Transcript for the Background Discussion "Digital Disinformation – the New Default in Online Campaigning?"

- Start of Transcript -

Julian Jaursch: Good evening everyone and welcome to SNV. I am very happy to have you here and I’m very happy to have Sam Jeffers here next to me. I’m glad that there is a lot of interest in the topic that we’re going to discuss today. My name is Julian. I work here at SNV on a project that deals with tackling disinformation online, together with a couple of other colleagues. We cover a range of different aspect of this topic, but one of the issues that we are concerned with at the moment is political advertising online. The opportunities are clear but there also are challenges that come with it. And that's why I’m very happy that Sam is here. So just a quick rundown of what we’re going to do tonight.

I’m just going to briefly introduce Sam, introduce the topic and in the end, we’ll have a lot of time for questions from you guys that you can ask in English or in German.

We are recording this conversation because there will be a full transcript available on our website after this talk. Any questions from the audience will be made anonymous.

So, let’s jump right into it. It’s I think a very timely conversation that we’re having, as just this morning there was an independent UK watchdog report out that specifically deals with online targeting and had a big section on political advertising.

And that is exactly what I would like to talk to Sam about tonight. He has analyzed many online election campaigns worldwide: most recently the European Parliament elections in May 2019 and the UK general election, in December of last year. Tonight, we are going to discuss how political ads are used online, how hard or easy it is to figure out how they are used and what potential challenges might arise from that.

But before we jump into the discussion, I would like to briefly introduce Sam, so we get to know you a little bit. Sam is a political scientist, a musician, a consultant, an activist. Together with Louis Knight-Webb, you founded Who Targets Me, an organization that helps the public, researchers, journalists and academics to better understand how political ads work online. And we'll get to specifically what Who Targets Me does and how they do it in just a second. But first, is there anything else that you would like to add introducing yourself to these folks?

Sam Jeffers: Yeah, sure. The origin of the project is that I used to work for a company called Blue State Digital for eight years that used to run political online campaigns. Most notably, they ran the two Obama campaigns in 2008 and 2012. We worked on the Labour Party campaign in the UK in 2015.

I suppose it kind of helps to blend this into the origin story of why I do this. In 2015, the polls in the UK suggested that the election would be quite close and that there would probably be a hung parliament and that maybe there would be a coalition again. Instead, the Conservative Party won and like in all elections, winners get to write history. The history that they wanted to write was that they spent a million pounds on highly targeted Facebook ads in specific seats in the UK and that the media had not been able to cover this. We are certain that the other side of that campaign knew nothing about it. For this reason, they managed to surprise everybody.

That set the idea in my head, although at that time I was still working on campaigns. I left the company in 2016 and then the Trump election and the Brexit referendum happened. So, this began to become an issue and I thought, it began here and now it’s here and we need to understand how it works.

Julian Jaursch: That’s exactly what I wanted to start with, to just briefly lay the foundation for the conversation. The topic of political advertising online has gained some notoriety. You mentioned the 2015 election that you worked on, but it also was discussed very prominently in the Brexit referendum, in the 2016 US elections, where it was found that political ads by candidates or by parties tended to be fairly negative. A lot of the time they tend to demobilize voters. There were very specifically targeted messages that tested different images for different audiences. In the case of the US, there’s also proof that foreign governments paid for advertising to polarize society even more.

But it’s not just the US and UK. All major political parties in Germany use ads online. If you use Google and you search for a political topic or if you use Facebook or Twitter, you might see an ad from a political party. It is important to say that this is not necessarily a bad thing. You can reach out to voters and it's kind of an extension of the idea that you can talk to people and convince them like you do in other media.

Online ads have come under scrutiny because they are different from ads in traditional media, while still being governed by the rules that were made for traditional media.

But online ads have come under scrutiny because they are different from ads in traditional media, while still being governed by the rules that were made for traditional media, for TV, for papers, for radio. These differences are exactly what I want to talk about with you. So, suppose that I’m a political candidate or a political consultant, I want to tell people to go vote or I want to inform them about my policy agenda. In that situation, Google, YouTube or Facebook, for example, give me really good tools that I can use to target a specific audience and I can reach them with my message. What’s so bad about that?

Sam Jeffers: Generally speaking, nothing, if you are looking to reach people and you are transparent about your intent and your spending and the way that you are using data, so that people can get some kind of understanding of it. Or even if individually, they don’t have the tools to understand it, at least journalists or other people who hold those types of messages to account can do so.

Generally speaking, you can have a healthy political ad but it’s the cleverness that this kind of thing encourages. It means that people start doing the sorts of things you’re talking about, for example: How do we make the electorate the right size? How do we tell some people to stay at home, other people to come out, show people some messages about this, other people about this? How do we do that in such a way that no one gets a grip of what’s going on? How might we run 10,000 ads at the same time? There’s absolutely zero chance that any journalist is ever going to be able to go through them all and understand what you're really saying to who and why you're saying it. Once that kind of sophistication starts, you begin to get these problems.

The Conservatives in the UK recently ran extremely expensive YouTube takeovers for a whole day. The ads were showed to everyone in the country who went on YouTube.com or the YouTube app for two 24-hour periods and they cost about 150,000 to 200,000 pounds. There’s no massive problem with that. Everyone can see the ads, everyone can interrogate the content. Everyone understands what these ads were about: Get Brexit done. The message was very clear. There wasn’t anything particularly nefarious going on and people could get a good reading of that.

Yet, they were running 10,000 ads simultaneously on Facebook, targeting individual constituencies and different groups of people and really mixing up things. And that made interpretation and accountability much harder.

Julian Jaursch: You already mentioned transparency and accountability. I think that leads us over to the question of what Who Targets Me does. How did it come about and how do you address these transparency or accountability issues that you alluded to already?

Sam Jeffers: At the time when I left this company in 2016 and while we had this kind of surprising snap general election in the UK in 2017, Facebook offered no transparency at all. You couldn’t see what advertising was being run, who was being targeted. And we just came off the back of the Trump election, with lots of stories about Russian interference. Also, people were talking about data brokers and profiling organizations that have been used in campaigns. The only way that we could see to get some transparency during that campaign was to build some software. So, we built a browser extension for Firefox and for Chrome, which anyone can still install to this day.

The browser extension grabs the Facebook ads from your feed and matches them against the list of political advertisers that we’re maintaining. That includes candidates, parties, nonparty groups, and it gives you as an individual a kind of playback of the campaign. It will break them down depending on what grouping they fit into. You can click on the tool and you can see a list of the ads that you've been targeted with. You can see the reason if you click “Why am I seeing this?” on the top right corner of the Facebook ad. You can see that piece of information and that way you have something to refer back to you individually. As a voter you can understand what your campaign looked like.

But we are also crowd sourcing that data. They are used during the election campaign by journalists, who are trying to do real time coverage and explain what’s going on and hold it all to account. It’s a data set that they can use for research. It informed a bigger conversation about what we should actually do about all of this stuff. That was the origin of Who Targets Me. In the first instance, it was just about whether we can we get any data at all and about what we can we see. And then it developed into this three-pronged approach: with the individual aspect and the journalistic and research aspect and then the advocacy and policy aspect of it.

Julian Jaursch: So, if each of us installed the app, we get a benefit of individual information about the ads, but we also contribute to providing research and transparency on that topic.

Sam Jeffers: Lots of projects don’t provide any value for users and they don’t exemplify how things should be. We want to find out what’s going on before we push our hypothesis forward. So, I think that was an important part of it, because it gave people a reason to take part.

Julian Jaursch: And how many people are taking part? Maybe you have the numbers for the UK election?

Sam Jeffers: Yeah. So, back in 2017, about 15,000 people took part in the UK. Altogether, we’ve had about 50,000 people take part over the last couple of years. We had 5,000 people for the Polish election last year. We had 12,000 again for the UK election and we tend to get a few thousand with every country, where we focus on that particular campaign.

Julian Jaursch: Does it want to be representative? Because when considering the entire electorate, it’s still a relatively small number.

Sam Jeffers: We only collect ads on Facebook, right? So, is Facebook representative and are Facebook users representative? No, it’s not, and we don’t try to be. As I said, it’s a piece of software that has some value to the individual. In some sense, it’s a sort of passive activism, so we want people to install it. We have done some projects where we basically used polling samples to recruit users for Who Targets Me. It’s a fabulously expensive way of doing things, but if any pollsters are here in Germany and would like to do something for next year’s election, you are very welcome to talk about it.

There are the American elections and there are everyone else’s elections.

Julian Jaursch: If all goes as planned, Germany’s election will indeed be next year. Are you planning for that already? Is there an outreach or are you working with researchers or any media here in Germany for the upcoming elections?

Sam Jeffers: Not yet. That will happen as soon as there is this other very big election in November that we are trying to do some work for as well: It’s an interesting one, because what we tend to see is that there are the American elections and there are everyone else’s elections. And everyone else’s elections tend to start quite late. Obviously, the amount of money involved is much smaller everywhere else, so the sophistication tends to be less. I think that they are starting later is not actually so much of a problem as it is in the US, where we probably should’ve been working much more since last summer.

Julian Jaursch: I see. Let’s get into the UK elections, because it’s the most recent one and it’s where you’re from. If you take a step back and look at the elections that you have covered, maybe including the referendum, what’s your take on how digital campaigns are run? Is there already a change that you see from 2015 or 2016 to today?

Sam Jeffers: The 2019 election in the UK is a quite confusing one. The 2017 election was very misleading, as the campaigns were very important in 2017. The Conservatives started with quite a lot of points ahead in the polls, lost that entire lead and ended up not really winning the election at all.

And so, the narrative around the 2019 election was, would that happen again? Everyone was looking for signs that that would happen again. But historically in the UK and in lots of countries, campaigns don’t make a tremendous amount of difference. You’re unlikely to start ten percent behind and have a great campaign and end up ten percent ahead. When the polls are close at the beginning of the campaign, then you need a good campaign, you need to perform well, then you’ll move a few percent. In the UK in 2019, none of that happened. In a way, you’re looking at all those campaign techniques and you’re asking whether any of them mattered. There were a bunch of interesting new things that happened in that UK election, though.

The first is that the Conservatives did a lot more broadcast-type advertising, which is very different to the prevailing narrative around data-driven elections and micro-targeting. For the most part, they had one message: Get Brexit done. A three-word campaign and nothing else. The idea that there was an incredibly sophisticated, tailored, personalized campaign taking place, by the ultimate victor, I don’t think there’s much evidence for that.

That slogan emerged fully formed. I mean, it wasn’t something that they found during the campaign. It wasn’t something that they tested or optimized, that came out of focus groups. In some sense, that’s quite old school, that’s the same as would have been the case in the 90s. You own this kind of media narrative, you say the same thing over and over again and by the end, no one can forget what you stand for and you win the election.

The other interesting thing was the Brexit Party, which was standing at about half the seats in the UK and was running more ads than anyone else by volume. They won zero seats. In some sense, when history looks back on the campaign, they had no effect, no MPs. But it seems they successfully targeted Labour voters and encouraged them to stay at home or to switch to the Tories in a specific set of seats in the North and the Midlands that were kind of Brexit-supporting.

So, you have this interesting third party that's playing this role in suppressing vote over there.

Julian Jaursch: And that worked?

Sam Jeffers: They said it did. Again, these things are hard to fully understand. What the motivation was. But they believe that they had a very good model for what a typical Labour voter looked like, particularly one who supported Brexit. And they relentlessly ran ads attacking the Labour Party.

Julian Jaursch: In specific places?

Sam Jeffers: In specific places, that’s exactly how they felt that they could make a difference. Therefore, what you have is a kind of coordination between the two, a sort of symbiotic healthy relationship. And then the final thing that was interesting was, we had American-style Super PACs in the UK for the first time: these kind of political action committees, if not quite the same legally. But one of the lessons that the Conservatives learned between 2017 and 2019 was that they had no friends.

It’s a bit like there are no funny right-wing comedians in the UK. Right-wing parties don’t really have friends online in the same way. The Labour Party has people who will share their message and they have other groups that will fundraise and support and they have trade unions and all the rest of it. The Conservatives realized that they had no friends. And so, people with some money decided that they would be their friends and set up these small organizations with quite strange niche topics. So, there was a kind of very pro-landlord organization and there was a pro-private schools organization, there were some that were representing interests in business and the City. And they all popped up their Facebook pages right at the beginning of November, so the moment that campaign started, these pages appeared and they started running ads and pushing these kind of slightly off-the-wall issues that were nowhere else in the campaign.

My interpretation of that is that it’s the Steve Bannon line, which is: Flood the zone with shit. What these pages were trying to do is to try and push out other messages and confuse what was going on and make you feel like you might have to react to these organizations or spend time understanding it or to make all of their material less trustworthy. They all spent quite a lot of money and on election day itself, our data suggested that they reached more people than the main parties themselves. They were actually reaching twice as many people as the Conservatives or Labour Party on election day.

The experience of being a voter on election day was that you weren’t seeing messages that were saying, get out and vote. You were seeing these strange things that you didn’t really know where they came from and that were quite destabilizing to the rest. Those are the things that felt different to previous elections. Some of that is technique, but some of that is electoral law. Some of that is organizational dynamics. It's not just about the ads, it’s often about them as a tool and who’s using them and who’s putting the money into them.

Julian Jaursch: So, so they are part of a broader party strategy.

Sam Jeffers: That’s what it like felt to me, yes.

Julian Jaursch: Have parties in the UK fully embraced digital campaigning, ads being one small part of it, but in general using social media, using apps?

Sam Jeffers: I think in Britain, this is the first time. Everyone’s been waiting for years for the first digital election. What’s the first social media election? They kind of all have been quite similar for the last ten years or so. But this was the first one, where it felt as if every actor was equally matched. They all had roughly the same amount of money to spend. They were all using roughly the same techniques. They were all good at organic content, they were all good at paid content. It felt like the kind of first mover advantage for digital, which maybe the Democratic Party had in the US in 2004 and 2008 and that maybe the Labour Party had in the mid-2010s, had evaporated. Essentially, you get back to everything canceling itself out and the fundamentals of campaigns mattering, which are good leadership, a solid program, trust and all the rest of it, which won the election for the Conservatives in the end.

Julian Jaursch: When we talk about disinformation online, there are many angles that you can tackle. There could be a false          piece of news or a misleading piece of news. What type of disinformation was there in the UK elections or other elections you’ve covered? What do you think were the most prevalent parts? Was it really just fake narratives that were put out there? Was it misleading stuff? Was it obscured sources?

Sam Jeffers: I think in the UK context, we are quite used to political figures stretching the truth. Boris Johnson is someone who speaks at the third stage of truth, but generally it doesn’t go too far over the actual line, and we're kind of used to that. You’re talking big numbers, you’re talking vague, long-term things, you avoid questions, you’re clever with your words, these are the sorts of things that people expect. Not necessarily positively, but we expect politicians to do so. There’s a fair amount of that going on.

There were organizations looking at the claims made in some of the ads as to whether or not they were truthful or not. Those very famous ads in the Brexit campaign around this bus with the National Health Service, NHS, and the 350 million pounds all sit just about within the boundaries of acceptable political speech. For me, it’s sort of untrue, but it’s not so far that you shouldn’t be able to interrogate those claims and hold them to account.

The Brexit campaign was a fault of the media for not interrogating those claims. But they were on the bus every day. You couldn’t miss the bus. The bus wasn’t a micro-targeted dark ad, the bus was the bus. That sort of sets the tone. I think the Conservatives did a couple of things that were interesting, it was a kind of mainstreaming of disinformation techniques. One is that they rebranded one of their Twitter channels to be a fact-checking channel during one of the debates. They would dispute the claims that their opponents were making the whole time and they tried to take the Labour Party down. It was designed to get people talking about it. I think it backfired, but it didn’t backfire to the extent that people didn’t vote for them in the end.

And they also did quite a lot of video work that was edited in a very specific way. In British politics, everyone who is involved wants to be like the American political strategists who invented clever ads and did clever things. Everyone wants to be Bill Clinton’s campaign team from 1992. So, they made these videos and occasionally they become controversial.

But in a way, the controversy is louder than the actual number of people, who might have seen it or be misled by it. What we didn’t see were the sorts of things that have become expected in this world of bot accounts and amplifying trends. We didn’t see too many new pages or viral content that was going off in completely misleading directions. And I think that is different to 2017. In 2017, there were a couple of stories that were public the whole election around animal rights stories for example. These stories went extremely viral online and they really hurt the Conservative Party, because the Conservative Party was seen as pro-hunting. And the stories sort of tapped into this kind of little cultural undercurrent and it was never talked about on TV, it was never talked about on the radio, never talked about in the newspapers, but it was online, this stuff was flying. That didn’t happen in 2019.

Julian Jaursch: It was interesting that you mentioned what you saw in the most recent election was a lot of broadcasting, so it was not very targeted. But at the same time, there also were other parties apart from the big ones that used targeting. What is the issue with this? You could argue that it’s great that you can show your policies to a specific group that you know is interested in them. Knowing that not every party does it and measuring it is difficult, but what are some of the challenges you see with that?

What you end up with is a campaign that only tries to speak to people in incredibly tailored and specific ways. You put so much effort into that group of people that you lose the wider democratic instances.   

Sam Jeffers: I think the first one is that in the first-past-the-post electoral system, so the UK and US, there’s a lot of value in identifying small groups of swing voters in specific places. With every UK election, there usually are 50 seats that matter out of 650 and within those seats, a couple thousands of votes either way makes a difference. So, you're talking 50,000 to 100,000 votes that determine the entire election. If you believe in the power of data and you think you can identify those people whose mind is not yet made up and who could go either way, that’s the logic behind your campaign. Now, obviously, you still need to get your voters out, you still need to raise money and motivate people to volunteer.

I think that logic is something that people are still trying to get the most out of. And that’s where, if you’re a consultancy, you come in and help a campaign. You’re trying to tell them, “we can work these people out for you, we have a model for all of that.” So, I think there’s something in that that’s problematic, because then what you end up with is a campaign that only speaks to those people or tries to speak to those people in incredibly tailored and specific ways. You put so much effort into that group of people that you lose the wider democratic instances.

In that sense, we have a bad electoral system, but the way it can be abused is even worse. I think that happened in 2015, with the Conservatives targeting a smaller number of seats. I think they tried to do that again in 2017 but targeted the wrong seats. So, it can be a very costly thing if your data model is off and you’re doing the wrong sort of thing. They didn’t really try and do it in 2019, it was much more straightforward. They did have advertising targeted to individual seats, but it wasn’t that they were showing one thing particularly. It wasn’t a completely different message or different campaign in different places. But the risk is there, and I think that’s an issue.

The second thing is that people still don’t really understand how the data gets used in any sense on the internet. Specifically, once you start thinking about people trying to persuade you and your political views. I think that’s much more serious than some of the other forms of targeting that takes place. If I’m looking for shoes online and I see some ads for shoes online, I kind of understand the bargain that was made. If I am relentlessly profiled over a period of many years in order to try and put me in a particular bucket of voters, so that they can then serve me specific messages, I probably don’t have much understanding of what’s going on there. And therefore, my rights to be treated as an autonomous citizen have been infringed to some extent.

Julian Jaursch: These risks that you mentioned, I understand they are hard to measure with the tools that are out there, provided by the companies. So, the obvious question is, what do we do about this? Trying to strike a balance between allowing people to voice their opinion and to promote their positions, but at the same time trying to mitigate the risks that you just mentioned, when it comes to privacy issues, data profiling and data collection, when it comes to targeting very specific groups of people with specific messages. I am not suggesting that you have all the solutions for us tonight, but you’ve been working on those for quite some time. I was wondering, when it comes to transparency, what type of specific measures would you like to see in place. Do they concern the platforms? Do they concern political parties, the advertisers? Who’s responsible here?

Sam Jeffers: What we obviously have are things like the Facebook ad library: Now you can at least see all the political ads. Or rather all ads that Facebook decides are political will end up in a place on the internet for seven years. And you can see some data about how much money was spent, roughly which demographic stats they were showed to and the content of the ad. And that’s been useful. You can now see what’s going on almost in real time. Occasionally, they miss stuff and all of the platform companies have slightly different definitions of what’s political or not.

Also, now all of the platform companies have taken slightly different views on political ads. Twitter tried to ban political ads without ever really properly defining what they are. Facebook has decided to allow them all, but put some of them in the library, again without ever really defining what they are. Google has a different view on it: They tried to limit targeting options, again, without really defining what political ads are in the first place. So, you have these kinds of issues of definition, of what libraries are, whether they work: the further we get into this, the more kind of granularity about what an ad library is, what data it should present, while in turn still preserving people’s privacy. There is basically all the content, all the spending, the targeting, and the interaction really all there. So, they understand something about what people’s response to it is.

There’s a lot of incentives on social media to get people to try and share things and comment and spread the message for you. We know nothing at all about that really with advertising at the moment. But we clearly know their incentives, because it’s free advertising and it’s almost like money laundered through people’s reputation when they share it for you. At the moment, we know nothing about that. I think that sort of stuff would be useful. There are too many ads out right now. It’s basically an impossible job for anyone trying to scrutinize this material, to work out what’s really going on. Like the example I gave earlier of the Conservatives and Liberal Democrats in the UK, both running upwards of 10,000 ads during a six-week campaign.

I think we should just limit the number of ads you can run at any one time. That would make practical scrutiny so much easier, but it wouldn’t harm the speech implications. You can still say what you want. You could still target the ads however you want. You can still spend whatever you want, but at the same time you will be held to account for those things, because any old journalists can go into an ad library and have a look at what you’re doing and ask you questions about it. And I think coming to kind of these more practical limitations is where we have to go next. Because I think we’ve won the argument on transparency as a single word, but we haven’t yet won the specifics of what the ultimate program and policy should look like.

Julian Jaursch: That’s what I was wondering. If you say, “we won the argument on transparency”, who do you trust with having the final say of what transparency looks like? Do you think the companies themselves, be it Google, YouTube, Twitter, Facebook, whoever, have to do that voluntarily on their own? Which has been the case so far with some sort of commitments at the EU level in this field. Should there be a strict regulatory guideline from governments on the national level? Is there a mix of it? Who do you trust with achieving this transparency?

I used to have dreams of a global solution. Some sort of global observatory of political advertising.

Sam Jeffers: That’s really hard, because on one hand the platforms should be good at it. You know, they have all the data, they have people who are good at making products that people want to use. And there should be some standard for a normal voter out there that says, “Do I have the transparency I need?” And they could test their products and they could build them around that set of user needs.

Similarly, for journalists, are these the tools that journalists and researchers actually need? I think at the moment, companies’ outreach to those communities and to normal voters hasn’t really been sufficient. They’ve been building it and this is what we’re comfortable with. Yet, if you shove that into the hands of governments, you’re asking governments to build software using data that they don’t really understand. You know, generally you don’t have the expertise to really work that out.

We have been asking the UK and other governments to engage with the idea of what’s the specification of APIs that have this data, so that people could work with it. And everyone’s goes back to the core concept of transparency. There is a list of 10, 15 things that would be useful to provide with data that would mean that this data will be brought together somehow. It could be displayed in different ways, it could be explained to people in different ways. That kind of opening up of this information would really help, rather than saying Facebook has designed it this way and Twitter designed it this way and so on.

I used to have dreams of a global solution. Because I’m not sure this is a national problem. Platforms are too big in a way to kind of wrap your arms around from the national level. Some sort of global observatory of political advertising: They would engage seriously with the technical issues, they would understand that there are different circumstances and different countries and different electoral systems and so one would kind of treat them as an expert in that way. A small couple of percent levy on the political advertising on social media platforms might pay for it.

A group of people could follow that around the world and kind of learn from each election and continue to evolve the policies and so on. So, it would be sort of semi-self-regulatory. It would have the kind of arms-length independence to make decisions. And in the end, what Facebook has done with its new supreme court isn’t too dissimilar, I think.

Julian Jaursch: Okay, interesting. I got a ton more questions, especially on this point of a global observatory, but I want to also open it up to the floor. So, maybe I lead over to the Q&A session here with an outlook that I would ask from you. You mentioned earlier that there are a lot of other important elections coming up before the German one next year. One would be the one in Ireland, which is this Saturday. You’ve been following that very closely as well for the past couple of weeks or months, since it was announced. Maybe you can give us an impression of what’s been happening in the Irish elections? What do you see when it comes to online ads, the targeting strategies and the content that’s out there?

Sam Jeffers: Sure. So, I live in Ireland actually, as of the last few months, which has meant a crash course in Irish politics, which is different to what it looks like from the outside. That has been an interesting thing. From the outside, certainly from the UK, Irish politics looks like they are only negotiating about Brexit. When you live in Ireland, it’s all about domestic issues and no one cares about any of that stuff at all really. I wasn’t expecting it to be a very interesting election, but there was a poll yesterday that suggested that the political wing of the IRA is set to be the largest party by vote share in the Irish election. And this has brought up a lot of questions, which is actually really nice.

Most of the parties signed a fair play pledge to play by the rules and not mislead and to use data responsibly. And many of them had ads running. Tens of ads or maybe 50 ads at most, which has meant that it’s been really easy to see what they’re saying and who they’re saying it to. The Facebook ad library, for some reason, has a better geographical breakdown in Ireland than it does in the UK. So, you can see what’s happening in individual counties as opposed to in the UK, which is England, Scotland, Wales and Northern Ireland, which is not a useful breakdown at all.

And it feels very politics as usual and quite healthy in that respect. That seems quite encouraging. And that’s quite different, they repealed the abortion referendum a year and a half ago. And that was pre-Facebook-advertising-transparency. There was a lot of concern about American Christian groups buying ads, there was a lot of pressure on the companies, which have headquarters in Dublin for tax reasons. Lots of the staff within those companies are very uncomfortable about the policies that the companies had. So, Ireland’s been an interesting case, because it’s quite a small country and not always at the center of everything, but when it comes to tech companies in Europe, this employee activism can make interesting things happen and I think that’s been beneficiary.

Julian Jaursch: Thanks for those insights. There’s a lot of topics that we covered and that we can go more in depth now depending on what the audience is interested in. We talked about some of the potential of political ads online, some of the issues with them, especially with targeting. We talked about specific elections, UK, Ireland. So, I want to open it up to the floor and get a couple of questions in on the topics we discussed or something else that you want to bring in.

Guest 1: You mentioned the US elections and everybody else’s elections in the beginning. What would be different in these elections in terms of political ads? The second question is, you mentioned that you can have 10,000 slightly different ads all targeting different people. What do you know about how effective it is?

Campaigns often get talked about in the media as if their job is only to be persuasive, like their job is just to change the minds of the people. But their job is also to raise the money that you need for the campaign, it’s to get people sharing your message on your behalf.

Sam Jeffers: I think that depends a lot on what your goals are and what your timeframe is. Campaigns often get talked about in the media as if their job is only to be persuasive, like their job is just to change the minds of the people. But their job is also to raise the money that you need for the campaign, it’s to get people sharing your message on your behalf. It’s to get people signed up to volunteer and spend time talking to other people and making phone calls and text messages and all of these other things. You have these money-mobilization layers of your campaign that change over time. In American campaigns, they are fundraising all the time until maybe like the last week, when they’re probably not fundraising anymore and it’s all into the kind of message and the mobilization phase. You’re building your volunteer base the whole time. So, you’ve got to look at the campaign and those layers and the ads serve different purposes within that as well.

Some of that goes into the content. You can take the Conservative Party in the UK, they tend not to fundraise with online ads, because they don’t really have a grassroots donor base. They get their money from industrialists and media moguls and other people like that, who give them big sums of money and they go on. Whereas the Labour Party spends a lot of time fundraising and encouraging people to register to vote. So, they may not be effective in that sort of changing minds perspective, but they might increase the size of the youth turnout, for example. That is obviously a really useful thing to have happen in your election as well.

There are quite a lot of studies that say that political ads aren’t terribly persuasive in terms of, “I just saw an ad and I used to think this and now I think that.” But they are persuasive in the longer run, in terms of building the mobilization in the sense that you have a viable campaign, that you have something to say.

I think some of the other things we sometimes see are people who really are playing a long game. For example, around Brexit, for a long time, a no deal Brexit wasn’t a thing. The first year and a half of Brexit, it was all about “How are we going make all this happen?” And then suddenly, I think about spring last year, we received these organizations, nobody knew who they were, they were starting to spend serious amounts of money asking people to write to them, about who is in favor of a clean break.

I don’t think that the correlation is absolutely there, but you start seeing some parliamentarians talking about it, the media starts talking about it and the whole idea begins to get some momentum. I’m not going to take the ads as the thing that necessarily got all of that going, but it seemed like there was a campaign over six months or so to spend some pretty serious money online, to try and normalize some ideas that didn’t previously exist.

So again, I'm not sure that there was a single ad that made anyone go from here to there. But suddenly, there’s a new concept out there, it begins to find some traction and it plays into some of the wider campaign themes.

As for the US, we have this idea of all these campaigns being terrifically mature. They have people who are very good at data, they have people who are very good at buying this stuff in a very efficient way. They have people that are very good at the messaging. The American data protection laws are a much looser than what you find in Europe and so people experiment a lot more with things that you probably wouldn’t feel comfortable doing here.

There are people out there making the argument that Trump’s rallies are really good for getting signatures off people’s mobile phones and thus kind of painting them: You came to a rally and then the campaign can reach them again after. The idea of your physical location and the kind of proximity to certain things is going to become a thing in American campaigns much more.

It’ll be the usual, just up another level. We didn’t have AI in 2016 and now we’ll have someone selling AI-based advertising optimization, which may or may not mean anything, but it’d be another technique that then by spring next year, there will be American consultants in Berlin selling it. It has this cycle.

It seemed like there was a campaign over six months or so to spend some pretty serious money online, to try and normalize some ideas that didn’t previously exist.

Julian Jaursch: If I may add to that: It’s very helpful that you explained that it’s not just about persuasion, it’s not just about getting a voter to vote for party 1 when they had previously voted for party 2. I think it was the former chief security officer of Facebook who said this and the head of the Trump campaign said this himself, too: The ads are also used to gather data about people. With an ad, you entice them to sign up for a newsletter or donate something, so you get their phone numbers. If you have the phone numbers, you can target them with very personal messages and you don’t even need an ad. Is that something that you see? Is the ad a means to an end to gather people’s data and then target them off of Facebook?

Sam Jeffers: Well, if you’re gathering data through Facebook, I think it’s not totally straightforward to target them off Facebook. The data you’re trying to get there is to reach people more on Facebook. But there are so many websites you go to. I have a thing installed on my computer, another browser extension called the Facebook pixel tracker, which is actually a Facebook tool. It shows you which sites have a Facebook pixel. The Facebook pixel will basically tie back to your Facebook profile and follow you around the internet as you travel and then allow targeted ads based on that. It will also combine with my Facebook profile where I am building up behavior and liking things and looking at certain content, which goes into the picture of who I am.

Then Facebook can sell a model of “Go and find me all the other people who look like Sam based on some opaque criteria.” If you look at the Trump campaign, which is spending $200,000 a day on Facebook ads and has been doing so for quite a long time now, it’s painting a lot of people with little touches of data. That seems to me to be the approach in the US at least: How do you build the biggest possible pool of people, who you can then do things with?

Julian Jaursch: So, some of the ads that you see, especially from the US, are not necessarily ads that say, “Vote for me.” It’s ads like “Sign up for this” and then you have the phone number and then what do you mean by painting? Knowing more about the voter?

Sam Jeffers: Yeah.

Guest 2: I’m from Ireland but living here in Berlin, so my Facebook profile and my own personal social media program, I suspect is linked to me here rather than there. So, I’m not seeing any of the ads at all. What I am seeing are sort of like memes that people put up of an event that’s happened. Somebody said something and that’s then being circulated amongst friends and so on. Is there any way of tracking these sorts of memes?

Sam Jeffers: We focus primarily on the paid content and that’s a relatively small part of what people see online. One of the other explanations of the 2017 election in the UK was that the Labour Party has so many people sharing their stuff that even though the Tory Party was buying five times as many ads then, you would still see five Labour-supporting posts for every one Tory post. They’d be essentially drowned out by all the organic stuff.

In the most recent UK election, the Conservative Party spent a lot more effort on making their own organic viral content. They hired two guys from New Zealand, who’s specialty was shit posting and trying to throw people off the set in terms of what’s going on and get people talking about that content disproportionately to how good or important it was. They were making graphics with terrible fonts and the sorts of things that you see your aunt and uncle sharing on Facebook.

I actually don’t think it was massively effective. It was effective in the sense that the media was really interested in it and wanted to talk about it. What I don’t really think happens is that there is a content factory with close links to parties. I think that happens in Russia, I think it happens in some strongman-type states. There’s not some basement in the party HQs full of people churning out memes.

I don’t even think we’re at the situation where parties have the resources to sit in local Facebook groups, for example, to any great extent and push content through those, which is a really big thing in India. BJP in India have WhatsApp captains whose job is to get into WhatsApp groups and push content through WhatsApp groups. Their job is to manage a bunch of those groups and make sure that the memes are floating around in those places. It’s the same in Brazil.

I haven’t seen that so much in the UK or Ireland. There’s an area to catch up in as a campaign. There’s no data protection law to make that more difficult for you. Just go in and post stuff.

Guest 3: There was a huge gap on spending for Facebook ads, or ads in general, in Germany for the elections in 2019 and at the same time a massive gap on the organic reach between the AfD and for example the Christian Democrats or the Social Democrats. Was there a similar situation in the UK? And how could you explain the huge gap between the spending for ads and the very high organic reach for the AfD?

Sam Jeffers: We never quite worked out what the pattern exactly is, but I think we see it quite often. If you take the Irish example at the moment, Fine Gael, who are the governing party, they are spending 10,000 Euros a day on political ads, which in the Irish context means that actually they’re spending nearly twice as much per head as Trump is spending right now. It’s not a huge cash sum, but relatively speaking, there is quite a lot of volume there. However, as I mentioned Sinn Féin is potentially leading in the polls and they are spending almost nothing on ads at all, but have people sharing their content and pushing it. If you look at the response to either party’s posts online, Sinn Féin gets positive comments and positive shares and everyone else gets negative comments.

One interpretation of that is: Traditional parties are really struggling to work out how to cut through online. They don’t have exciting, internet-friendly leaders who are very authentic and natural in that space. They don’t have people who feel like their issues never get covered in the mainstream media.

There was a Belgian election at the same time as the European parliament elections last year and there was a right wing nationalist Flemish party, who the media won’t cover, right? “They’re too racist, we can’t cover these people.” They spent record amounts on Facebook to get around that. They absolutely went around the media to try and get as much reach as possible that way. So, they were both successful organically, but they were spending more money than anyone else by miles on Facebook. It’s just often a case of the practicalities. As a political consultant, there are limited resources, there’s limited time, there’s limited capability, often campaigns are chaotic and no one really knows what’s going on anyway.

It seems to me that the center right, particularly traditional liberal parties, struggle to get the same reach online as people who are saying slightly different things outside the political mainstream.

Guest 4: What is the main driver for the reduced use of disinformation you mentioned in the UK? I have another question, because you mentioned Brazil and India. Do you think about strategies that can be thinkable from the perspective of NGOs or companies to tackle the problem there?

Sam Jeffers: One of the questions is how to get the data in the first place to find out what’s really going on. I mean, we have a small tool that’s installed by a small number of people, where we can kind of see through their eyes. We now have ad libraries for the paid stuff, but no one knows what’s going on WhatsApp at all. There’s no ability to protect elections as they’re happening in any sense. There are people who are beginning now to have some ideas around how you might do that kind of technical ideas.

But it’s not a straight-forward tool without unleashing potential really horrible unintended consequences at the same time. We might not like where we are, but there is some risk about just kind of heading off in another direction. You see laws popping up around fake news which are obviously going to be used as oppressive tools at some point by governments.

There should be so much more of small tech taking on big tech. That’s a dynamic that we need to foster.

You know, one of the arguments in the UK often is about doing more fact-checking of political ads, that ads should be pre-cleared for truth. They have something similar in India that allows the government to snarl up or position advertising for months in kind of red tape. I’m not sure we want those types of things happening – as well intentioned as they as they seem. So, I think one of the first calls is, “Can we get the data we need to be able to answer these questions more authoritatively?”

This question how important all of this is is really fair. We have concepts like filter bubbles, which pop up and then are debunked. If we had made big filter bubble laws, would we have been wasting our time doing that? There are so many parts to this, where we don’t have enough in terms of the answers. If you take Who Targets Me as a very prototypical solution, it’s not the answer by any means, but we think we are testing different things out and heading in the right direction.

There should be so many more of those small tech organizations taking on big tech. That’s a dynamic that we need to foster. And exploring technology with technology is a thing that we need to do as well. It's a lot of time spent looking at how do we get our arms around technology with law and with bullet points in policy documents and these and they don’t work in the same way. The way we use our phones doesn't always feel like the way we might read in a policy document or a proposal from a government. I think there's a whole load more explorations to do, and I think nobody can say they can answer these questions very well. It's only one year of a five-year project.

Guest 5: I have a question about your extension. How much do you know about the people that are using it? And then if you do get any demographic data, is there a certain type that’s using the extension and what does that say about the dataset?

Sam Jeffers: We don’t know very much. It’s all voluntary anyway. We ask people how old they are, which electoral district they live in and their age and where they think they sit on a left-right political spectrum. Essentially, there’s a little bit of demographic data, not enough to work out who it actually is. Associated with that, we see a sort of timeline of the advertising that they saw and so then you can draw your conclusions across that to some extent.

Our user base is too male, a little bit too young, too urban, too left-wing. As I said, it’s never been our goal to try and create perfectly representative samples.

Where we have been successful in a given country, we might have enough people that we might be able to normalize the data more. But again, each of those users use Facebook a different amount. Each of those users is targeted, there’s so many other things going on. I’m not sure that the statistical value of the sample doesn’t work independent of what we come up with.

Guest 6: You said that you had looked at Facebook and WhatsApp. I was wondering about the other social media platforms like Instagram, LinkedIn and Snapchat. Which insights do you have into what’s happening in those platforms?

Sam Jeffers: None really. What we’ve tried to do is to use Facebook as the kind of battering ram, because for every 10 Dollars, Euros, Pounds spent on Facebook on political advertising, a dollar is spent on Google and 10 cents or a cent is spent on LinkedIn. Snapchat is very small as well. Facebook is the place where many campaigns go, because they know they can get audience. So really, it’s been trying to exemplify what the platforms should do.

We see a Facebook ad library and Google has some kind of transparency tool that could be a lot better. Snapchat publishes the data and for a while, it had a kind of ad library. All of that data becomes something that we can do things with. We have just built a tool, it’s basically a personal ad notifications tool. You could sign up and say, “I want to see all the ads from the CDU. I want to see all the ads on this topic.” And it will email you each day with all the new ads that popped up today. The goal is to try make it easier for journalists to write about this stuff and hold it to account. Once the data starts to become available, you can deal with it. Now I can see in real time how much money is being spent on this platform versus this platform versus this platform. And how much is going into younger people over here and old people over here and those sorts of things. That is kind of where we want to head with this as a society or as a project. But it’s all Facebook all the time for us at the moment.

Guest 7: We see that a lot of right-wing populist influencers moving from open platforms like YouTube when they are banned. They’re moving to Telegram, for example, which is a big issue. Is there any kind of evidence of a party moving from YouTube, from Facebook or Instagram to Telegram and using private chats to share ads or something like that?

Sam Jeffers: It’s just a question of scale and practicality, right? You could reach 30, 40, 50 million people on Facebook. The reason why some of these right-wing populists can’t afford to buy food anymore is because they are on Telegram. It’s changed the dynamic of some of these people’s lives there. Probably the easiest way to track this would just be these companies reporting their monthly active users. Facebook still has a growing number of users, people use it a lot, they make more money every quarter.

As long as that’s the case, that’s where political parties, mainstream actors will go, because it’s the most efficient place to spend their money. I suspect that there are young people that are in different places and they’re doing things in different ways. I mean, parties in the US support ads inside Xbox games, places that don’t have transparency.

Guest 8: I wonder whether you could say something about that report about foreign interference in the Brexit that was not published.

Sam Jeffers: I wish I knew.

We’ve not seen much foreign interference in two and a half years of tracking political advertising in the UK. We’ve seen one or two instances of things that look weird. For about one percent, not even one percent, we’re like, “What is that?” We’ve seen a couple of things that were attributably Russian. They were in some sense following the theory of create division. There was that really horrible fire in the tower block in London, where all those people died. This group made a documentary, saying it was a conspiracy and that the Conservative government wanted it to happen and effectively killed all these people. You could see that it got some response and they bought some ads and they’ve been making other documentaries about other things in other places.

But that’s an extremely expensive way of trying to sow division. These are 30-minute documentaries, they’re good production values in the US today and it is not probably a good use of anyone’s time. And it was really obvious, you could see who it was and everyone who was working there was a former Russian agent. It was a pretty transparent effort of interference.

I’m not sure the Russian report is going to show anything particularly. I don’t think it’s been suppressed because it shows widespread infiltration of the information environment.

Guest 8: And the financing.

Sam Jeffers: The financing of parties is different. London is full of Russian billionaires and they throw parties and they give money to people. How well orchestrated that is is way outside the scope of what I know about that.

Guest 9: You’ve just touched the issue that I wanted to mention about Russian interference. We know that we have these well-equipped and well-staffed groups sitting in St. Petersburg and trying to influence American and European elections. What do we do in order to carry on legitimate information that you’ve mentioned and getting the audience to distinguish between the bad guys and the good guys? What solution do you have to offer?

Sam Jeffers: I think it’s the T-word again: Transparency is important. For example, these new Conservative Party pages that were set up at the beginning of the election campaign, where affiliated people were funneling money into Facebook ads. The problem there is the lack of financial transparency about where that money comes from. Do we have enough disclosure about that as it happens? At least now thanks to Facebook transparency, I can find out something about who those pages are, but I still can’t find out who actually paid for anything. And that really is the main part: money. Money makes the world go round and it makes all of this stuff happen. We still don’t know enough about that. You could imagine platforms being much stricter in their financial disclosure for people running ads.

We do have to be quite careful about how we talk about some of this stuff in that it might affect people’s trust in the overall process. If they think there are some things you can’t trust, maybe that spreads and becomes a worse virus than the original problem.

Already for the US, Facebook is trying to get people to register at the Federal Election Commission and have numbers against their ads. It’ll be much more straightforward to track it against the spending records in the US and people buying ads from outside the US would be no longer allowed. There probably are a bunch of things that can be done to tighten up where the money comes from and who ultimately pays for all of this.

But again, the scale is often really confusing. I don’t think in 2020, we will see anything like the scale that of what happened in 2016. It’s no longer open season for foreign actors to run around on social media platforms causing trouble. I think the bogeyman of it happening is almost more of a problem now then the reality of that happening. I think people will talk about it more than it actually happens and if it does happen, crisis points in the media will go crazy about it.

We do have to be quite careful about how we talk about some of this stuff in that it might affect people’s trust in the overall process. If they think there are some things you can’t trust, maybe that spreads and becomes a worse virus than the original problem.

Guest 10: You explained how the social media campaign in the UK was very successful for the Conservatives, but we have, at least in Germany, 20 to 25 million people who are above 60, 65 who are not necessarily using all these social media tools. I was just wondering if you can also somehow reach these people directly through social media or indirectly over acquaintances on social media? Another question: The Trump team in 2016 more or less scrutinized every state and every electoral district in a very deep way. Is this the same thing that you estimate for their November election this year that the Trump team will try to see what Democrats will do?

Sam Jeffers: Regarding the first question, older people definitely use social media a lot. It’s sort of spoken like a joke, but you know, there are a lot of programs that say we must teach media literacy to young people that they must understand that not everything on the internet is true. Well, no, actually, we should teach older people individually because they’ve been confronted by a new technology that they haven’t grown up with and they don’t understand and they think everything on the is true and they share it. And then Brexit happens, you know, one, two, three.

That is a real issue and so this is one of the things that what we have to do: Institute big media literacy programs. And it has to filter through schools and it’s going to take all this time. In the meantime, that will allow six or seven electoral cycles, where the information environment is terrible. We probably need to think a lot more about, “How do we deal with this stuff at source?” That’s true for all audiences.

On the second question on the US, you know, the Hillary campaign in 2016 was doing all the same stuff. They’re trying to work out where their voters are, who’s a swing voter, but their polling was off. They missed a couple of states and they lost the election. There isn’t a particular reason to believe that they won’t have good polling, good data and all the rest of it. Whoever the ultimate nominee is will have the same tools.

Trump might win the election, but he won’t win it because he raises more money than the Democrats. He won’t win it because he has more volunteers than the Democrats. He won’t win it because he runs better ads than Democrats. He will win it because he has this enormous bully pulpit and he has a successful economy, he has all these other things that are going in his favor. It might be really close and maybe some of the campaign techniques will be really important in how it ultimately plays out. But there’s no particular reason to believe that Republicans are massively better at this stuff than their opponents. You know, their opponents were world-leading ten years ago and there’s no reason why they wouldn’t be world-leading again at some point in the near future.

Guest 11: You said that the Conservative Party pinpointed the 50 seats that were persuadable and that they had very specific targeted ads in those regions. Did they actively use disinformation or do you have any insight on what ads they actively used?

Sam Jeffers: In this election, it was much easier to see where they were targeting things because the ads would literally say the name of the place that they were targeting in the ad. You would just go and look at the ad library and they would say, “We are targeting Warrington.” It would say, “Vote Conservative at Warrington to get Brexit done.” There’s nothing there other than just the message repeated again and again and again.

The sorts of things that we saw in terms of tailoring the campaign were much more subtle. In the North, they didn’t tend to use pictures of Boris Johnson in their ads, but they did that in the South, because he’s less popular in these traditionally Labour-supporting places. He’s a rich posh old Etonian type and that doesn’t play so well in poorer, working-class towns. So, you just take the picture out of the ads. They were doing things like that, but it wasn’t massively sophisticated in that respect. It was the same template played out 50 or a hundred times, swap out the name of the place, take out the picture of the prime minister, put something different in.

In 2017, it was much more interesting because this data wasn’t available. What we could see were the seats that were targeted, where they were running their ads. In Britain, 326 seats is the winning line. And they were targeting seats way over the winning line. They thought they were going to win by miles and they should’ve been defending around the line. And so, you have this clear example of how you can do this wrong and totally misunderstand your data.

Guest 12: This issue of ads being used and people lacking trust in these institutions that traditionally they’ve always believed in. Do you think this is perhaps the greatest issue with Facebook, dark ads and issues that surround how people receive their information? You mentioned that one day, so many people are going to feel disenfranchised and unable to trust in any sort of organization that that in itself creates further divisions in a way that foreign actors couldn’t necessarily do.

At the moment, the platforms say, we don’t have the answers to these challenges, but we’re not going to share anything about it. And that is a really bad place to be.

Sam Jeffers: I don’t think that’s the only future. The other future is one that goes back to the first question: In the end, we need more data to get released, and maybe then the research suggests this isn’t that important. Maybe people are more informed than they were before. Maybe we’re still just living through the embers of the financial crash in 2008 and therefore, it’s kind of instable as a result of economic conditions and a lack of agency. And we move into a new historical era and everything turns out okay.

We talk about this stuff endlessly without really knowing what’s going on. People become very destabilized in what they think truth is or isn’t, what they can trust. There’s a sense of democracy not being as popular as it was in the immediate aftermath of the fall of the Berlin Wall, right? Democracy had a 15-, 20-year honeymoon period and now the question is, “Does this deliver for us?” If we can’t trust the process of it, that’s even worse and you end up in a totally different place. I don’t think we have to end up in the kind of hell scape. But I do think we need the platform companies to be forced to take a bit more responsibility about the pace of change that we are all comfortable with.

How do we, as companies, say we don’t have all the answers to these things and therefore genuinely open up to help people find out what those answers are? At the moment, they say, we don’t have the answers to these things, but we’re not going to share anything about it. And that is a really bad place to be. But I still hold some optimism.

Julian Jaursch: That’s a great way to end. First of all, it’s an optimistic way to end and it’s also a call to action to the platforms, which was interesting to hear again. You pointed out a responsibility for them to open up to be more transparent themselves and create transparency.

I think that brings us full circle to the point that we are dealing with a different type of ad. And I think through the questions that you answered, you really worked out what the differences are. They might not all be negative, but they are radically different from a paper or a TV ad. I think the example that you gave with just switching out the image of a politician very easily on a large number of ads, that’s just not possible in traditional media inexpensively.

I really appreciate that you answered all the questions and worked out these specific characteristics and challenges. I want to thank all of you for coming in and for your questions and your curiosity, it was really nice and very engaging. And thanks again to Sam.

Sam Jeffers: Thank you so much.

- End of Transcript -