick there is that all of the data that you hand in to those services are kept on different databases. They do not go into a central database. That makes quite a big difference to how much people trust such a system. The other thing that makes a difference is whether people know that their data are being collected or whether they are being remotely collected, either by remote biometrics or by commercial monitoring on the internet. Having a certain level of awareness and a decentralised storage system really makes a difference.
Stephen Metcalfe
So identity assurance providers who collect data and assure you that it will then be protected are not the answer, because that would be one central database.
Professor van Zoonen
There is a level of choice already. Lessons were learned, so people can choose to whom they want to give their data. That is an important difference already. They are aware that they are giving data and that presumably there is a purpose for giving those data as well that is not breached. Dr d’Aquin: It is interesting that you connect the two. The identity assurance framework explicitly states distribution of information as a core principle, for the reason that it helps to establish trust by ensuring that no one private organisation will have total control over the entire system. Care.data does exactly the contrary. As I understand it, the current proposal gives the entire provision of software services to support it to one private company. That is one private company that people may or may not trust, but the understanding of how that private company might gain benefit from having such total control over their data is certainly not helping with the trust issue. Going back to your original question, it is true that for a lot of the researchers and journalists who are looking at this aspect the core computing infrastructure—where the data are stored, how they are processed and how secure they are—is a very important issue. Based not on concrete research but simply on an anecdotal understanding of research in other areas, I would say that these technical aspects are far from the concerns of the majority of users. The one question is, “What are they going to do with it, and how can I understand what is going to be done with it in such a way that I can be reassured that the interpretation, understanding and use of the data will not go against my interest?”
Stephen Metcalfe
How would you ever arrive at being able to reassure someone that the data would be used in a way that they could even begin to understand? We have heard how complex some of the links are between different pieces of data. How can you persuade someone that what is going to be done will be done ethically, sensibly and safely on their behalf? Dr d’Aquin: It is a good idea to start by saying what is going to be done.
Stephen Metcalfe
But there are people out there who will instantly say, “Don’t believe them. Don’t believe them.” Dr d’Aquin: There are mechanisms that do not necessarily achieve that amount but allow a certain feedback mechanism to be included in the whole system. They include some of the elements that have been talked about in the mydata program, which give access not through a complex procedure of requesting it under data protection but by the simple click of a button. A single authentication into one system gives access to the full set of data somebody else might have access to, showing what kind of analytics can be achieved on those data. In our own research we have done that, if only on a small scale, for the relationship between employees and employer. Our research has shown that when employees are faced with the data that their employer collects about them, the majority of them react by saying, “That is reassuring. I knew data were collected and that some things can be done. Seeing them makes me confident that I can entrust my employer with this type of data.” That is a very simple mechanism that is reasonably obvious. Giving the data back to the citizen—making them accessible to the citizen and attaching to them information about organisations and individuals who might have access to every part of them—seems an obvious mechanism, in addition to distribution, of course, to try to improve trust in these sorts of systems.
Professor Robertson
The issue is really, “What is the evidence base?” If somebody says, “Why should I trust this?” you should have some evidence. The important thing I want to add is that it is not completely a black art. There are things that you can apply that are reasonably well understood. You can have various mechanisms that are just to do with the data artefact that protect it. You can have various protocols. There are all sorts of ways you can try to make sure that that is secure in the narrow, technical sense. There are things to do with the process by which you move data around. You can have processes that are inspectable, that you could accredit and so on. Those are quite often used in health care when you are interested in doing that sort of thing. The third point is the authority argument, which is to do with professionalism. Just because you are using social data, it does not mean that everyone has to use it. You can have communities of people who have a very high level of trust, which is what an awful lot of this sort of thing runs on at the moment. All of these things are pretty well known. In my personal opinion, they tend to screw up where they are done without respecting the domain. People do them thinking that one size fits all and they can do the same thing as they move between different areas. That is where you get all sorts of sociological and cultural problems.
Stephen Metcalfe
Thank you.
Stephen Mosley
While the focus of the last question was very much on trust, it started to move on to things like security and safety of the data as well. Is it possible for the public ever to be sure that their data are secure?
Professor Robertson
No. I suppose any of us could probably expand on that a little, but let me be the first to do so. I suppose the more detailed answer is “not on any scale that we would care about.” Pretty much as soon as the data start to contain significant content, it is very difficult to guard against that content being able to be used to breach various aspects of your privacy and security. That is it; that is the basic problem we are all fighting against.
Stephen Mosley
I will assume the answer is pretty much the same for all of you, but could I expand it a bit? Have the Government done enough to make sure that individuals’ data are secure?
Professor van Zoonen
What we find is that people are very concerned about things like identity fraud and privacy breaches. They do not have a real clue about how that works, but it is a big concern. In addition, saying that their data will never be secure is a recipe for quite a lot of distrust. The issue is not making the data more secure but making sure that, once your data have been stolen or mistakes have been made with them, there are mechanisms to correct that. You bring your car to the garage for a yearly check-up to see whether it is still all right. You need to have something like that for your personal data as well, so that they are checked on a yearly basis to see whether they are still all right everywhere and have not been stolen. We hardly ever talk about that kind of mechanism and whether it is a way to repair identity loss, identity fraud and privacy breaches.
Chair
That would have to be predicated on the citizen having the right of access to data.
Professor van Zoonen
Exactly.
Chair
And on having some sort of ombudsman role that presumes the citizen is right, rather than waits for—
Professor van Zoonen
Yes, it would be the other way round—trusting the citizen to come up with the right personal data, instead of distrusting—
Chair
I have been arguing that for years.
Professor van Zoonen
Yes, but an ombudsman is much more on a collective level. There should be individual mechanisms as well. In that sense, we know very little about what kind of crime identity theft is, for instance—who are the most vulnerable people, who are the most likely perpetrators and so on. That would be part of bringing back confidence in Government doing stuff about this, because it is one thing that is very high on the agenda of the people we spoke to, at least. Dr d’Aquin: There is an element that relates to transparent traceability—putting in mechanisms that make citizens able almost to audit by themselves how their data are being handled. It is not about having unnecessary technical complexity, but simply about being able to demonstrate the level of security that is being given to the data. You must also respond directly to any possible threat or issues. We had a discussion before about whether or not Government were leaking more data than private companies. I disagree respectfully with the answer that was given. I think private companies leak personal data much more. Almost every month one of the very large companies is shown to have been hacked and to have had data stolen from its system. The difference is that those companies are starting to get better at putting on a direct response by locking out all the accounts that may have been affected and getting in contact directly with the users who may have been affected. It requires very complex and demanding mechanisms to be able to trace back any security issue and how it may have affected users directly. That is the sort of mechanism that could make trust in the security of Government systems with respect to personal data more effective. You can never achieve complete trust in the security of a system, but you can improve trust in the people who are trying to make it secure and the traceability and auditability of the system.
Professor van Zoonen
It is interesting that in the last two years many of the big companies and consultative groups have issued reports about gaining online trust from their customers. Basically, they have the same discussion that we are now having, about how to assure trust and the secure handling of personal data. For them, it has become one of the key issues in maintaining a good business case. They recognise that if that trust is not gained from their customers they will lose business. It is a very simple economic argument.
Emma Carr
Going back to your original question, not only can you never be sure that anyone will keep your data safe but it would be completely dangerous to try and allude to that. Sometimes both the public and the private sector can overpromise about how safely data can be kept, whether those are being stored in the UK or outside. Going back to your question about the safeguards, the regularity framework around the Data Protection Act is pretty good, certainly for data in this country, despite its being fairly old. The one thing I would say about it is that there could be far harsher penalties available for people who breach section 55 of the Data Protection Act by knowingly disclosing information. That is something that could very easily be rectified. You could have custodial sentences available. Under section 77 of the Criminal Justice and Immigration Act, the Home Secretary has the ability to amend the Data Protection Act to include those. If people saw harsher penalties were available, rather than just very small fines—sometimes we see fines of less than £1 per piece of information that has been lost—that would give them more faith that the safeguards were there. We also should not rely just on the Information Commissioner to enforce the Data Protection Act. Too often, they have had to prosecute under the Computer Misuse Act or for the offence of misconduct in public office, because harsh penalties are available and it is easier to prosecute. We could allow civil cases to be put if I know my data have been lost, whether it is by the Government or a private sector company, as is possible in the US.
Stephen Mosley
You have mentioned the Information Commissioner. He has noted that individuals lose data rights once the data have become anonymised. Do you think that should be reconsidered? It depends on the situation, doesn’t it, but is it getting easier to un-anonymise data?
Professor Robertson
In narrow technical terms—to give you that kind of answer—this is a very difficult one. It is not as if data just sit there, you then do something with them, they go through and you can easily spot the routes by which they go through. Typically, very early on in the analysis of data they will be transformed into something else, with the same identifiers, different identifiers or a different structure. It is a bit like taking two numbers, multiplying them together and then trying to guess what the original two numbers were—it is not always easy. It is quite difficult to disentangle that. You can go the other way, of course, and take data where it would not have been obvious that somebody was identified, or maybe not obvious even that the structured data were there, because they were obtained from mining Twitter feeds or from unstructured text. All of those algorithms are getting really quite good now. All of that is generating data that in some sense maybe you never put in—or certainly that you never put in intentionally. That is being pushed in all sorts of different directions, because a lot of those data were already on the open web. That makes it quite a complicated topology. That is just technical. As far as I am aware, I never said anything that remotely verged on the sociological—it only gets worse. My point is that it would be very difficult to think of some gold standard or single route for doing it. Basically, it is about continuous vigilance and continuous work.
Emma Carr
I would certainly like to see the ICO address further anonymisation, the potential of re-identification and pseudonymised and de-identified information. From my awareness, as time goes on and more datasets are created, it becomes easier to re-identify information that has been de-identified. That is something that I would like to see the ICO address before we come up against any potential problems in the future.
Professor Robertson
I also do not want to give the impression that it is completely hopeless. One thing that you want is an appropriate level of de-identification. You want to reduce the probability of identification to some reasonable level, given the benefits involved. It is a bit like safety cases for aircraft. Occasionally an aircraft will indeed fall out of the sky, but not so often that typically it feels like we should ground every aircraft.
Chair
We are practical as well. We have suggested in the context of some health data, where sharing for research is valuable, that the data could be held in a safe repository and released under controlled circumstances, case by case. Those sorts of mechanisms are feasible, aren’t they?
Professor Robertson
Yes—those sorts of proportionate mechanisms are. The property of these domains is that it always looks really bad in the general case, but it is possible to make progress just by shifting the parameters a little bit, if culture allows. That is the crucial point.
Emma Carr
All too often getting the terminology right here is very important. When I have entered meetings, people have been talking about anonymised data when they mean pseudonymised, and pseudonymised data when they mean de-identified or statistical. Far too often they say “anonymised” when that is not what they mean, when we are talking about data sharing in practice. There needs to be a greater awareness within Government Departments, if they want to use data, of what the terminology actually is, so that when they try to inform people of what is going to happen they are informing them correctly. Dr d’Aquin: That is exactly the point I wanted to make, but formulated in a slightly different way. I agree with the Information Commissioner that truly anonymised data should not be subject to data protection, but I say that knowing that anonymisation can mean an awful lot of different things. There are other examples; the ones that have been mentioned are completely right. I have just read one about a city in the US releasing anonymised data about the taxi trips done during the year in that city. Those anonymised data were entirely and fully de-anonymised within two hours. It was not because the technology is better or because people are getting better at re-identifying data, but simply because the anonymisation mechanism that was used was provenly and very ignorantly weak. It was a big technical mistake. I like the idea of anonymised data; it can be extremely useful, if it is done properly. However, if you are talking about data that are personal or useful, I do not agree. What is needed is not only a proper definition but a proper framework—standardisation of what anonymisation means and what an anonymisation level is. There are mechanisms that exist in which single datasets can be assessed with respect to anonymity. Any discourse or release of anonymised data should come with a proper assessment of the level of anonymity and should be considered as having to follow data protection or not on that basis. That is a very important point. Basically, anonymisation can mean an awful lot of different things.
Graham Stringer
Not everybody uses social media, and there are parts of the country that are not covered by these sorts of electronic communications. Is it sensible for the Government to use information from social media as a basis for forming policy, because those datasets will always be inadequate?
Professor Robertson
While they are thinking, I will give you my academic’s answer. There is a different weight. There are some demographics, for example, that use less social media, but there will be pretty strong drivers for some of those to do so soon. For example, when you look at the size of the problem that is emerging with the elderly, the frail and the infirm, it is very hard to imagine some notion of technical social networking not being used. That will have to happen, because otherwise it is not going to work. If that does happen, you will see it shift. It would be brilliant to be prepared for that, rather than to try to exclude people from that particular ethic. We should be approaching it the other way round and trying to push it in, with the aim of making the other problems to do with demographic shift rather less difficult. That would be a more positive way of putting it. I realise that, being an academic, I have not really answered your question.
Graham Stringer
It was interesting. Dr d’Aquin: The most naive answer would be, “No, it is unreasonable.” At least, it is unreasonable to do it in a completely naive way. It is obvious that any data collection mechanism, including social media analysis, includes a bias, but the proper use of social science methodology includes assessing the bias and the effect it may have, and including any other complementary mechanism that can help to reduce that bias. While David’s answer might be to try to reduce it by making sure that people join in, there is also an element of simply assessing what is the effect of the bias introduced by the data collection.
Professor van Zoonen
In addition to what has been said about vulnerable groups not being on social media, it is important to realise that there is an increasing movement of people to getting off the grid altogether. That has nothing to do with vulnerability and everything to do with people not wanting to be traceable by energy companies, Governments or whatever and wanting to live off the grid. It is a serious social movement that would no longer be captured by social media analytics.
Graham Stringer
That is right. Where do you think the data on social media could give us most benefit in extra information and as an extra base for policy making?
Professor van Zoonen
There is a fantastic example in the health context, although it comes from the Netherlands, of failure of a policy because of social media. It relates to a disease for which girls have to get an inoculation when they are 13. I cannot remember what it is called, but when they are 13 years old girls need to get a vaccine against a certain type of cancer. In the Netherlands that policy failed because of a social media campaign by parents, who were very concerned about what was going to happen to their children. The year after that, the Dutch Government started their own social media campaign. Whenever there was a story about the vaccine being dangerous, they would put in their expert and explain what was going on. Since a lot of social media analytics is about marketing, advertising and campaigning, it is there that the Government could probably gain most—not by using the data but by being on there to explain their policy.
Graham Stringer
Is social media best—or only useful—when it is used in relation to information from other sources, such as CCTV?
Emma Carr
It depends on what you want to achieve. One suggestion is that social media data could potentially help the police with information that is usually gathered by the census or something like that. However, as has been said, the accuracy of social media data has been questioned in academic studies. I know that the ONS is looking at Twitter to see what it can ascertain from that that it could not necessarily ascertain from the census and the questions it has to ask every 10 years—or in addition to that. It depends on what you want to do with social media. If you want to inform people via social media, that is one thing; if you want to analyse people via social media, the safeguards around that are obviously different.
Professor Robertson
It is starting to develop enough outside the public sector that there may be something to be learned from looking at applications that were not developed there. It would take too long to go through all of the examples, but there are canonical examples to do with emergency response, where you very quickly build up maps and so on from just the population, because there is nothing else. There are commercial examples to do with navigating roads and so on. You want to get really fast, up-to-the-minute information about where there are blockages on roads, so you do it through the social network rather than by trying to model the road network. In the commercial world and that kind of funny world of people just doing stuff, there are an awful lot of those kinds of examples. The bigger question for this kind of audience is, how many of those would translate and survive were they to be used in Government? Government is a big lever, because it could potentially use data that it has acquired to jump-start some of these things or to provide a bit of an incentive for people to be involved. However, as you may be aware, there is rather less expertise on how you do that. Dr d’Aquin: I cannot talk generally, but from conversations we have with local government, with the police about crime and with people working on disaster management and things of that sort, I can say that there is an increasing awareness that social media data have two extremely good qualities: first, there are a lot of them; and, secondly, they are fast. In many cases—I take the example of crime reporting—that is a double-edged sword. It comes with needing the ability to process the data in very large amounts and to make sense of them, to reduce noise and to enable early signals of something to look at in more detail to emerge. In many cases it cannot replace longer-term, richer and more sophisticated ways of data collection. What I see emerging as a pattern is the use of social media to detect in a very rapid way early signals of a phenomenon that then has to be validated through the use of other data. In crime reporting, especially, the example you give of CCTV is an absolutely obvious one. Looking at Twitter to detect that an issue may be happening somewhere at that moment means that someone can go and watch a real stream on a CCTV camera to validate that something is indeed happening and then take appropriate action. In many cases false reporting, noisy data or simply misinterpretation of what in the end is a very large amount of unstructured data will show that such validation mechanisms are absolutely necessary.
Graham Stringer
Thank you.
Chair
Stephen has a couple more quick questions.
Stephen Metcalfe
I want to go back to the legislative framework that is currently in place to control the way our data are used. The EU is proposing some changes to the Data Protection Act. There seems to be a mixed reaction to that. TechUK said it would have a “negative impact” and Big Brother Watch said it would improve individuals’ control over the amount of data. Briefly, what is the purpose of data protection legislatively? What is your view on these changes? Is there a balance to be struck somewhere that will mitigate against the negativity while allowing individuals some control over what is done with their data?
Emma Carr
One of the things that Big Brother Watch quite liked about the new EU regulation was that it shifts the burden of evidence away from the individual to the researcher. No longer will the individual have to explain why those data should be kept by the researcher—the researcher will have to justify why they should be kept. That is one of the safeguards. It gives individuals more control over their information, which is always a good thing. EU data protection is incredibly important because of the online world we now live in. I would like to see some kind of global principle for how our data are used and shared, because the internet has completely changed the way in which those data transfer across borders. This new regulation is also about additional safeguards for when data are moved outside the EU. I believe there is a process where the national data regulator has to get involved and to sign up to some safeguards saying, “If these data enter our country, we will be responsible for them.” That higher threshold is important and is reflected in the individual nations, which tend to adopt the same principles. Anything that can be done to give individuals more control and to have a higher level of safeguards is a good thing.
Professor van Zoonen
In this respect it is interesting that, when we asked our respondents what they feared most in the future about data sharing, their biggest fear was that they would have to pay for privacy and would always lose out to big business interests. If you contrast the EU policy and citizen protection with what techUK is saying about the economy suffering if we do not allow this data sharing, that is exactly where this public fear is located—that the citizen interest will lose out to big business interests. It is necessary to take on board that kind of sensitivity and what people are thinking about these issues. Dr d’Aquin: I am not profoundly familiar or at least as familiar as my colleagues may be with the changes that are proposed, but one of the key arguments is that, while, in principle, giving more control to users, these changes will introduce a bigger need for companies to invest in data infrastructure—in mechanisms such as responding to requests for the right to be forgotten. These mechanisms of this investment might hamper the economic growth of Europe in this particular area and might simply introduce additional obstacles to the business and economic use of social media analysis. That is a valid concern, from a purely economic perspective. In response to that, the important aspect is to provide support for businesses to make this investment and, especially, to factorise, understand and show good practices and how to make this investment adequately so that the economic benefit of social media analysis is not reduced by the need to comply with the regulation.
Chair
There are other examples beyond the US, but my final question relates specifically to the US Patriot Act. You will be glad to know that the notes we have on you on these iPads are not out in the iCloud. The House took that decision consciously, because it is not held within the framework of legislation the UK has control over. Clearly, legislation like that is going to impact on ordinary citizens. Do you think the Government should be doing more to help people understand how their data are held in other countries and what rights they have in respect of those?
Professor Robertson
Yes.
Emma Carr
Yes.
Chair
I do not think I have caused any controversy there—I am just curious. I might have if the American ambassador were here. Looking at it from the British Government’s perspective, should we adopt a Patriot Act kind of approach to gain access to social media data?
Professor Robertson
The things that are important are the evidence base and transparency. That is the difficulty with the Patriot Act. You can understand how it was rolled in at a very emotional time, why it was felt to be necessary and so on—that would be a very long conversation—but the way that it has played out has been very murky. The effects of that are pretty wide ranging. It is not just the Patriot Act but everything that went along with it. Even if you look at the research community in pure science, you will find that quite a lot of people working in security in that community feel personally undermined in their work by some of the things that are believed to have been going on. A lot of these things are beliefs—you do not know exactly what happened—but it has certainly demolished a huge amount of confidence. If you had to think of how to demolish confidence with public bodies in that area, that would be a good way to do it. I could not have made it up. On the other hand, we have just talked about the EU directive and so on. In my personal opinion, that is another example of something that is quite broad-brush, so it is very difficult—in the other direction, but still a bit unwieldy. Both of those things feel slightly uncomfortable to me. Somewhere there must be a pragmatic middle route, because we have to do something. As you say, part of it has to involve simply explaining, “These are the things that we believe we do, to the best of our ability and our ability to tell you. Here’s how they are laid out.” That would be a tremendously positive thing to do.
Emma Carr
If the question is, “Do we need a Patriot Act to get access to communications?” the response is no. We already have things in place in this country to allow people to have access to them, such as the Regulation of Investigatory Powers Act, the Telecommunications Act and the mutual legal assistance treaty, for access to communications from other countries. We have to be very careful when we think about the Patriot Act and its repercussions for the United Kingdom. As has been said, it is not just about communications. We must remember that in this country we have a different legal framework from the US. Dr d’Aquin: One thing that was quite misguided about the US Patriot Act was that it was realised in a slightly naive way, possibly because it was rolled out very quickly, at a time when it was felt to be necessary. Under the Act, the only approach to data collection to fight terrorism is the bulk collection of all information about every citizen and filtering that out a posteriori—after or when it has been collected—which sounds really naive. The argument about the need to collect data for security is not one that I would feel confident to have myself, but it requires complex and sophisticated mechanisms to ensure that data are not misguidedly and naively collected, in a way that might have bad repercussions, that data are collected at the time when they are needed and that the conditions under which data are collected about specific individuals are made extremely clear to the public.
Chair
This is a debate that will go on. I was going to come back to Professor van Zoonen to point out that we pay for privacy already, in the sense that if you do not want a Clubcard you do not get the discounts and so on. There are plenty of examples like that. Thank you very much for a very good evidence session. Oral evidence: Social media data and real time analytics, HC 245 34