Oral evidence - Fake news

Examination of Witnesses

Witnesses: Professor Stephan Lewandowsky, Professor Vian Bakir and Dr Caroline Tagg.

Q189 Chair: Good morning. Thank you for joining us for this further session of the Digital, Culture, Media and Sport Select Committee inquiry into fake news. My apologies that we are slightly later starting this morning but I thank the witnesses for their patience and for joining us.

The purpose of today’s session is to look not just at how fake news works but why it is potentially such a powerful tool and how influential is it on people who are exposed to messages delivered to them in this way. I would like to start off by asking the members of the panel if they could just give us their own perspectives on why they think fake news is potentially a powerful tool and influential. A lot of the evidence we have taken so far has looked at the architecture of fake news, how it is distributed, how people are targeted, but do you feel that it is an effective way of distributing information and why is it that people are susceptible to it? Perhaps if I could ask Professor Bakir, if you would like to start, and we could move along the panel?

Professor Bakir: I do not think I am the most qualified on this panel to answer this. In terms of misinformation, it is such a complex thing for us. There are psychological issues. There are sociological issues as to how and why people spread misinformation, why they remember it, and there is also a whole load of issues about whether people can recognise deception in the first place. I know you will speak to this extensively so I will not speak on that much. There are all sorts of biases, which I am sure you will speak of—cognitive biases, availability bias; that is another one. If people are exposed to stuff a lot they tend to remember it. If people are exposed to stuff that confirms their pre-existing views they tend to go for it. These are psychologically recognised empirical things, which have probably been talked about. [Interruption.] There is a bird in the sound system.

Chair: It is good to know we were not being hacked by the Russian Government. It is only a parliamentary IT problem.

Professor Bakir: Then people are quite bad at recognising deception as well. People like to believe that stuff is true when it is presented to them. Research shows that people are gullible; confirmation bias again. People just take from stuff what they want to see. So all of these reasons. Over to you, Steph, because you know this better than me.

Professor Lewandowsky: I think the most important problem we are facing is that once people hear something or take in information it is extremely difficult for them to unlearn that and to undo that. If I tell you that the moon is made of green cheese, chances are you will not forget me having said that. If I now say the moon is not made of green cheese then you will hear me say it again except with a negation in there.

It turns out that people are extremely challenged by processing simple negations. If I tell you that John is not tall you will probably think of him as being taller than if I tell you he is short. Negating inherently is a very difficult process for people to achieve. With misinformation we can show this over and over again. There are hundreds of studies on this. There is no ambiguity about the outcome. Under most circumstances, if we try to correct people’s beliefs based on what they have heard they may adjust their belief slightly but there is a lot of evidence to suggest that they continue to rely on that information nonetheless. That is just a basic cognitive problem. We are just designed in that manner. It is very difficult to get around that unless you provide people with an alternative or you make people suspicious of the source.

There is data to suggest that jurors, for example, will not disregard tainted evidence even though they say they do because cognitively that is extremely difficult. However, if you tell the same jurors that that tainted evidence was introduced because the prosecutor was trying to get promoted or something and there was a suspicious motive underlying the information in the first place then people can dismiss it because they have an alternative explanation.

We know how we can get people to dismiss information but generally it is extremely difficult. So the cognitive consequences of fake news are pervasive and we can take them quantitatively in lots of different contexts.

Q190 Chair: Would it be right to say, from what you are saying, speed is an underlying factor in fake news because if you can get to someone with the disinformation before they get the facts then it is more likely to be influential?

Professor Lewandowsky: Precisely, and we call that prebunking or inoculation. If we can get to people before the misinformation does then there is evidence to show that they will be able to filter it out better. One example is a recent study that I published with colleagues last year where we told people about the way the tobacco industry in the 1950s and 1960s was trying to create the appearance of a scientific debate about tobacco when, in fact, the science was quite clear.

Once you remind people of that precedent, they then became extremely resilient to misinformation about climate change, which followed the same playbook. So it is possible to give people an inoculation, like a vaccine almost, against misinformation by pointing to specific rhetorical strategies that are misleading but you have to get to them first.

Q191 Chair: That is quite interesting. So your view then is a strategy to combat fake news is to talk about fake news quite a lot and make sure people are aware that it is out there?

Professor Lewandowsky: Ahead of time, yes. The more you get to people ahead of time and educate them about discernment—how can you tell a Twitter troll from a real Twitter person—chances are that has a positive effect but I do not think it’s the only way to go by any means. We can train people. We can teach people, we can educate people—all of that. I think we also need to change the infrastructure of the whole ecosystem of online information to make it harder for misinformation to spread.

Dr Tagg: I think my views are largely in line with my fellow panellists but what I would point to is the fact that beliefs are not individual but are very social, people believe the things their group believe. My research, which I have been carrying out with my colleague Phillip Sargeant, who is sitting behind me, has been looking into how and why people communicate on Facebook, looking at Facebook specifically. We have been using surveys and interviews to get at people’s perceptions of Facebook, their ideas about Facebook. So, why they post on Facebook in the first place, what they think Facebook is for, what they think Facebook is not for, what they think the norms of communication are and then exploring how these ideas shape their behaviour on Facebook.

One of our findings has been that what is important to people on Facebook is maintaining social relationships. To a lot of our participants in our study, Facebook is not a news media site, it is a place where they carry out quite complex maintenance and management of their social relationships. We can talk more about this later, but Facebook is very complex in that regard because when you are posting a status update on Facebook, you are talking to people you know but people you know in very different ways. You have lots of different social relationships in the same space that you are trying to negotiate.

The relevance of that is that when someone is deciding whether to share an item of news or a post or whether they are choosing to flag it up, to denounce it in some way, they are doing so in light of what impact that will have on how they come across in their different relationships. I think that does potentially make fake news particularly complex to deal with and means that any way of addressing the situation—and we would agree that the infrastructure and so on needs to be addressed. It also needs to take into account the kind of social situation that people are operating in.

Q192 Chair: Yes. Do you mean that it is not just that people are receiving a source of fake news not necessarily from the originator but from a friend?

Dr Tagg: Exactly. Yes, one huge consideration is not only, “Oh, I’ve received it from someone who is a close friend” or, “My boss at work” or—but also the wider social situation. So this might be a friend of a friend who you do not want to offend or upset but also it is all about how you come across to all the people on your Facebook friends list. It is very complex in terms of how a decision someone makes as to how to respond to something that might potentially be fake or misleading.

Q193 Chair: Leading on from that, Professor Bakir, you have written about algo-journalism. I wonder if you could just explain to us a bit more about that. Is that stories that have been written with a particular audience in mind?

Professor Bakir: Yes. It is basically automated journalism. Journalism that you can farm out to a computer to do. It tends to be done with very data-heavy stories, so sports. If you have a whole load of sports results across the country you can train a computer with an algorithm to report on who won what match very quickly. So you do not have to send reporters out to cover all the football matches in the country. You just have the stats, you put it into your algorithm, your programme, and that will generate the story.

You can do this for anything where there is lots of data. You could do it for financial reporting. You could do it for elections if you have the data on polling and where has it gone. It is basically a very quick form for journalists to be able to farm out lots of stories and then those stories can be targeted according to who your audience is. If you have loads of people across the country, they might be interested in a local football match or a local election result. They do not care about the rest. You can target that result and that story to them. It is increasingly coming into journalism as a form of speeding up the process with fewer resources. You do not need as many staff on the ground to write the story.

Q194 Chair: How is that different from alerts? I use Google search on my phone. If I go on my search screen it will tell me the score of the Manchester United match because it has guessed that I am interested in Manchester United because I search for them and it will feed me information like that. The kind of small bits of news that are event based, data driven, it is quite easy to do. Can you use that to send people more sophisticated bits of information?

Professor Bakir: There have been experiments over the last year where the sophistication of the story around the data has increased. I think The Washington Post has been experimenting with this across the last US election, so you get more of a narrative placed on top of the story. Certainly other outfits and people who are into all of this are experimenting with things like the tone of voice you might use to portray that information. It is not just pure facts, right? You get a story in emotional tone as if a real human being had written it and thought about; what does this type of audience want to know when they read about this sports fixture or this election result? We do not respond very well just to pure fact anyway. We are inherently story-driven people. So if you can generate a story in a human way, with something that is receptive for the human being, it is going to resonate more basically.

Q195 Chair: I have seen some work that someone has done looking at hyperpartisan content and tracking a breaking news story and then how it has been re-reported through different platforms in an increasing hyperpartisan way in terms of tone of voice. From what you are saying, presumably you could use AI to do that, to change the language around a story.

Professor Bakir: I do not think it has been done yet but I think it is just around the corner if it is not. I do not know if you know anything different from me but I think there is an absolute possibility for this. Emotional tone plus targeting plus whatever fact or misfact or misinformation you want and you could potentially distribute that nationwide.

Professor Lewandowsky: Can I connect to that just to amplify the point? There is research out there suggesting that if you know a person’s Facebook likes then if you have enough of them you can predict their personality better than their own spouse can. There are algorithms that if they get 300 likes of a Facebook user they can outperform that person’s spouse in identifying their personality traits automatically.

It only takes about seven or 10 likes to do as well as your colleagues at work. It is possible to target specific messages if you have access to the information that Facebook is holding about all 1 billion of us, at least, who are using that platform. It is quite clear that during the election in the US in 2016 that capability was used to some extent to target messages to specific users.

We now know that as of November Facebook says 150 million Americans were exposed to Russian-sponsored Facebook ads. We do not know how customised and how targeted they were but we do know that the knowledge exists to target them on a specific person based on their personality.

Q196 Chair: I know the colleagues want to come in, and I want to bring them in. You said, “We know the knowledge exists”. How does the knowledge exist?

Professor Lewandowsky: In the research I just mentioned, experiments have been done where you compare the performance of algorithms looking at Facebook likes to a person’s spouse, and then you find that the Facebook technology, the likes, is outperforming the spouse in predicting a person’s personality.

Q197 Chair: So that is externally done. That is someone just looking externally at the platform at an individual user’s likes and preferences and then guessing what else they might like based on that.

Professor Lewandowsky: Precisely. This was done in an experiment where they—I can give you the full reference later about the paper that did this but we do know that the technology is there to do that.

Chair: Okay. If you give us something later that would interesting.

Professor Lewandowsky: Sure. Could I make one more comment, connecting to what you were saying about the social environment and the importance of that?

Chair: Sure.

Professor Lewandowsky: I think again that is under-appreciated, how that has changed with the technology. One of the things we know with great certainty is that people’s opinions are held to the extent a person thinks it is shared by others. We are social animals. If I think everybody else thinks the world is flat then I am not going to change my mind on that. The problem is that 300 years ago or in the middle ages the proverbial village idiot in Gloucestershire, or whatever, might have thought that the earth was flat and everybody kind of knew, “Yeah, right”, and he knew that everybody else thought, “I am not quite aligned with everybody else”.

Today, that same person, no matter how absurd the belief is will find a community on Facebook because with a billion users there will be somebody else in Denmark who thinks the earth is flat and then they are joined by somebody from Turkey and all of a sudden they have an epistemic community. The moment that happens their opinion becomes entrenched because they see no reason to change it. The social signals are telling me, “Oh, everybody else feels the same way”.

We have data on that. In a recent Australian study on climate change only 8% of people were found to completely negate the idea that the climate is changing but those 8% thought that their opinion was shared by half the population and that was because they were all in this echo chamber and talked to each other and felt their opinions confirmed. I think that is a novel problem that is inherent to the technology. That people think, whatever they think, everybody else thinks the same way.

Q198 Ian C. Lucas: I was very interested in this idea, the dissemination of stories based on data, for example football results, you were talking about. It depends on the veracity of the data. Are there any examples of inaccurate or misleading data being disseminated in that way?

Professor Bakir: Not that I know of because the stories that tend to have been farmed out to the data driven journalism does tend to have good data and it is easy data, did you win or not, how many goals did you score, or did this candidate went on—

Q199 Ian C. Lucas: You mentioned financial reporting. Potentially that could be extremely precarious.

Professor Lewandowsky: I guess it could. I have not heard of any problems with it yet but, yes, it always depends on the quality of the data, always, when you go into this algorithm data-driven journalism as well as whatever the algorithm is.

Q200 Ian C. Lucas: There is no process for checking the data through this. What is the process at the moment for ensuring that the original data is accurate?

Professor Bakir: Other than standard journalistic processes of checking; is your story truthful or accurate or your data sources truthful and accurate? Beyond that I would not know.

Q201 Giles Watling: It is fascinating. I read somewhere a report that people resist information that challenges pre-existing beliefs and that people’s voting habits tend to go that way so fake news can clue into this and leadership seems to be somebody who steps outside and posts bravely and people tend to follow. I just wanted to ask—people almost seek out fake news in a sense as long it supports their pre-existing beliefs. How do they go about finding information that supports their pre-existing beliefs? How do they do that?

Professor Bakir: I think the information finds them with the algorithms because—

Q202 Giles Watling: It is the other way round.

Professor Bakir: If your behaviour online shows that you like certain products, parties, ideas and belief systems, which it can show, then you can be targeted.

Q203 Giles Watling: These communities reach out to individuals who show certain propensities.

Professor Bakir: That is a more deliberative way and that will happen as well. For instance, in political campaigning you will definitely find that people are encouraging other people, “Yes, let’s go into this Twitter room and let’s promote this particular hashtag and this particular message”, for instance, for this particular political party or candidate or idea. It is an extra. You are being targeted by the algorithms, which are commercially funded by them.

Q204 Giles Watling: Of course, it can be much more serious in a sense.

Professor Bakir: You are also being deliberately pushed at certain times by political persuaders of all ilks.

Q205 Giles Watling: It could affect financial markets as well.

Professor Bakir: Yes, some of the submissions to this inquiry have pointed to examples of fake news adversely affecting the financial markets.

Q206 Rebecca Pow: Just on the Facebook issue that—I cannot pronounce your name, Professor—

Professor Lewandowsky: Neither can I—Lewandowsky.

Rebecca Pow: Lewandowsky. Facebook have said that they are going to just focus on family and friends in the future. Do you think that will have any effect on this data and in curbing misinformation, or do you think they will still be gathering lots of knowledge that is still going to drive what can turn into the fake news being reciprocated?

Dr Tagg: Our focus was more on news that was being shared by friends and family, so in that sense what Facebook are doing now does not address that aspect of the problem. What we saw in our research was how users’ actions contributed to the creation of the filter bubbles so obviously the main culprit is the Facebook algorithm that filters out things that it thinks users are not going to be interested in.

What we saw users doing was if they came across a view or a piece of news that they disagreed with rather than engaging with it or trying to argue with it or even denounce it, they did not want to do that because of the social situation that I talked about where they are managing lots of complex relationships and they are all so concerned with how they come across and also—this is on Facebook—trying to maintain conviviality. They see Facebook as being a place that should be light-hearted and fun. So for all those reasons rather than engaging with things they disagreed with they would quietly block them.

Depending on who the person was who posted it they would either block the post, which means that posts from that user no longer appear in your news feed but they do not know that you have blocked them or in some situations where they felt they could they would unfriend the person concerned or in most cases they would just ignore it. So these actions, which are a way of people dealing with the social situation of Facebook, to some extent contribute to that filtering out of views you disagree with, which does not directly have a bearing on fake news but it does increase the filter bubble effect, which increases polarisation of views, which enables fake news to then be circulated more easily, if that answers your question.

Q207 Rebecca Pow: It is just that they have said they are going to try to filter out so many of these viral videos, things like that, as if it is going to help. My point was will it help spread fake news or this data that is being collected about these people? It can be used, but you suggest that you do not think that is going to have any effect.

Dr Tagg: No, not necessarily. We have not looked specifically—obviously this change has only just happened.

Rebecca Pow: Yes, that is right.

Dr Tagg: We were not looking specifically at that. I suppose our interest was more in what happens when news is shared.

Professor Lewandowsky: There is data on this. There is a lot of data on this now and the studies that I know of pretty much say that sharing behaviour on Facebook or Twitter is not affected by the veracity of the content. So whether something goes viral or not is determined, effectively, just by emotional appeal—by the user’s preferences themselves—but it has absolutely nothing to do with the content, whether it is fake news or whether it is real news. Would you agree with that as a first approximation? There is no trace of veracity as being a determining variable and people sharing the—

Dr Tagg: We would say that we cannot assume that veracity is the most important criteria guiding users’ behaviour. There are lots of other concerns they are thinking about, including affiliating with particular groups and aligning themselves with particular viewpoints and coming across in particular ways. I do not think we can assume that people are necessarily guided primarily by “Is this true?” when they make a decision as to whether to share something.

Professor Bakir: I think Facebook are trying to make the fake news problem go away. I think it is just—they are trying stuff, basically.

Dr Tagg: Yes. That was what I was—

Professor Bakir: They are just trying stuff. I think they do not want news to be the key feature of Facebook. I think they would love the fake news problem to go away, and it not be their problem because, from what I can understand, they do not want to intervene as an editor. They do not want to be taking those decisions about what you are seeing. It seems quite genuine. I think they just want it to go away as an issue.

Q208 Paul Farrelly: Going back to the last statement Professor Lewandowsky—I thought you were a footballer when I first sat down.

Professor Lewandowsky: He is spelt with an “i”.

Q209 Paul Farrelly: I have to suspend my belief or disbelief at what you have just said about that in terms of veracity not being there, because it either speaks volumes for the way the research is conducted or even louder volumes for the mentality and intelligence of people who are the most active users of social media.

Professor Lewandowsky: I know it is hard to believe, but I have the papers right here. There have been two or three papers recently—massive analyses of Facebook sharers that come to that conclusion, so unfortunately it may well be true.

Q210 Paul Farrelly: Could you share those with us because that is quite important? Would you send them to us afterwards?

Professor Lewandowsky: Yes, sure.

Q211 Paul Farrelly: I, unfortunately, in this day and age, have to be on Facebook, which I originally joined as something like Friends Reunited but have found out since it is a lot more. I am forced to be on it because politics dictates it these days, and general elections. When I get messages, another plague, on Facebook that are telling me to do something, which are quite obviously and stupidly false, I just think the people who are sharing are idiots and I take no action. Are there any intelligent people out there using the platform?

Professor Lewandowsky: I think honestly people are as intelligent as the environment makes them, up to a point. Let me give you one example. This is something that I have been arguing for recently. We have to solve the problem, not just through education but also by changing the infrastructure. I call that techno-cognition because it is taking together cognition, what we know about cognition and the technology. One example is the Norwegian equivalent to the BBC, which is probably called the NBC or something, which has introduced a facility whereby if you want to comment on a news story on their website you first have to pass a quiz, a multiple choice quiz, about the contents of the piece you want to comment on. If you pass that quiz then you are allowed to comment.

To me that is terrific. It is a wonderful idea because you are not censoring anybody but you are slowing people down so they have time to cool off plus you make sure they at least know what they are going to comment on. That is one example of how you can put in something that is not censorship but nonetheless is likely to raise the quality of online debate.

Q212 Paul Farrelly: One of the things that we have prepared for these evidence session is suggested questions. We do not stick to the script. All the suggested questions have the term “people” in them and you used the term “people” and there are six billion of us, at the last count, so it is a generalisation. Is there, from your research, just to try and hone this down, a spectrum, a graph, a bell curve of what you might call a scale of gullibility or lack of scepticism among people and how does that correlate, if you have done any research, with the most active users in particular ways and shares in, say, Facebook data, which of course is only held by a company and platform like Facebook?

Professor Lewandowsky: It an excellent question and we are just beginning to look at that. I wrote a review article recently, which I think I circulated before the meeting, where I touched on the available research. This is American data. This is specifically written about the American public. There is some evidence that conservatives in America are more gullible, quote unquote, more susceptible to endorsing nonsensical statements to find them profound.

Q213 Chair: It might be a political decision whether you find them nonsensical or not, I suppose.

Professor Lewandowsky: I think the data are quite tentative on that. I report them because they exist and they are large-scale studies. I am not sure that that is generally true, but the big question you are asking is can we identify people who are more prone to misinformation. We are just starting that endeavour, so I cannot give a definitive answer on that. There is some evidence suggesting that people who are inclined to even pseudo-science, sort of homeopathy and non-scientific approaches to healthcare, nutrition and so on may be more susceptible to misinformation. Again, we are just in the early stages of that. I wish I could answer the question, but I do not have enough data.

Dr Tagg: I was just thinking that the question presupposes that the most important thing for the user is whether the story is true, which might seem to us a perfectly reasonable assumption, but if you think about what the user is doing and possible anti-establishment beliefs that they might hold, there are any number of reasons why a user might choose to like a post even if they think it is not true.

We have been thinking about and writing around the subject of fake news—not so much as false news, which was a distinction that Facebook made, and misinformation—as a way of taking a stance of denouncing a media source. We have seen quite high-profile cases, but one can now dismiss something one does not agree with by just saying it is fake news. So fake news in that definition has become—it has become more than just misinformation or something that is not true.

It has become something quite worrying for democracy and for national debate, in the sense that it undermines our shared factual basis, and that is again driven by social considerations, people acting within groups to achieve social things that have nothing to do with whether something is true or not.

Q214 Paul Farrelly: I just have one last question on this, because it is quite interesting. I can quite readily believe that a person who repeatedly points their finger and cannot control their digits and shares things that they either know or could never believe would be possibly true umpteen times belongs to that category of person that might be identified by the platforms that have all the information inside, as one of the people least likely to ask of a particular story, how do you know it is true, or if someone is relating something, how do they know it is true so I know it is true.

I see, Dr Tagg, from your research you have done a book on taking offence on social media, so that much be a rich seam. Also lots of your research has gone into the role of Facebook in fake news. What are your conclusions about the role of Facebook in fake news?

Dr Tagg: I should start by saying that our research has been just on Facebook, and I think it would be very different on different sites. Facebook is a very specific and quite complex platform, in that you tend to friend people who you know, which is one difference from other social media platforms, and people you know from different offline contexts. This is talked about in terms of context collapse. All the different contexts that you operate in in your offline world collapse into Facebook, which becomes a problem to manage in terms of self-presentation and how you come across to people.

One thing that we found was that the participants in our study were very focused on was being convivial. Creating a particular kind of atmosphere was important—so not offending people, not upsetting people, seeing Facebook as a space where you should not be doing that kind of thing. It is a place for doing trivial stuff and just being light-hearted and that environment, potentially, led to people not challenging things that may or may not be false. I think that is an environment where you can envisage people letting something go. It was almost, in some respects, a kind of indifference.

Q215 Paul Farrelly: Okay, that is an effect. That is a consciousness born of the environment rather than Facebook itself promoting fake news to those people.

Dr Tagg: Yes. It is interesting what you were saying earlier about Facebook not wanting to be a media company. In fact, this is a good example of where Facebook has become something that its users want it to be, which we have seen with other platforms too. SMS text messaging was never set up as something for normal people to use and Twitter has changed a lot in response to what its users are doing. It is interesting to think that Facebook has become something through its users’ actions and agency, and become something that Facebook itself would rather not be because it is so problematic.

Chair: I think if I can bring in a few of the colleagues and if there are points that you have a chance to make at the end I am very happy to take them but we want a bit of progress now I think.

Q216 Simon Hart: We have touched on a lot of these areas before, but is there any evidence out there at the moment which would suggest that older people might be more believing of some of the material on Facebook than younger people? Has any work been done on any suggestion that people in different age groups, backgrounds, area of the country, may take a different view on these things?

Professor Bakir: I think backgrounds are important—backgrounds in terms of political affiliation, for instance: if you strongly believe in something or you strongly support something. If you are very partisan, for instance, I think you will be filtered by your tribe, your partisan tribe, and fake news coming in and recirculating through that has been demonstrated in America.

An extreme right-wing media ecology is something that Facebook and Twitter are part of, but so are other extremely propagandistic news websites like Breitbart and Fox to a certain degree, and if they are all circulating the same fake news messages, disinformation propaganda or whatever, then evidence shows that people are quite happy to retweet and repost those messages within that community? Even if they connect to more mainstream news as well they will only take those bits of the mainstream news that confirm, because that have a critical bit in mainstream news that confirm what they already think.

I think it depends on your identity—your identity that you, as a group, belong to in that online space. I do not know if it is about age as much as identity. I think that there will be differences there.

Simon Hart: You are suggesting—sorry, there is a bit of head shaking on your—

Professor Lewandowsky: No, the head shaking is to say, “Well, I don’t have a good answer to your question about age”, but I totally agree that identity in the worldview in all of these things is the overriding factor.

Q217 Simon Hart: The reason I asked the question—let us turn our attention to the world we inhabit, elections and democratic process. I have a constituency where over 30% of voters are pensioners. As an example, during the election—and this is not a criticism, it is an observation—people were more inclined to believe what they saw on Facebook if they were in the 75 to 85-year-old age bracket, with my mother being a case in point, and perhaps there was a younger, more cynical group; we do not know. I hope I am not stereotyping or being too condescending. There was definitely an element of, “I saw it on Facebook therefore it is true”.

The reason for my question is this: you do not have to influence a lot of people. If you can identify a sector of society and target it quite cleverly you only need to change, or sow a seed of doubt in, relatively few minds by use of misinformation to very successfully alter the outcome of an election. Have I become too cynical or is there a chance that that is true?

Professor Lewandowsky: I would agree with you. I think that specific targeting is an extremely powerful tool, and I am guessing that as time passes we will learn more and more about how that was done during the American election and to a lesser extent here in the UK during Brexit—the specific targeting of specific voters with specific messages. I totally agree with you on that, but again I cannot answer your question about age. I just do not have any data on that.

Q218 Simon Hart: Does anybody know? The lack of information, should that worry us?

Professor Lewandowsky: I think if I had a long look at the literature and did an intensive search I would probably find a few things that speak to that issue. It is just outside my knowledge right now.

Q219 Rebecca Pow: What about gender as well? Do you have any knowledge—

Professor Lewandowsky: The same answer I am afraid. The overriding variable that I discover and other researchers discover is political ideology and the worldview. People’s political leanings determine about 80% of everything that is going on out there on social media or in their attitudes towards climate change and so on. Then, yes, I look at age and gender in my studies but it is like either nothing shows up or it is a very small effect.

Q220 Simon Hart: A distinction between fake news and misinformation: fake news tends to be the equivalent of a BBC news story that just does not happen to be true, whereas misinformation can just be a subtle and rather derogatory reference to somebody in public life, which may or may not have an influence particularly around election times.

Is anybody making that distinction, because I think one of the panellists spoke earlier about Facebook trying to retreat into an area of family and friends only? That, to some extent, I do not think will make any difference whatsoever to some of the issues that we deal with here because what we are looking at is situations where individuals are politically motivated and are offering misinformation about candidates and elections, which is having maybe a 1% impact, which is enough to affect the outcome. Are fake news and misinformation two very distinct things?

Professor Lewandowsky: Yes, totally. I think there is work now on categorising different types of misinformation. In fact, ine one of my recent papers we have come up with five different categories of misinformation. I do not have time to get into them all, but basically I would differentiate between two things. There are specifically targeted curated lies that are designed to achieve a particular purpose on the one hand but that make a claim to reality. The issues that come to mind would be what I call climate denial, where there is a systematic attempt to claim the same reality with an alternative story.

Then on the other hand we have a sort of free-for-all constructivist shock and chaos misinformation in fake news that do not even make a claim to reality any more. I think a lot of the information that is being pumped out by Twitter bots is so absurd that you would think, “Why do people even believe this?” I think the purpose or the consequence of that is that we have moved from a world where there was misinformation to a world where the very concept of the knowability of truth is under attack.

Q221 Simon Hart: If 0.25% of readers believed it and acted on it could that not have an impact on the way a country is—

Professor Lewandowsky: Absolutely. All of this has a big impact, but I think the most important development right now is that we have moved from a world where people competed for a truth to a world where people are increasingly saying there is no such thing as truth. You may remember Kellyanne Conway, one of President Trump’s advisers, talk about alternative facts: “We have alternative facts”. Then in this country Katie Hopkins recently, in connection with a car accident, tweeted that it was a terrorist incident. When she was corrected she said on TV, “Well, there is no such thing as truth any more. There is only what I think the truth might be therefore I can tweet anything I want.” That appealed to the view basically that there is no such thing as truth.

To my mind, that is novel. That is not something you would have heard 10, 15, 20 years ago and that is what I would call a a shock and chaos approach to fake news, undermining not a specific factual issue but the whole notion of truth, which I find is much worse.

Q222 Christian Matheson: One very quick question, Chair, if I may. You seem to be suggesting that the problem with fake news, aside from the point you have just made about undermining the notion of truth, is reinforcing people’s existing prejudices, as opposed to changing their mind. If that is the case, how can the fake news phenomenon be useful to those who would seek to subvert political or democratic processes, if you are not changing people’s opinions?

Professor Bakir: A complex question. Of the few studies that have been done on recent elections, America being the most studied, is that changing people’s opinions? Maybe it is not changing the opinions of the people in the filter bubble, but the filter bubble can then get the mainstream media to cover what they are thinking and they can start setting agendas for the mainstream media. This is what has been demonstrated in the American election.

The extreme right-wing media ecosystem, largely online, with all their fantasies about Clinton’s e-mails and corruption and all this sort of stuff, because they were making such prolonged and repeated claims about these, they forced the mainstream media in America to address these issues. A content analysis of what the mainstream media covered in the 18 months up to the election showed that they predominantly covered what the Trump campaign wanted them to cover, which was Clinton’s e-mails and scandals around the Clinton Foundation. So you might not be changing the minds of the people in the filter bubble, but you are infecting wider mainstream media with the ideas.

The mainstream media may pooh-pooh those ideas, but they are still talking about it, so there are people whose minds are perhaps not so convinced who are now suddenly exposed to these ideas through the mainstream media because they have listened to what the filter bubble has said. There is definitely a mechanism where you can change people’s minds—not necessarily the filter bubble people but people outside of it. That has been shown.

Q223 Brendan O'Hara: Slightly picking up on what you said about that loss of what we regard as traditional gatekeepers and how fake news affects people’s response to what we could call traditional journalism, if all views are equally valid and everyone’s opinion is given the same weight and credibility what happens to objectivity and balance in those circumstances?

Professor Bakir: We do not have any—end of story. They do not care though, the people in the filter bubble, because they believe what they believe. The people outside the filter bubble care and they try to correct the filter bubble, but the filter bubble people are not engaging with the people outside the filter bubble, so that does not work either. There are lots of techno-cognitive fixes you might try to make that happen a bit more, but it is difficult and then the people outside the filter bubble did not realise that the filter bubble was happening and then Brexit happened.

This is the problem and the mainstream does not know what an extreme position is doing and holding and how much influence they might be inserting back into the mainstream surreptitiously. It is a real problem. As to what happens to objectivity and truth, I think you can strive for impartiality. Objectivity is a difficult one. In objectivity you represent all views equally. It is not possible in a small-time format such as a news story; that has never been possible. Impartiality is possible where you have at least two sides and you try to give them equal weight across time, but we are in a much more complex universe now. There are not just two sides.

Dr Tagg: We think part of the solution has to be education, and we would point to the ideal position of higher education in addressing that. Part of the issue, as you suggest, is that the world is changing so much and so quickly and the mediascape is so different from what it was just a few years ago.

While people are aware that newspapers have a particular bias and have some sort of media literacy I think it is taking time for people to realise how social media works and how information is now circulated in society and what impact their actions might have. That alone does not solve the problem of fake news but it has to be part of it, I think, that people are helped to have more awareness of how things operate now. I think that needs to involve higher education, which can arguably have a more immediate impact than schools, because its outreach programmes can impact people who are consuming news today, but it also has to involve social media platforms and journalism and we would argue applied linguists and communication specialists.

Q224 Brendan O'Hara: How does fake news affect the way in which people react to expert opinion? Expert opinion seems to be so utterly devalued in our society, and I know from my own personal experience of teenage daughters who spout some absolute nonsense they have heard at school and you say, “No, that is not true and this is why it is not true” and they say, “No, no, it is true and I believe it to be true.” How do we as a society overcome that?

Professor Bakir: With a multi-faceted approach; there is no silver bullet. It is a really hard problem and I think every person who is involved in that media ecology needs to take responsibility and do something. It is education at all levels, but it is also political leaders, Trump and others, who spout ridiculous bullshit, for want of a better word, because they do not believe in whether it is true or false. They spout fake news to delegitimise anything that they disagree with, and then their followers follow that. Political leaders have a huge role to play here as well. They can convince large sections of their followers that something is right or wrong, that the world is flat or round or that Putin is a great guy or not.

Professor Lewandowsky: I have some data on that, which I have found absolutely fascinating. That is the approval rating, or the positivity ratings of Vladimir Putin among the American public. What has happened with Republicans is that during the last two years that has risen from 12% to more than 30%. There is a 20% increase in Republicans’ empathy or liking of Vladimir Putin over the last two years. Democrats have remained flat. To me that 20-percentage point jump is incredible because

I am old enough to remember the Cold War and how Republicans felt about the Russians and all that, and all of a sudden there is a 20% jump. I think it is pretty obvious, we all know who or what caused that, and the lesson I draw from that is that leadership can be incredibly influential. If Donald Trump can come along and pull the Republicans 20% up in liking Vladimir Putin all by himself, effectively, you may not like that or you may be concerned about it, but that shows the power of leadership. We cannot walk away from how important leaders are.

One good thing about leaders is there is some research by a colleague of mine at the University of Exeter who ran an experiment in the United States with state legislators who were randomly assigned to two conditions in the lead-up to state elections across the US. A random half of those legislators received a letter telling them, “Your reputation will suffer if fact checkers find that you are inaccurate during the forthcoming election campaign” and the other half of the legislators did not get a letter or got some letter just telling them nothing. It was a control condition. Then they were all monitored and their public utterances were subjected to fact checking and guess what happened?

The legislators who were warned about their reputational damage if they told an untruth ended up being far more accurate than the people who were not warned. In other words, politicians were shown to be sensitive to the reputational cost that they suffer if they tell porkies. To my mind, that is also good news, because it tells us that political leaders are sensitive to being held accountable; at least they were when the study was done in 2015.

Q225 Brendan O'Hara: What happens now? We are in a situation where many of the political leaders benefit from the propagation of fake news.

Professor Lewandowsky: That is the problem.

Professor Bakir: You need to regulate that. That is a problematic thing; how do you regulate political speech? When it is outright deceptive, yes, you have freedom of expression and it is very important, you have to protect political speech, but outright deception, there must be a mechanism of accountability and if there is not one you damn well should find one. This seems pretty important.

Q226 Chair: That is why we are all here. It is testament to the quality of the evidence that the longer it goes on more and more people want to come in with questions. What I might do in fairness to Committee members is bring in one or two who have not yet asked a question and then we will see if we can wrap up with any loose ends at the end. Julie Elliott and then Jo Stevens.

Q227 Julie Elliott: I am finding this session fascinating. I want to come back on something you said there about the movement of Republicans to having a more positive view of Putin. Has any research been done on which Republicans are thinking that, because there is this new kind of army of Republican voters who 20 years ago, 10 years ago, were not Republicans and then there was what I would call the Republicans’ traditional supporter base, who are not particularly enamoured with their president? Has there been any breakdown on that?

Professor Lewandowsky: Excellent question. On this particular question I do not know, however generally you are absolutely right, the Tea Party, which is what I would call that new phenomenon, the Tea Party members and conventional Republicans tend to differ on many issues and a lot of pollsters and political scientists in the US are now thinking that really you have to talk about a three-party system. You have the Democrats, you have the Republicans and then you have the Tea Party, which is bolted on to the Republican Party but it is very distinct. On many issues they are quite far apart.

Q228 Julie Elliott: I think it is wider than the Tea Party, because the Tea Party have been around for a decade but there is this group that voted Republican in the 2016 presidential election who had never aligned themselves to the Republicans and who I would say probably on many social issues are not Tea Party people. They are very white, working class, often unemployed. They are not Tea Party people; they are different people but they are very definitely aligned to Trump’s Republicans, but has nothing been done on that breakdown?

Professor Lewandowsky: I should be careful here, because if you look at the election results in detail then it is not the case that poor, working class people voted for Donald Trump. In fact, they voted for Hillary Clinton.

Julie Elliott: They did in some states.

Professor Lewandowsky: Exactly. It is very nuanced and there are exceptions to this but if you look at the data, this generalisation that it was poor, unemployed, frustrated people who lost out who voted for Trump no, that is not the case.

Q229 Julie Elliott: It is in some key states. It won the election.

Professor Lewandowsky: In some key states, if they were white. If they were white, okay. So it is very nuanced. The moment you take out race and just look at income in the aggregate and social status then it totally flips around and all these people voted for Hillary Clinton, but there is an underlying variable of race that splits them apart.

Q230 Julie Elliott: But has nothing been done on that?

Professor Lewandowsky: There is work on that. That election has been analysed very thoroughly.

Q231 Julie Elliott: No, I mean in relation to your issue about Putin. That is quite worrying.

Professor Lewandowsky: Putin? No.

Q232 Julie Elliott: Professor Bakir, I want to come back on something you said. This influencing of news and people believing leaders—that was a huge issue in the referendum campaign when you were talking to people on the doorstep. They simply believed what certain individuals had said and it was whichever mainstream political party you were from. You might be a Leave supporter or a Remain supporter and you knew that factually this was nonsense, but it was impossible to break through that once people believed it. What do you think can be done? That is not just about leadership. It is a bit wider than that. Can you think of anything that can be done to penetrate that kind of thing? Certainly for somebody who has been involved in political campaigning for almost 40 years this is very new in the last two or three years.

Professor Bakir: I think we would need to research it. I do not know what can be done off the top of my head. There is a disconnect between what people do and what people say as leaders and then what people think they do and say and why people vote. Identity comes into this again, political identity. I do not know off the top of my head. I really think it is worthy of research, though.

Q233 Jo Stevens: I agree with Julie Elliott. It has been absolutely fascinating but incredibly depressing. The more I hear, the more I think Facebook seems like an uncontrollable monster that is sucked into our lives. I am interested to know if there was one thing that you thought could be done not just in relation to Facebook but to mitigate the negative effects of fake news. If you had a magic wand and could do one thing what would it be? Any or all of you?

Dr Tagg: Without the magic wand, we have talked already about education. We do feel that more qualitative, ethnographic research needs to be done. It has come up quite a few times where we have gone, “What? People are doing that?” and I think the more quantitative research needs to be done too, but there is space for doing research that tries to understand exactly how people are interacting on social media sites and how they are consuming needs and how they are reaching these decisions, alongside the general patterns, to get a sense of what is going on: why people are doing what they are doing, which does seem to us to be difficult to understand. That research is really necessary to feed into solutions. Without understanding why someone would post that, why they would believe that, I think that needs an ethnographic, in-depth investigation.

Professor Lewandowsky: I do not disagree with that at all, but I want to add another component. That is that as legislators you do have the tools to regulate those giants, Facebook, Google, Twitter, whatever, to do things a certain way. We know that can work because I believe one in six of Facebook’s moderators now works in Germany. They have to, because the Germans passed hate speech legislation that requires Facebook to remove offending content within 24 hours.

Guess what? Facebook hired the people to make that happen. So it is possible for a regulation or a law to be passed that is telling those IT giants how to behave. Now, I am not saying they should be censored, but I think we should not shy away from this and say, “Oh, my God, Facebook is a monster.” Well, yes, maybe, but it is ours to shape. Legislation can intervene. We have precedents for that, in Germany for example. We should not lose track of that: that there are powerful tools available.

Professor Bakir: Fake news really took us by surprise, didn’t it? The thing about fake news is it is going to continue to morph and it will continue to take us by surprise as technology changes and as people adapt to that technology in unpredictable ways. I think what you should have is a monitoring group, a working group that is monitoring what is going forward with fake news, what is the development of fake news. Of the many solutions that many countries as well as different technology companies around the world are enacting, see which ones seem to be working. I would say over the next couple of years get a working group, get your academics on board and other parties where possible and monitor the hell out of this thing and keep an eye on it, and see what is around the corner. What is around the corner may be much more worrying than what we have experienced to date.

Q234 Jo Stevens: Can we afford to wait two years to do that?

Professor Bakir: We can do regular updates. You could have monthly reports or something from this group. It could be open to the public and the world to comment on, rationally and civilly.

Professor Lewandowsky: I agree with that as well, but I also should maybe point out that I was in Italy last week at a fake news workshop organised by the European Union; they have precisely that monitoring group and they have that idea that Professor Bakir has just proposed. That is in place. That is being built. They are working on this. They created a division, EU vs Disinformation, I think it is called, that is dealing explicitly with Russian propaganda with respect to Ukraine so there are a lot of European initiatives out there.

Professor Bakir: It sounds like they are targeted towards the Russian security threat.

Professor Lewandowsky: This particular outfit is, yes. This particular outfit is, but I think they take a broader approach overall. There is some fantastic work done by one of their researchers on the French election, which was the first western election held after Brexit and Trump where we knew what was going to happen ahead of time.

It was absolutely fascinating, the research they did, on how Macron was attacked with fake news and what the sources were and who retweeted them and how the whole thing was done. That was absolutely masterful and I think that made me a little more optimistic, because I could see there were some very strategic people doing incredibly good research on precisely how elections are being undermined by propaganda efforts and how they can be countered, because Macron countered all that stuff very well.

Q235 Paul Farrelly: I think education is important. We do not have to go into the fourth pillar of education to realise that it is important that children should be taught how to ask the right question or to consider how to ask the right question and why they should care. If I went to a school as a former journalist to do that, I would love to do it, but I can imagine it would cause some problems for the usual bog standard run of the mill religious studies period. I wanted to ask one final question. Your categorisation of “filter bubble” pretty much sounds to me like your ordinary run-of-the-mill devout church congregation, so why should we be concerned more by one than the other?

Professor Bakir: I suppose some deny the existence of facts and truth and some do not. Some have perhaps more democratically problematic, scientifically problematic beliefs than others. That would be one way of putting it. I will not speak to the religious comment. It is far too inflammatory.

Dr Tagg: Technology is a complicating factor in that it gives views greater reach, greater visibility, gets picked up by the media very easily and quoted and also is still not well understood; the research has not fully caught up with the changes. In some respects that is all problematic. I guess in some way we should not pick out social media as being very distinct from other social activities.

As well as schools being important, we also feel there is a role that higher education can play, as well as teaching children at an early age, in trying to reach out to people who are currently accessing their news through social media. So the Open University, for example, where we both work, has a platform, OpenLearn, for reaching out to the general public and that is part of the Open University’s mission.

We have been working on some open educational resources, some animated videos, designed to raise awareness among the general public as well as for use among our students. Obviously, that is a big body of people as well. That is just one thing and the Open University is doing other things as well, through the library and so on, and we feel that platforms like that are also very important. That was not answering your questions.

Chair: I think we have one final question from Ian Lucas, then we need to close.

Q236 Ian C. Lucas: One specific brief point. Is there any evidence, particularly with Facebook, that the people who use Facebook are less involved in the community: in how we see the community traditionally? For example, involved in community groups or actively involved in their community? Is there any evidence of that level of involvement of people who do Facebook?

Professor Lewandowsky: None that I know of. It might exist, I can look for it, but at the moment I do not know.

Dr Tagg: A lot of the research that I am familiar with is almost the opposite of that. It is increasingly looking at how social media is a tool that people use in their everyday lives, so people are likely to be part of communities, taking photos and putting them up on Facebook. We all know how embedded social media is now in offline activities and social activities, because we see people using their phones during them.

Q237 Ian C. Lucas: What is striking to me as a Member of Parliament in the community that I represent is how a lot of the people who communicate with me on Facebook, first, I do not know and secondly do not operate through other community organisations. I guess I am displaying my prejudices but my idea is that those people are active in the Facebook community but not active in their real community.

Professor Lewandowsky: I will follow up on that and send you some information. It is a very interesting question.

Dr Tagg: Certainly there is the idea that people operate in different ways on Facebook, so they operate in networks rather than communities. We have been talking a lot about communities in groups, whereas Facebook and other social media sites encourage you to have individual links with people and have your own network, which can operate separately from the idea of a community. So you are linked to relatives in America and a working group that lives here and so you have a different way of interacting that operates separately from communities and groups.

Chair: That is great. Thank you very much. It has been a really fascinating evidence session this morning. Thank you.

Examination of Witnesses

Witnesses: Dr Charles Kriel, Adam Hildreth and Matt Breen.

Q238 Chair: Good afternoon, and thank you for joining us for the second panel in this evidence session this morning. I am sorry we have run slightly longer than we anticipated, so thank you for your patience. I think you were all there pretty much for the previous session as well.

If I could start off, and I would welcome the views of all members of the panel on this, we have heard quite a lot so far about understanding people’s behaviour online, particularly in the social media space and how through understanding that you can help organisations to target people with particular messages they might find believable or credible because of the way in which fake news works as a tool for media distribution.

We have not heard quite so much so far about the data analytics part of it. Do people release campaigns with the hope that they reach an audience through viral connection? That could be one way of doing it, although pretty hit and miss, but beneath all of this seems to lie pretty sophisticated data profiling techniques. It would be interesting to hear your views based on your different experiences in the organisations you work with, looking at how people go about building up the profiles of the people they seek to target online and how effective do you think those tools are? Perhaps Dr Kriel would like to start us off.

Dr Kriel: Thank you very much, and thanks for having me along today. It is a really vital issue, because at the moment Facebook and Twitter monopolise online public discourse, and so to operate in this way within this environment has an incredible influence over our democracy and it is a huge threat to our democracy as well.

There are three pillars of fake news in general, I think, for operators who are trying to conduct influence campaigns. One is behavioural science and behavioural economics, which operates across many platforms. The other is data analytics, natural language processing, being able to analyse sentiment in terms of conversations with communities online or among communities online and then highly addressable advertising technology. How do we hone in and target specific groups of online users in order to influence their opinions, their actions and often, even more important, to discourage certain actions, like going out to vote?

Q239 Chair: Facebook in particular is a closed system, so are people analysing externally millions of interactions that people have with the site and seeking to build up profiles and characteristics of certain types of people that they can then target using the tools that Facebook has created for advertisers? Is that basically how it works?

Dr Kriel: It is a closed system, but once you have bought into the advertising side of it there is a great deal of analytical and targeting information available to an operator within the space. It is not hyperbole to say that, in certain iterations of Facebook and its targeting tools, I could target you, if I knew enough about you, and reach you with a specific message.

Q240 Chair: Yes, but how would you target millions of people like me? When we are talking about elections or big campaigns, that is effectively what we are talking about; it is not just the micro-targeting of a very small number of people based on age or geographical location, but a whole range of people who hold different views.

Dr Kriel: There are a number of ways to do that. It is based around targeting people based on their interests, on what they have liked, on geographic data. There is demographic data as well that you can use to target people. The more sophisticated operations of companies like Cambridge Analytica or sometimes state actors, or actors who are supported by the state, will also use five-factor personality profiling on individuals and then group those individuals and target them specifically. So it is to look for people who are vulnerable to certain types of messaging, or receptive to certain types of messaging would be a gentler word.

Q241 Chair: Can you tell us a little bit more about five-factor profiling?

Dr Kriel: Five-factor profiling is a little bit like Myers Briggs, which uses fewer factors in order to determine a person’s personality. It is typological profiling. Five-factor uses openness, conscientiousness, extroversion, agreeableness and neuroticism in order to define a person’s personality. This is called the OCEAN system for each of those items that I have just named, and that is in combination with data analytics around what a person has liked. It was referred to in the last session that if you had a certain number of likes for a person, you could predict their behaviour better than perhaps their spouse or a co-worker might do, depending on the number of likes. Those two in combination are really powerful for targeting not only individuals but groups made up of those individuals with these personality traits.

Q242 Chair: You cited Cambridge Analytica as a company that works in this way.

Dr Kriel: They are just one of many.

Q243 Chair: Yes. How would they bring this information together? Is the external profiling just based on external opinion research that is done, understanding characteristics and traits of people in a certain location, a certain country or city, and then applying that to data that can be derived from the platforms? How do you bring those elements together?

Dr Kriel: It is a combination of publicly available information, information that is on public records published by Governments, often, but then also getting access to the data of an individual or a number of individuals based on online techniques in Facebook and other social media platforms. There is a rather famous case of Cambridge Analytica and their associated companies acquiring a great deal of data about individuals, and I am trying to make this as short a story as I possibly can—

Q244 Chair: We are with you. Don’t worry.

Dr Kriel: Acquiring this data through an Amazon service called Mechanical Turk and Mechanical Turk lets me say, “I need survey data. Will you workers work on my survey data? I will pay you one penny for each piece of information that you give me” and so you try to give me as many pieces of information as you can.

So through Mechanical Turk a number of people were paid to take a survey, but to take that survey they needed to agree to like a particular Facebook page. In liking that Facebook page they not only exposed all of their own data, but they exposed all of the data of all of their friends online, and when Amazon found out about this of course they immediately cancelled the account as it was against their terms and services but through that was acquired the information that became the baseline that Cambridge Analytica now claims to hold where they have 5,000 data points on 220 million American citizens.

Q245 Chair: So that is probably the biggest data set outside of Facebook?

Dr Kriel: It is pretty phenomenal, yes. I am not sure that it would be the biggest data set outside of Facebook, because Twitter would hold data as well and has far more users than that, but the claim is that it is on every voting American.

Q246 Chair: Yes, but I suppose outside of a big tech company, in terms of the platform owner, it is the biggest data source anyone else would have.

Dr Kriel: It would seem so.

Q247 Paul Farrelly: Just on these sources of data, I have never had a store loyalty card because I do not want the likes of Experian following me around.

Dr Kriel: Understood.

Q248 Paul Farrelly: Political parties now use Experian because they do not have the bodies across the patch anymore as comprehensive canvassing bases, so firms in particular like Experian use these categorisations and sell to the Labour Party and the Conservative Party. They are not choosy. What role does sales of data by firms like that play? Experian as one example.

Dr Kriel: That is outside the parameters of my research, but a number of players who have data are selling it to people conducting influence campaigns and electoral campaigns as well. I would like to give you more detail on that but I cannot in terms of Experian.

Q249 Rebecca Pow: I am touching on something you mentioned. I know, Dr Kriel, you focus a lot on behavioural economics. Can you expand a bit on how that is being used and what evidence there might be that that is being used—for example, people’s emotions and all those kinds of things, to effect misinformation and fake news, ultimately?

Dr Kriel: There are a couple of points that I would like to address on that, although I do not want to dominate the panel. The first one would be, and this carries over from the earlier panel, that it is very important to distinguish between misinformation and disinformation. There is a great deal of intentionality in disinformation and I think some of the scope of this panel would be very concerned with disinformation, especially with malicious actors, for want of a better way to put it. There was an illustration that I think you all have in front of you. That is more or less about how Facebook works and it is very much based on a behavioural economics perspective on how Facebook might work. I will see if I can remember it in my head, as I do not have it in front of me.

Facebook works around a target, an action, an intermittent variable reward and an investment. What this does is create what is called a compulsion loop. This is a very addictive pattern and it is addiction by design—it was not stumbled upon accidentally—as a model for the way that online platforms work. It is used across the gaming industry but as well in social media platforms.

A trigger might be, for example, I am walking down the street and I am feeling lonely, or I am feeling an emotion that I do not want to feel at that moment. So I would take out my phone and I would look at, for example, Facebook. The reward I would get could be almost anything. One reward might be someone has posted something that I find myself interested in. So then I make an investment and that investment contains the seed of the next trigger.

My investment is I write a pithy comment about their post and then I put my phone back in my pocket and I walk about 10 more steps and I get my phone back out because I would like to know if anybody liked my pithy comment that I made about that post. So I am in this loop but it is not addictive yet. What makes it addictive is the intermittent and variable qualities of the reward because the reward that I get could be anything from something negative—someone has disliked my comment—all the way through to that person I fancy a lot has just friend-requested me and I am really very happy about that.

So in this range of rewards I do not know when it is coming and I do not know how good it is going to be, but I know when it is good it is going to be absolutely amazing, and this is what makes the cycle so incredibly addictive.

Q250 Rebecca Pow: Excuse my complete ignorance, but did the brains behind Facebook know about this model, or did it happen by accident that it has become addictive in this way?

Dr Kriel: It has not happened by accident at all. These are common practices. I am not going to use this to plug my book but I have written a book about using behavioural economics in gaming-platform design, which I think I have also submitted to the Committee to have a look at if you wish. These are tropes. They are regularly used and a part of the industry. It is as common for Facebook as it is for Candy Crush Saga, for example.

Q251 Rebecca Pow: Taking this, obviously you can get people to become addicted, so how dangerous is that for then taking them and the knowledge about them and using that to manipulate them, which is a lot of the things we are talking about, say in the American election or our referendum? How easy is that, to go from one step to the next?

Dr Kriel: Dangerous enough to destroy the fundamental structure of Europe.

Q252 Rebecca Pow: Do you really think that?

Dr Kriel: Dangerous enough to influence the directions of countries and politics. I have just written an article for NATO, NATO’s strategic communications journal, about the role of this in the EU referendum and in the US election, but principally in the EU referendum. There was a great deal of influence campaigning, and that is fine, because influence campaigning is what we do when you try to win elections. We do it here and we do it abroad, but it was there and there was a lot of confirmation bias happening.

What was really interesting earlier was when you asked for recommendations about how to convince people of the truth. What Remain was not very good at doing was telling an emotional story that was compelling, but Leave was very good at it. Just looking at it analytically without taking any sides, Leave told a powerful emotional story and as opposed to that, using logic and facts to try to tell stories is really very difficult.

Q253 Rebecca Pow: Is emotion stronger when you are trying to influence people than the facts?

Dr Kriel: Emotion is far stronger. A good example would be that doing work in Donetsk in Ukraine we found that Russian propaganda was spreading the story that if a Ukrainian national soldier took a house they could have the women as sex slaves and they would crucify any infants in the house. That is a very good and very viral story. That is going to be spread around. I would spread that around. I would be shocked and appalled if I thought that was true. A good civil society journalist in trying to tell the truth about that story could say, “No, they don’t” and that is not really a powerful emotional story. So it is a challenge for the civil society and supporters of the civil society of how we tell these fundamental truths about our basic beliefs and have them compete in a viral marketplace of information.

Q254 Rebecca Pow: Is it too late for us to deal with this?

Dr Kriel: No, not at all. I think that we are in the infancy of this problem, but it is a problem that has existed for hundreds of years. It is propaganda just distributed computationally.

Q255 Chair: Thank you. There are quite a few people wanting to come in, and I want to bring in Matt and Adam as well. Do we have a copy of the article you have written for NATO?

Dr Kriel: You do have a copy, but I can make sure that it is sent over to you again.

Chair: That is great. Thank you very much. Just briefly, Paul and Ian.

Q256 Paul Farrelly: On your article and your research, I have said this a number of times: during the referendum campaign you could have a decent conversation with a middle ground of people who you would have sworn, because you know them, were going to vote Remain but with their heads and thinking about jobs, the kids, and investment. In the last three weeks that changed, as soon as the big red banner posters, “Turkey population 76 million is joining the EU. Vote Leave”—and it had the same sort of effect in changing the questions you were asked, because people just said, “Well, what are you going to do about the Turks?”

It was like what the Conservative posters did in the 2015 election showing Ed Miliband in the pocket of Alex Salmond, because then the question became, “What are you going to do about the Scottish Nationals?” and it changed the terms. You could not talk to people. If there were any similar examples, and I did not see them on social media either, of adverts that were running like that and any data on their audience reach and the age profile of people seeing them I would be very interested. I am in complete ignorance because I did not use social media in the election.

Dr Kriel: May I just take 60 seconds to address that? I do not want to be rude. One of the problems with these adverts on social media is the bias is usually what is called “dark ads”. Dark ads allow highly specific targeting in Facebook; those who are not targeted cannot see the ads and will have no awareness that they are there. So as long as online political advertising is not subject to the same scrutiny as other types of political advertising then we will not be able to see them, we will not be able to gather the research around them. I specifically haven’t but I am certain there must be some who have. I know The Guardian had a project for a while to ask people to submit ads that they saw on Facebook around the general election.

Q257 Ian C. Lucas: Propaganda is not new, we all know that, but what seems to me to be new, from what your evidence is, is the combination of propaganda and addiction, because propaganda and addiction together mean that people who are being addressed by the propaganda will continue to be addressed again and again and will go back to it. It seems to me that that is novel and that is what is novel about this aspect of social media. Do you agree with that?

Dr Kriel: I agree with that and I would add the factor of the ability to highly target people according to how receptive they are to your message. It is a common campaign technique to give canvassers on the streets an app that door by door will tell you how receptive this person is to your message.

Ian C. Lucas: We know.

Dr Kriel: Yes, I am sure. It suggests questions that you can ask them. While it is very difficult to change a person’s mind in these environments, it is very doable to either get somebody to do something or discourage them from doing something. This works across the range of discouraging someone to go out to vote all the way to encouraging someone to strap a bomb on their back.

Q258 Ian C. Lucas: What we have never been able to do before is get the canvassed people to come to us, and the addiction enables that to happen.

Dr Kriel: Indeed. Indeed, I would agree.

Q259 Giles Watling: I think I am going to neatly be moving on to Mr Hildreth in a moment so this is quite good. I speak as someone who does not use Facebook. I am naturally reticent about giving out information about myself so when anybody asks for any profile I just do not do it if I can avoid it at all. I always object when banks seem to want to know my inside leg measurement or something; it is just nonsense.

People are desperately trying to trawl for information the whole time. That is what I see and that is how I react. Because this information comes from the Facebook user, should we be teaching people some sort of reticence in some way so that they are more resilient? Is that a way forward? Is that a way of perhaps educating people when they are young, which is why I am going to come on to you in a moment, Mr Hildreth?

Dr Kriel: Yes, that is absolutely true. Critical thinking skills around this are really important, and around one’s data privacy. I also believe that users should own their own data and that they should legally be able to rescind the use of that data. Although I do not believe in legislation that limits or potentially limits free speech or access to information, because this is a fundament of our civil society, there are a lot of legislative options including limiting the use of individual’s data or limiting the length of EULAs—end user licence agreements, the things that everyone clicks agree to without reading—and limiting the automatic opt-in to the next end user licence agreement when you log into online services.

Q260 Giles Watling: So that is legislative and that is what—

Dr Kriel: Yes, I do think there are many legislative approaches.

Q261 Giles Watling: Moving on quickly to Mr Hildreth. You founded a company when you were, I think, 14.

Adam Hildreth: Long ago.

Giles Watling: Long ago, yes. It was essentially to protect people, like yourself. Online security is what it was all about.

Adam Hildreth: Yes, I think that having listened to everything so far it is probably interesting why we founded Crisp in the first place, which was to focus on what we call bad actors. So fake information, fake news, whatever we want to refer to it as, has been generated by someone for whatever means. That person is out there to exploit, they are out there to defraud, whatever words we want to use to describe them.

Our view is that we should be going after the bad actors. Bad actors happen in everyday life, whether it is e-mail scams that go on, whether it is trying to convince someone to sign up to something they shouldn’t, whether it is fraudulent sales, we see fake news as just another methodology, although a very advanced methodology, of that. Our view is not to focus on why do people click on fake news because it is really easy to be exploited, I would guess everyone in this room has been exploited one time during their life, whether it is by clicking on an e-mail that they should not have done, even clicking on an actual fake news article. It happens all the time.

Our view is that our focus should be on the people that are behind it and identifying those accounts. It is very interesting that you mention behaviour profiling. Behaviour profiling really, until companies like ourselves, is very much used in an advertising sense. You are seeing it from how do you profile someone you want to target. We see behavioural profiling as how do we find the people that are behind the fake news in the first place. Ultimately, whether it is child exploitation, fake news distribution, cyber criminals, only 1%—a tiny, tiny minority—of people using the social web are out there for that, but they are causing 99%, if not 100%, of the issues there.

So I think concentrating on the people behind this and techniques and helping the social platforms find those people is what is critical. If you do not find the people, they are just going to create new methodologies, new techniques, new ways of manipulating end individuals to get to where they want to be.

From my perspective, we founded Crisp to deal with child sexual exploitation 12 years ago. We did incredibly well at that and have since moved on to other areas. Our latest one is fake news. We are not fully there yet, but our belief is that we have to go after the people that are behind it, not the individual pieces of fake news.

Q262 Giles Watling: Can I turn that on its head then and say instead of finding the actors, the bad actors as you call them, Crisp spots users who are at risk, does it not?

Adam Hildreth: Sort of. Only by looking at the people that are targeting them. So by identifying the bad actors we can then do things to intervene when those bad actors are targeting people. That is our view that concentrating—because that is who the war is really with. This is a war in terms of there are people out, albeit a tiny minority, who are extremely sophisticated, extremely good at what they do but are sat there right now while we got teams of people, and Facebook have teams of people and so has Twitter, working out how to stop them.

Those people are working out how to deliver their next bit of fake news that is going to get through filters, get through the mechanisms that are there. Any help about how we identify accounts that are out there that are essentially fraudulent exploitative, for multiple purposes, is the critical bit because once you understand the people that are out there doing it, you can then intervene to maybe stop the information being sent.

Q263 Giles Watling: How do you do that? What is your next move? How do you then intervene?

Adam Hildreth: That is hard at the moment. That is the bit that we do not have right yet. How do you intervene? You have this whole freedom of speech issue. Who is there to decide that that person is a bad actor? Is it a general consensus that this makes you a bad actor when it comes to fake news distribution or disinformation or whatever it maybe? Or is it the social platforms individually do that?

We have the technology now that will very precisely identify the people that fake news originates from. Those people create fake accounts, they are stealing other people’s accounts, they are buying dark ads, these are bad people. It is not really a social platform issue in the sense it is an entire community issue about how we stop these people from exploiting that.

Q264 Chair: Without asking you to give away trade secrets, how do you do that? Obviously you are analysing Facebook externally, because you do not have access to Facebook’s data. How are you able to identify likely sources of fake news by analysing the site externally?

Adam Hildreth: We license our technology to social platforms, which we will not mention in terms of the ones that are there, and when it comes to fake news we have spent a year, 18 months, looking at how we would approach it. Because of the freedom of speech issue, because social media is all about sharing, we have to be really careful as to how that is implemented. But ultimately you need to profile people, the same way that people have done for advertising reasons.

If we can get to the point where literally you can pinpoint individuals based on their unique interests, we can absolutely pinpoint the bad people that are behind it, whether it is fake news, child exploitation, whether it is scams or reward schemes. You can do that but you do need access to the data. This is a problem that the social platforms want to fix. It is just incredibly hard because there is not a single point solution where you can go, “We have fixed it”. The next day, wherever these criminals, exploiters may be, they are going to be finding out mechanisms to go and combat it.

Q265 Chair: Hypothetically, if Facebook said we want to understand who on the site is involved in child exploitation, by analysing the data you could help them predict where those bad actors are likely to be found?

Adam Hildreth: Absolutely. We have been doing that for 12 years and the social platforms are also very good at it. Whether they use our technology, their own, whether they use a hybrid, they are doing a lot but there is still a lot more to do because the reason why it is so hard to solve this problem is normally when we are looking at AI we try and help it. Whether you have Alexa or anything at home, you are trying to tell it what you want. When we deal with AI to try and combat bad actors, to go after the bad people, they are trying to avoid detection just like criminals do in the real world, they are trying to avoid it, so it is a difficult challenge to solve but we are getting there and so are the social platforms.

The critical bit from my perspective is they want to solve this problem because trust and safety is critical for them. Without trust they will lose their user base. Without having a safe environment they will lose their user base.

Q266 Chair: One of the issues the Committee has taken an interest in is the role of Russian agencies in propagating disinformation and fake news. From what you are saying it should be relatively easy to—well, I say relatively easy but it is technically possible to analyse the data to identify likely activity by Russian agencies on the site, based on the characteristics of what we now they have been doing so far?

Adam Hildreth: Yes, whether you could get to the point of knowing who was actually originally behind it, but by identifying the activities and their behaviour, yes. There are very advanced ways of doing it, it is just this is a different type of risk that we are now facing.

Q267 Jo Stevens: I just want to ask about mitigation and also what we talked about in the first panel, which was around digital literacy and educating people about how to use platforms. We had a conversation after the first session about it does not matter what your age is, you can be 86, but if you are using a device for the first time, you are using a platform for the first time, what education do you need in order to use it safely and effectively? I am interested in your views about how that can be addressed, if you think it should be addressed but also then about mitigating the negative effect of fake news and what do you think is the best route to that?

Adam Hildreth: My view is that educating, whether it is children, young people, adults, all of us, about understanding what the source of the information is. The big issue—and it was talked about on the first panel—is that social media is people that you inevitably trust, or you think you trust. They are sharing something from somebody that they think they trust but very, very quickly that can spread like wildfire but everyone believes because it has been shared by a friend, shared by someone they trust, that they should trust that information.

We need to get to a point where you understand where the information originated from and we can understand if it was “BBC News” I probably can start to trust that; if it is some random website somewhere that I have never heard of and no one else has heard of, then I probably shouldn’t. I think the education should be on how you can use tools, how you can use mechanisms for understanding who created it. So that is very different.

We used to buy a newspaper or have a certain set of TV channels that have very strict standards so there was a certain level of trust already, depending on which newspaper you read. Online we do not have that, so there is going to be a series of technologies and mechanisms where we can get to that trusted source. That is the education bit, understanding where it came from.

Matt Breen: I completely agree. I have looked at some of the stats that Ofcom released at the end of 2017 and it said that 41% of over-75s now have a social media account of some sort. That is about four in 10 over-75s and it is the best part of 50% over 65, so the education needs to be throughout the whole spectrum, not just from young adults, millennials and even younger, who are spending about five or six hours a day on social platforms all the way through the older generations.

It is very much about education. It is very much about also looking at the source of the news, but I do say that with some trepidation because we all know that fake news can come from a newspaper as well as from a new media publisher. There are many times where newspapers are forced to retract news that simply is not true because they wanted to sell papers, they wanted to get front of mind on the top of the newsstand. It is very much about making news sources accountable for the news that they produce, that would be a key part for me, as well as educating people throughout to understand what indicators they need to look out for that are likely to be a high risk of fake news.

One of those is certainly sensationalism. There are a lot of posts on Facebook, on Twitter and on other social platforms that are encouraging people to click through because the headline is almost beyond belief. There is very much a common sense approach that needs to be shared and encouraged where to say if something is too good to be true or if a headline seems too sensationalist, then it probably is.

I think there is something that can be done where platforms identify the headline in itself and then the webpage that it is clicking through to. There are many instances where a headline will be sensationalist but then will not reflect the website that you land on, the reason for that being that the more people that land on a website, the more advertising revenue it generates. There is certainly a job to be done to identify the landing page versus what is posted on the social platform itself to make sure there is some consistency between those.

Q268 Jo Stevens: Just picking up on the point that you said there about print media and printing corrections, we know that the practice is that the headline could appear on the main page but the correction is a tiny, little paragraph on page 26. Is there any reason to think that it would work any differently with social media and that its impact would be any different? From my perspective, when corrections are published it is too late then. You might reach some of the people that read the stuff in the first place but you will never get to all of them.

Matt Breen: Yes, and I think that is one of the most interesting parts about fake news. We talk about the hypertargeting, we talk about dark ads, but in reality fake news spreads and it becomes viral. The reason we are all talking about it today is that it goes beyond micro targeting. It goes beyond a specific location, a specific demographic. It spreads wider than that.

I think what we need to do, to your point specifically, is try to catch it at its earliest possible point. You take Twitter and Facebook, both social platforms but they work very differently. Twitter is very much in the real time. Things trend instantly and things will be trending for maybe an hour or two hours but then it will come down as something else trends. Therefore, you have to be extremely reactive in understanding what is trending right now and is that a source of reliable news. Whereas Facebook, because people do go multiple times a day and the news feed changes often, there could be a story that has been running for 16 hours that only then catches fire because there are a few larger accounts that share it.

Probably to Adam’s point as well, it is very much about identifying the sources of these news, closing down those accounts, and identifying what those type of people look like and what they act like to identify the bad actors of tomorrow, of the day after.

Q269 Chair: How easy is it to game the news feeds, though, to try to ensure that your content is highly ranked and, therefore, more likely to gain virality?

Adam Hildreth: People have made entire careers out of doing that with search engines. It is incredibly easy to game AI and that was sort of my point before: getting AI so it understands good and bad is an incredibly difficult challenge, much harder than just creating it for the good. Because when people tell it what they want to find or you are being honest with it, it learns very well. What it does not understand is when you are not honest with it, and that is what the people who are exploiting these—it used to happen with search engines, search engine optimisation—it happens in exactly the same way now, whether it be fake accounts, retweeting or resharing a particular news article so it looks like people are reading it and doing that, or whether it is exploiting individuals to physically like it themselves. There are a lot of techniques and they do change every day, every hour.

Dr Kriel: I was just going to say that it was a common technique of Islamic State, in fact, who were quite accomplished operators in the social media sphere, to use false accounts and bots in order to spread their propaganda. It has been thought that there were more bots spreading their propaganda than, in fact, real people were at some points.

Matt Breen: Just to add to that as well, there are certainly ways—it is known as hacking—in which you can increase the likelihood that your post will be seen by your user base. One of those could be understanding what Facebook is looking to prioritise. At the beginning of last year when Facebook focused on live, we knew that live videos would be incredibly popular and Facebook would prioritise those within the algorithm so they are more likely to show to users. You then found some advertisers or industries where they would take a normal video and label it as live in order to make sure that more videos were seen, as an example. There are certainly hacks.

To the point about bots as well, you do get bot farms where you have thousands of phones that are all driven by a single computer to like and share. What I would say is that the platforms, in the conversations we have, are working every day to clamp down on fake bot accounts. Again, we talk about indicators for advertising and targeting. They have indicators for what is a bot account. They are getting better and better at closing those down, making sure that the IP addresses from which they originate are also restricted so that those cannot replicate in the same area, and that you are making it harder for those people who want to create bot accounts and who want to benefit from that and make it harder for them to start up again.

Dr Kriel: If I could make a point about bots as well, the Oxford Internet Institute has been running a computational propaganda project. They have looked at computational propaganda in a number of different countries. One of the things that they found is that Russian public discourse on Twitter is almost completely dominated by bots. This could happen anywhere, in fact.

Q270 Ian C. Lucas: What strikes me is that when you are talking about dealing with the problems that we are discussing, essentially the platforms themselves are the only means of dealing with them. You just said the platforms are doing what they can, and we are talking about issues that we all abhor, like Islamic State, for example—we all accept that those should not be allowed to propagate their policies online. Are we entirely in the present system in the hands of the platforms on this?

Adam Hildreth: I do not think so. If we take financial fraud as an equivalent, it is not a credit card processing company’s fault that someone steals credit cards or creates fake ones. What legislation has done and what Governments have done is helped the industry out, helped them to combat it, because they understand it is not their fault but both sides want to solve the problem. I very much see it as the same thing.

There are things that can be done that enable social platforms to find the bad actors faster. There are things that can be done that give them more information. There are lots of ways that I think it can be helped, but it has to be a joint approach, very similar to other industries.

This problem is not really a social platform problem. We spoke about propaganda earlier. It has happened for years and years. It is just the latest mechanism. If it was Royal Mail delivering propaganda, it would not be Royal Mail’s fault necessarily. They would try to do stuff to deal with it and we would probably try to help them with that. It is very similar to social platforms, which is a fantastic, brilliant thing that has been created over the last 10 years, but now those people that were exploiting the offline world are similarly exploiting the online world. It should be done in partnership, I think.

Matt Breen: Yes, I agree with that as well. Obviously, I will not be speaking on behalf of them, but where the role of the platforms differs to the role of Government is setting those guidelines. We have talked about fake news. We have talked about discommunication. We have talked about a number of different of meanings. The big challenge today is that the meaning of fake news is not consistent to every user.

We talked about confirmation bias and unfortunately with certain senior political leaders in some countries, when they have fake news awards and they are awarding the top award to The New York Times, I think most people around the world would probably debate that The New York Times is a reputable news source. The definition of fake news has elevated itself and it has become a parody.

The role certainly for Government, in my opinion, is to make sure that those guidelines are very clear to give some really clear facts and criteria for what is fake news, what should and should not be defined as that, and then working with the platforms to help them to shut down the things that we clearly believe are of no benefit to anyone.

Dr Kriel: I do not necessarily agree with everything that both of you have said, but I agree with most of it, because I err on the side of free speech and access to information, which is a fundament of the civil society. But I do think that you are not powerless in the face of these platforms that are dominating discourse, that have a monopoly over this discourse, precisely because they have a monopoly over this discourse. If we were talking about any other large, dominant player that cornered the size of the market that they have cornered, then there would be legislative approaches to that historically. I think that those options are also available now, but it requires a state-level address.

Q271 Paul Farrelly: It is a similar issue for search engines: to what extent should they take responsibility for pirate sites that they draw and show people to from the music industry and the film industry? To what extent would it be helpful or not or using a sledgehammer to crack a nut to classify, for instance, these platforms as publishers so we have legal remedies? Matt, I notice that you worked for the Mail Online.

Matt Breen: I did previously, yes.

Paul Farrelly: The Mail Online was very successful in its early days in expanding its reach because it collected all the stories that were fantastic, strange, gory and grotesque from around the world and had them all on one site. It has got a bit more sophisticated now. They were always taken aback years ago when I said it was every story, apart from that one, “World War II Bomb Found on the Moon”, which was demonstrably false. At least with the Mail Online you can sue them if they publish something that is false and libellous. To what extent should regulation and the law encroach to that degree on platforms?

Dr Kriel: I am not necessarily in favour of classifying these platforms as publishers. I think it is a debatable issue that I have not come to a conclusion on yet. I would note that Facebook in the past several days in response to fake news investigations has been saying that it will begin curating trusted news. If you are curating trusted news, then you are editorialising trusted news and you are moving dangerously, for them anyway, into the space of being a publisher. Whether it means it is time to declare them a publisher or not I do not know. Given the amount of information that is being published minute by minute by these platforms, it seems, even with the help of AI, nearly unmanageable and very whack-a-mole, for lack of a better metaphor.

However, if they were smaller organisations, then perhaps they could manage this approach to that data better. If Facebook, for example, is now editorialising, then could they not enter better relationships with the real news providers and real content providers that could be more supportive of traditional approaches to news and journalism rather than, in fact, charging them to distribute their stories?

Q272 Paul Farrelly: Where does the balance lie, Matt and Adam?

Matt Breen: Just to pick up a couple of points, I have obviously seen the same stories, but I think the Facebook curation that is coming up—and I think they have deliberately done this—is they are going to use a trustworthy measure. I am not sure if you have seen this, but they are going to ask their users to deem which sources of news are trustworthy. They will then give them a trusting score and that score will then impact the algorithm.

Dr Kriel: That cannot be gamed.

Chair: Lo and behold, RT might have been the most trusted news source in the world.

Matt Breen: Just to point out for the record as well, I worked for the commercial team at Mail Online, not the editorial team.

I think it is an interesting comparison, to come back to your point. That type of website is very much a standardised user experience. Every user sees the same story in the same order. If all of us in this room went on to that website, you would see the same story in the first position and the same story in the fifth position, because it is a very long home page.

Where other platforms have done so well, like Facebook and Twitter, and we also talk about other platforms like Sky because I am sure a number of people have set-top boxes at home, all of those are personalised experiences. They are then traditionally algorithm-driven so, therefore, your likes, your dislikes, and the stories you share will be promoted to you and delivered to you more often than you would otherwise see. For me, I think that is a consideration to think about as well. The personalised experience is going to be more of a challenge to police than a standardised experience.

Adam Hildreth: I think the publisher debate is a really interesting one because I have seen it from both sides. For 12 years we have been trying to keep the web safe, so trying to remove offensive commentary, toxic commentary, trolls, and we have a lot of broadcasters as customers. They know what they have to do. They have to moderate essentially 100% because they are so worried about the offline world of broadcast if they take that online.

We saw the complete opposite reaction, though, for numerous years where people would not consider themselves any kind of publisher so refused to moderate anything because they did not want to have the legislation say, “You are a publisher so we want you to do all of this” because it was either too costly or was not going to work. That was a disaster because that is how we ended up with so much horrendous hate speech and toxic content online: if you are not curating it and you are not moderating it, you are technically not responsible.

I do not think it is a full shift to go over to say, “You are all now publishers” because the good things that they are trying to do, whether it works or not, which is about trust ratings for news, they are trying to do things. Going all the way and saying everyone is a publisher where you are curating everything will have a negative effect, I think. That is certainly what it did a few years ago. It is just treading gently on that particular ground.

Chair: That is great. We probably could carry on all day, but that concludes our questions for this morning. I am extremely grateful for your evidence. Sadly, we have a statement on traditional media, on Fox-Sky, that we have to attend in the Commons Chamber. Thank you for that. Through the Clerks, it would be great if we could stay in touch during the course of the inquiry as well. We may have other questions we would like to follow up on. Thank you.