0:00:00.000,0:00:19.030
36C3 preroll music
0:00:19.030,0:00:26.500
Herald: OK. So inside the fake like[br]factories. I'm going to date myself. I
0:00:26.500,0:00:32.980
remember it was the Congress around[br]1990,1991 or so, where I was sitting
0:00:32.980,0:00:38.550
together with some people who came over to[br]the states to visit the CCC Congress. And
0:00:38.550,0:00:43.230
we were kind of riffing on how great the[br]internet is gonna make the world, you
0:00:43.230,0:00:46.970
know, how how it's gonna bring world peace[br]and truth will rule and everything like
0:00:46.970,0:00:57.259
that. Boy, were we naive, boy, where we[br]totally wrong. And today I'm going to be
0:00:57.259,0:01:03.470
schooled in how wrong I actually was[br]because we have Svea, Dennis and Philip to
0:01:03.470,0:01:08.980
tell us all about the fake like factories[br]around the world. And with that, could you
0:01:08.980,0:01:17.670
please help me in welcoming them onto the[br]stage? Svea, Dennis and Philip.
0:01:17.670,0:01:28.810
Philip: Thank you very much. Welcome to[br]our talk "Inside the Fake Like Factories
0:01:28.810,0:01:35.899
". My name is Philip. I'm an Internet[br]activist against disinformation and I'm
0:01:35.899,0:01:38.719
also a student of the University of[br]Bamberg.
0:01:38.719,0:01:45.039
Svea: Hi. Thank you that you listen to us[br]tonight. My name is Svea. I'm an
0:01:45.039,0:01:50.219
investigative journalist, freelance mostly[br]for the NDR and ARD. It's a public
0:01:50.219,0:01:55.759
broadcaster in Germany. And I focus on[br]tech issues. And I had the pleasure to
0:01:55.759,0:02:01.280
work with these two guys on, for me, a[br]journalistic project and for them on a
0:02:01.280,0:02:04.289
scientific project.[br]Dennis: Yeah. Hi, everyone. My name is
0:02:04.289,0:02:09.009
Dennis. I'm a PhD student from Ruhr[br]University Bochum. I'm working as a
0:02:09.009,0:02:16.160
research assistant for the chair for[br]System Security. My research focuses on
0:02:16.160,0:02:21.349
network security topics and Internet[br]measurements. And as Svea said, Philip and
0:02:21.349,0:02:26.660
myself, we are here for the scientific[br]part and Svea is for the journalistic part
0:02:26.660,0:02:31.790
here.[br]Philip: So here's our outline for today.
0:02:31.790,0:02:38.550
So first, I'm going to briefly talk about[br]our motivation for our descent into the
0:02:38.550,0:02:45.160
fake like factories and then we are going[br]to show you how we got our hands on ninety
0:02:45.160,0:02:50.780
thousand fake like campaigns of a major[br]crowd working platform. And we are also
0:02:50.780,0:02:56.080
going to show you why we think that there[br]are 10 billion registered Facebook users
0:02:56.080,0:03:04.360
today. So first, I'm going to talk about[br]the like button. The like button is the
0:03:04.360,0:03:12.150
ultimate indicator for popularity on[br]social media. It shows you how trustworthy
0:03:12.150,0:03:18.620
someone is. It shows how how popular[br]someone is. It shows, it is an indicator
0:03:18.620,0:03:26.520
for economic success of brands and it also[br]influences the Facebook algorithm. And as
0:03:26.520,0:03:31.710
we are going to show now, these kind of[br]likes can be easily forged and
0:03:31.710,0:03:38.580
manipulated. But the problem is that many[br]users will still prefer this bad info on
0:03:38.580,0:03:45.960
Facebook about the popularity of a product[br]to no info at all. And so this is a real
0:03:45.960,0:03:53.780
problem. And there is no real solution to[br]this. So first, we are going to talk about
0:03:53.780,0:03:58.990
the factories and the workers in the fake[br]like factories.
0:03:58.990,0:04:04.210
Svea: That there are fake likes and that[br]you can buy likes everywhere, it's well
0:04:04.210,0:04:09.660
known. So if you Google "buying fake[br]likes" or even "fake comments" for
0:04:09.660,0:04:15.100
Instagram or for Facebook, then you will[br]get like a hundreds of results and you can
0:04:15.100,0:04:19.989
buy them very cheap and very expensive. It[br]doesn't matter, you can buy them from
0:04:19.989,0:04:27.790
every country. But when you think of these[br]bought likes, then you may think of this.
0:04:27.790,0:04:34.960
So you may think of somebody sitting in[br]China, Pakistan or India, and you think of
0:04:34.960,0:04:40.240
computers and machines doing all this and[br]that they are, yeah, that they are fake
0:04:40.240,0:04:47.630
and also that they can easily be detected[br]and that maybe they are not a big problem.
0:04:47.630,0:04:54.880
But it's not always like this. It also can[br]be like this. So, I want you to meet
0:04:54.880,0:05:03.120
Maria, I met her in Berlin. And Harald, he[br]lives near Mönchen-Gladbach. So Maria, she
0:05:03.120,0:05:11.750
is a a retiree. She was a former police[br]officer. And as money is always short, she
0:05:11.750,0:05:19.670
is clicking Facebook likes for money. She[br]earns between 2 cent and 6 cent per like.
0:05:19.670,0:05:28.720
And Harald, he was a baker once, is now[br]getting social aid and he is also clicking
0:05:28.720,0:05:34.480
and liking and commenting the whole day.[br]We met them during our research project
0:05:34.480,0:05:40.930
and did some interviews about their likes.[br]And one platform they are clicking and
0:05:40.930,0:05:46.750
working for is PaidLikes. It's only one[br]platform out of a universe, out of a
0:05:46.750,0:05:52.070
cosmos. PaidLikes, they are sitting just a[br]couple of minutes from here in Magdeburg
0:05:52.070,0:05:56.990
and they are offering that you can earn[br]money with liking on different platforms.
0:05:56.990,0:06:02.410
And it looks like this when you log into[br]the platform with your Facebook account
0:06:02.410,0:06:07.300
then you get in the morning, in the[br]afternoon, in the evening, you get, we
0:06:07.300,0:06:13.260
call it campaigns. But these are pages,[br]Facebook fan pages or Instagram pages, or
0:06:13.260,0:06:18.240
posts, or comments. You can, you know, you[br]can work your way through them and click
0:06:18.240,0:06:22.930
them. And I blurred you see here the blue[br]bar; I blurred them because we don't want
0:06:22.930,0:06:29.800
to get sued from all these companies,[br]which you can see there. To take you a
0:06:29.800,0:06:37.310
little bit with me on the journey. Harald,[br]he was okay with us coming by for
0:06:37.310,0:06:44.280
television and he was okay that we did a[br]long interview with him, and I want to
0:06:44.280,0:06:50.080
show you a very small piece out of his[br]daily life sitting there doing the
0:06:50.080,0:06:53.540
household, the washing and the cleaning,[br]and clicking.
0:07:26.760,0:07:36.020
Come on. It could be like that. You click[br]and you earn some money. How did we meet
0:07:36.020,0:07:41.150
him and all the others? Of course, because[br]Philip and Dennis, they have a more
0:07:41.150,0:07:45.169
scientific approach. So it was also[br]important not only to talk to one or two,
0:07:45.169,0:07:50.120
but to talk to many. So we created a[br]Facebook fan page, which we call "Eine
0:07:50.120,0:07:54.210
Linie unterm Strich" (a line under a line)[br]because I thought, okay, nobody will like
0:07:54.210,0:08:01.080
this freely. And then we did a post. This[br]post, and we bought likes, and you won't
0:08:01.080,0:08:10.310
believe it, it worked so well; 222 people,[br]all the people I paid for liked this. And
0:08:10.310,0:08:18.259
then we wrote all of them and we talked to[br]many of them. Some of them only in
0:08:18.259,0:08:23.410
writing, some of them only we just called[br]or had a phone chat. But they gave us a
0:08:23.410,0:08:29.949
lot of information about their life as a[br]click worker, which I will sum up. So what
0:08:29.949,0:08:36.169
PaidLikes by itself says, they say that[br]they have 30000 registered users, and it's
0:08:36.169,0:08:41.070
really interesting because you might think[br]that they are all registered with 10 or 15
0:08:41.070,0:08:45.620
accounts, but most of them, they are not.[br]They are clicking with their real account,
0:08:45.620,0:08:57.529
which makes it really hard to detect them.[br]So they even scan their I.D. so that the
0:08:57.529,0:09:03.210
company knows that they are real. Then[br]they earn their money. And we met men,
0:09:03.210,0:09:09.760
women, stay-at-home moms, low-income[br]earners, retirees, people who are getting
0:09:09.760,0:09:17.850
social care. So, basically, anybody. There[br]was no kind of bias. And many of them are
0:09:17.850,0:09:24.890
clicking for two and more platforms. That[br]was, I didn't meet anybody who's only
0:09:24.890,0:09:29.370
clicking for one platform. They all have a[br]variety of platforms where they are
0:09:29.370,0:09:34.610
writing comments or clicking likes. And[br]you can make - this is what they told us -
0:09:34.610,0:09:41.580
between 15 euro and 450 euro monthly, if[br]you are a so-called power clicker and you
0:09:41.580,0:09:48.410
do this some kind of professional. But[br]this are only the workers, and maybe you
0:09:48.410,0:09:52.740
are more interested in who are the buyers?[br]Who benefits?
0:09:52.740,0:09:59.631
Dennis: Yeah. Let's come to step two. Who[br]benefits from the campaigns? So I think
0:09:59.631,0:10:06.089
you all remember this page. This is the[br]screen if you log into PaidLikes and,
0:10:06.089,0:10:14.490
you'll see the campaigns with, you have to[br]click in order to get a little bit of
0:10:14.490,0:10:25.370
money. And by luck we've noticed that if[br]you go over a URL, we see in the left
0:10:25.370,0:10:31.980
bottom side of the browser, a URL[br]redirecting to the campaign. You have to
0:10:31.980,0:10:40.700
click and you see that every campaign is[br]using a unique ID. It is just a simple
0:10:40.700,0:10:49.640
integer, and the good thing is, it is just[br]incremented. So now maybe some of you guys
0:10:49.640,0:10:56.570
notice what we can do with that. And yeah,[br]it is really easy with these constructed
0:10:56.570,0:11:02.670
URLs to implement a crawler for data[br]gathering, and our crawler simply
0:11:02.670,0:11:11.931
requested all campaign IDs between 0 and[br]90000. Maybe some of you ask why 90000? As
0:11:11.931,0:11:17.110
I already said, we were also registered as[br]click workers and we see, we saw that the
0:11:17.110,0:11:24.779
highest ID campaign used is about 88000.[br]So we thought OK, 90000 is a good value
0:11:24.779,0:11:30.540
and we check for every request between[br]these 90000 requests if it got resolved or
0:11:30.540,0:11:36.030
not, and if it got resolved, we redirected[br]the URL we present this source. That
0:11:36.030,0:11:42.431
should be liked or followed. And we did[br]not save the page sources from the
0:11:42.431,0:11:50.750
resolved URLs, we only save the resolved[br]URLs in the list of campaigns, and this
0:11:50.750,0:11:58.700
list was then the basis for further[br]analysis. And here you see our list.
0:11:58.700,0:12:05.740
Svea: Yes. This was the point when Dennis[br]and Philip, when they came to us and said,
0:12:05.740,0:12:12.000
hey, we have a list. So what can you find?[br]And of course we searched AfD, was one of
0:12:12.000,0:12:20.940
the first search queries. And yeah, of[br]course, AfD is also in that list. Maybe
0:12:20.940,0:12:31.149
not so surprisingly for some. And when you[br]look, it is AFD Gelsenkirchen. And the fan
0:12:31.149,0:12:39.589
page. And we asked AfD Gelsenkirchen, did[br]you buy likes? And they said, we don't
0:12:39.589,0:12:48.240
know how we got on that list. But however,[br]we do not rule out an anonymous donation.
0:12:48.240,0:12:55.410
But now you would think, Ok, they found[br]AfD; this is very expectable. But no, all
0:12:55.410,0:13:00.930
political parties – mostly local and[br]regional entities - showed up on that
0:13:00.930,0:13:09.250
list. So we have CDU/CSU. We have had FDP,[br]SPD, AfD, Die Grünen and Die Linke. But
0:13:09.250,0:13:15.390
not that you think Angela Merkel or some[br]very big Facebook fan pages just showed
0:13:15.390,0:13:23.800
up. No, no. Very small entities with a[br]couple of hundreds or maybe 10000 or 15000
0:13:23.800,0:13:28.390
followers. And I think this makes[br]perfectly sense, because somebody who has
0:13:28.390,0:13:35.370
already very, very much many fans[br]probably would not buy them there at
0:13:35.370,0:13:46.311
PaidLikes. And we asked many of them, and[br]mostly they could not explain it. They
0:13:46.311,0:13:52.040
would never do something like that. Yeah,[br]they were completely over asked. But you
0:13:52.040,0:13:56.690
have to think that we only saw the[br]campaign. The campaigns, their Facebook
0:13:56.690,0:14:03.110
fan pages, we could not see who bought the[br]likes. And as you can imagine, everybody
0:14:03.110,0:14:08.740
could have done it like the mother, the[br]brother, the fan, you know, the dog. So
0:14:08.740,0:14:15.160
this was a case we would have needed a lot[br]of luck to call anybody out of the blue
0:14:15.160,0:14:20.260
and then he would say, oh, yes, I did[br]this. And there was one, or there were
0:14:20.260,0:14:25.810
some politicians who admitted it. And one[br]of them, she did it also publicly and gave
0:14:25.810,0:14:35.339
us an interview. It's Tanja Kühne. She is[br]a regional politician from Walsrode,
0:14:35.339,0:14:40.260
Niedersachsen. And she was in the..., it[br]was the case that it was after an election
0:14:40.260,0:14:44.360
and she was not very happy with her fan[br]page. That is what she told us. She was
0:14:44.360,0:14:49.220
very unlucky and she wanted, you know, to[br]push herself and to boost it a little bit,
0:14:49.220,0:14:55.510
and get more friends and followers and[br]reach. And then she bought 500 followers.
0:14:55.510,0:15:02.870
And then we had a nice interview with her[br]about that. Show you a small piece.
0:15:53.829,0:15:59.760
Okay, so you see – answers are pretty[br]interesting. And she.. I think she was
0:15:59.760,0:16:05.180
that courageous to speak out to us. Many[br]of others did too, but only on the phone.
0:16:05.180,0:16:09.180
And they didn't want to go on the record.[br]But she's not the only one who answered
0:16:09.180,0:16:14.110
like this. Because, of course, if you call[br]through a list of potential fake like
0:16:14.110,0:16:21.120
buyers, of course they answer like, no,[br]it's not a scam. And I also think from a
0:16:21.120,0:16:26.180
jurisdictional way, it's it's also very[br]hard to show that this is fraud and a
0:16:26.180,0:16:33.209
scam. And it's more an ethical problem[br]that you can that you can see here, that
0:16:33.209,0:16:40.170
it's manipulative if you buy likes. We[br]also found a guy from FSP from the
0:16:40.170,0:16:45.269
Bundestag. But yeah, he ran away and[br]didn't want to get interviewed, so I
0:16:45.269,0:16:52.700
couldn't show you. So bought, or no[br]probably... He was like 40 times in our
0:16:52.700,0:16:59.100
list for various Facebook posts and videos[br]and also for his Instagram account. But we
0:16:59.100,0:17:06.730
could not get him on, we could not get him[br]on record. So what did others say? We, of
0:17:06.730,0:17:10.970
course, confronted Facebook, Instagram and[br]YouTube with this small research. And they
0:17:10.970,0:17:18.079
said, no, we don't want fake likes on our[br]platform. PaidLikes is active since 2012,
0:17:18.079,0:17:25.370
you know. So they waited seven years. But[br]after our report, at least, Facebook
0:17:25.370,0:17:32.549
temporarily blocked PaidLikes. And of[br]course, we asked them too, and spoke to
0:17:32.549,0:17:35.781
them and wrote with PaidLikes in[br]Magdeburg. And they said, of course, it's
0:17:35.781,0:17:41.620
not a scam because the click workers they[br]are freely clicking on pages. So, yeah,
0:17:41.620,0:17:47.640
kind of nobody cares. But PaidLikes, this[br]is only the tip of the iceberg.
0:17:47.640,0:17:58.520
Philip: So we also wanted to dive a little[br]bit into this fake like universe outside
0:17:58.520,0:18:05.780
of PaidLikes and to see what else is out[br]there. And so we did an analysis of
0:18:05.780,0:18:12.780
account creation on Facebook. So what[br]Facebook is saying about account creation
0:18:12.780,0:18:19.299
is that they are very effective against[br]fake accounts. So they say they remove
0:18:19.299,0:18:26.330
billions of accounts each year, and that[br]most of these accounts never reach any
0:18:26.330,0:18:33.000
real users and they remove them before[br]they get reported. So what Facebook
0:18:33.000,0:18:39.080
basically wants to tell you is that they[br]have it under control. However, there are
0:18:39.080,0:18:45.700
a number of reports that suggest[br]otherwise. For example, recently at NATO-
0:18:45.700,0:18:53.630
Stratcom Taskforce released a report where[br]they actually bought 54000 likes, 54000
0:18:53.630,0:19:02.220
social media interactions for just 300[br]Euros. So this is a very low price. And I
0:19:02.220,0:19:07.169
think you wouldn't expect such a low price[br]if it would be hard to get that many
0:19:07.169,0:19:15.880
interactions. They bought 3500 comments,[br]25000 likes, 20000 views and 5100
0:19:15.880,0:19:22.991
followers. Everything for just 300 Euros.[br]So, you know, the thing they have in
0:19:22.991,0:19:32.050
common, they are cheap, the fake likes and[br]the fake interactions. So we also have,
0:19:32.050,0:19:38.470
there was also another report from Vice[br]Germany recently. And they reported on
0:19:38.470,0:19:46.410
some interesting facts about automated[br]fake accounts. They reported on findings
0:19:46.410,0:19:50.980
that suggest that actually people use[br]internet or hacked internet of things
0:19:50.980,0:19:59.150
devices and to use them to create these[br]fake accounts and to manage them. And so
0:19:59.150,0:20:04.590
it's actually kind of interesting to think[br]about this this wa. To say, OK, maybe next
0:20:04.590,0:20:11.020
election your fridge is actually going to[br]support the other candidate on Facebook.
0:20:11.020,0:20:16.970
And so we also wanted to look into this[br]and we wanted to go a step further and to
0:20:16.970,0:20:24.660
look at who these people are. Who are[br]they, and what what are they doing on
0:20:24.660,0:20:32.200
Facebook? And so we actually examined the[br]profiles of purchased likes. For this we
0:20:32.200,0:20:38.390
created four comments under arbitrary[br]posts, and then we bought likes for these
0:20:38.390,0:20:46.500
comments, and then we examined the[br]resulting profiles of the fake likes. So
0:20:46.500,0:20:51.050
it was pretty cheap to buy these likes.[br]Comment likes are always a little bit more
0:20:51.050,0:20:59.520
expensive than other likes. And we found[br]all these offerings on Google and we paid
0:20:59.520,0:21:08.169
with PayPal. So we actually used a pretty[br]neat trick to estimate the age of these
0:21:08.169,0:21:16.490
fake accounts. So as you can see here, the[br]Facebook user ID is incremented. So
0:21:16.490,0:21:24.250
Facebook started in 2009 to use[br]incremented Facebook ID, and they use this
0:21:24.250,0:21:31.780
pattern of 1 0 0 0 and then the[br]incremented number. And as you can see, in
0:21:31.780,0:21:40.200
2009 this incremented number was very[br]close to zero. And then today it is close
0:21:40.200,0:21:49.559
to 40 billion. And in this time period,[br]you can see that you can kind of get a
0:21:49.559,0:21:56.770
rather fitting line through all these[br]points. And you can see that the likes are
0:21:56.770,0:22:02.710
in fact incremented, ... the account IDs[br]are in fact incremented over time. So we
0:22:02.710,0:22:08.670
can use this fact in reverse to estimate[br]the creation date of an account where we
0:22:08.670,0:22:15.340
know the Facebook ID. And that's exactly[br]what we did with these fake likes. So we
0:22:15.340,0:22:22.090
estimated the account creation dates. And[br]as you can see, we get kind of different
0:22:22.090,0:22:28.929
results from different services. For[br]example, PaidLikes, they had rather old
0:22:28.929,0:22:35.750
accounts. So this means they use very[br]authentic accounts. And we already know
0:22:35.750,0:22:41.370
that because we talked to them. So these[br]are very authentic accounts. Also like
0:22:41.370,0:22:46.660
Service A over here also uses very, very[br]authentic accounts. But on the other hand,
0:22:46.660,0:22:52.160
like service B uses very new accounts,[br]they were all created in the last three
0:22:52.160,0:22:58.280
years. So if you look at the accounts and[br]also from these numbers, we think that
0:22:58.280,0:23:06.510
these accounts were bots and on service C[br]it's kind of not clear, are these are
0:23:06.510,0:23:10.870
these accounts bots or are these[br]clickworkers? Maybe it's a mixture of
0:23:10.870,0:23:17.820
both, we don't know exactly for sure. But[br]this is an interesting metric to measure
0:23:17.820,0:23:23.390
the age of the accounts to determine if[br]some of them might be bots. And that's
0:23:23.390,0:23:29.340
exactly what we did on this page. So this[br]is actually a page for garden furniture
0:23:29.340,0:23:36.750
and we found it in our list that we got[br]from paid likes. So they bought, obviously
0:23:36.750,0:23:43.970
they were on this list for bought likes on[br]Facebook, on PaidLikes. And they caught
0:23:43.970,0:23:51.000
our eye because they had one million[br]likes. And that's rather unusual for a
0:23:51.000,0:24:01.260
shop for garden furniture in Germany. And[br]so we looked at this page further and we
0:24:01.260,0:24:07.390
noticed other interesting things. For[br]example, there are posts, all the time,
0:24:07.390,0:24:13.820
they got like thousands of likes. And[br]that's also kind of unusual for a garden
0:24:13.820,0:24:19.590
furniture shop. And so we looked into the[br]likes and as you can see, they all look
0:24:19.590,0:24:26.790
like they come from Southeast Asia and[br]they don't look very authentic. And we
0:24:26.790,0:24:32.460
were actually able to estimate the[br]creation dates of these accounts. And we
0:24:32.460,0:24:36.700
found that most of these accounts that[br]were used for liking these posts on this
0:24:36.700,0:24:44.130
page were actually created in the last[br]three years. So this is a page where
0:24:44.130,0:24:49.540
everything, from the number of people who[br]like to page to the number of people who
0:24:49.540,0:24:55.559
like to posts is complete fraud. So[br]nothing about this is real. And it's
0:24:55.559,0:25:02.380
obvious that this can happen on Facebook[br]and that this is a really, really big
0:25:02.380,0:25:08.309
problem. I mean, this is a, this is a shop[br]for garden furniture. Obviously, they
0:25:08.309,0:25:14.580
probably don't have such huge sums of[br]money. So it was probably very cheap to
0:25:14.580,0:25:22.170
buy this amount of fake accounts. And it[br]is really shocking to see how, how big,
0:25:22.170,0:25:31.179
how big the scale is of this kind of[br]operations. And so what we have to say is,
0:25:31.179,0:25:39.970
OK, when Facebook says they have it under[br]control, we have to doubt that. So now we
0:25:39.970,0:25:46.320
can look at the bigger picture. And what[br]we are going to do here is we are going to
0:25:46.320,0:25:52.700
use this same graph that we used before to[br]estimate the creation dates, but in a
0:25:52.700,0:25:59.080
different way. So we can actually see that[br]the lowest and the highest points of
0:25:59.080,0:26:05.090
Facebook IDs in this graph. So we know the[br]newest Facebook ID by creating a new
0:26:05.090,0:26:13.200
account. And we know the lowest ID because[br]it's zero. And then we know that there are
0:26:13.200,0:26:20.780
40 billion Facebook IDs. Now, in the next[br]step, we took a sample, a random sample
0:26:20.780,0:26:27.610
from these 40 billion Facebook IDs. And[br]inside of the sample, we checked if these
0:26:27.610,0:26:33.740
accounts exist, if this ID corresponds to[br]an existing account. And we do that because
0:26:33.740,0:26:39.360
we obviously cannot check 40 billion[br]accounts and 40 billion IDs, but we can
0:26:39.360,0:26:45.720
check a small sample of these accounts of[br]these IDs and estimate, then, the number
0:26:45.720,0:26:54.470
of existing accounts on Facebook and[br]total. So for this, we repeatedly access
0:26:54.470,0:27:02.770
the same sample of one million random IDs[br]over the course of one year. And we also
0:27:02.770,0:27:10.100
pulled a sample of 10 million random IDs[br]for closer analysis this July. And now
0:27:10.100,0:27:15.950
Dennis is going to tell you how we did it.[br]Dennis: Yeah. Well, pretty interesting,
0:27:15.950,0:27:21.160
pretty interesting results so far, right?[br]So we again implemented the crawler, the
0:27:21.160,0:27:26.530
second time for gathering public Facebook[br]information, the public Facebook account
0:27:26.530,0:27:35.730
data. And, yeah, this was not so easy as[br]in the first case. Um, yeah. As. It's not
0:27:35.730,0:27:45.059
surprising that Facebook is using a lot of[br]measures to try to block the automated
0:27:45.059,0:27:52.460
crawling of the Facebook page, for example[br]with IP blocking or CAPTCHA solving. But,
0:27:52.460,0:27:59.929
uh, we were pretty easy... Yeah, we could[br]pretty easy solve this problem by using
0:27:59.929,0:28:06.980
the Tor Anonymity Network. So every time[br]our IP got blocked by crawling the data,
0:28:06.980,0:28:14.480
we just made a new Tor connection and[br]change the IP. And this also with the
0:28:14.480,0:28:21.440
CAPTCHAs. And with this easy method, we[br]were able to to crawl all the Facebook,
0:28:21.440,0:28:26.020
and all the public Facebook data. And[br]let's have a look at two examples. The
0:28:26.020,0:28:36.890
first example is facebook.com/4. So the,[br]very, very small Facebook ID. Yeah, in
0:28:36.890,0:28:41.790
this case, we are, we are redirected and[br]check the response and find a valid
0:28:41.790,0:28:50.070
account page. And does anyone know which[br]account this is? Mark Zuckerberg? Yeah,
0:28:50.070,0:28:55.360
that's correct. This is this is a public[br]account for Mark Zuckerberg. Number four,
0:28:55.360,0:29:01.679
as we see, as we already saw, the other[br]IDs are really high. But he got the number
0:29:01.679,0:29:10.690
four. Second example was facebook.com/3.[br]In this case, we are not forwarded. And
0:29:10.690,0:29:17.760
this means that it is an invalid account.[br]And that was really easy to confirm with a
0:29:17.760,0:29:23.740
quick Google search. And it was a test[br]account from the beginning of Facebook. So
0:29:23.740,0:29:31.059
we did not get redirected. And it's just[br]the login page from Facebook. And with
0:29:31.059,0:29:38.500
these examples, we did, we did a lot of, a[br]lot more experiments. And at the end, we
0:29:38.500,0:29:46.970
were able to to build this tree. And, yeah, [br]this tree represents the high level
0:29:46.970,0:29:53.059
approach from our scraper. So in the,[br]What's that?
0:29:53.059,0:29:56.340
Svea: Okay. Sleeping.[br]Laughing
0:29:56.340,0:30:07.090
Dennis: Yeah. We have still time. Right.[br]So what? Okay, so everyone is waking up
0:30:07.090,0:30:16.680
again. Oh, yeah. The first step we call[br]the domain, www.facebook.com/FID. If we
0:30:16.680,0:30:24.650
get redirected in this case, then we check[br]if the, if the page is an account page. If
0:30:24.650,0:30:31.270
it's an account page, then it's an public[br]account like the example 4 and we were
0:30:31.270,0:30:39.890
able to save the raw data, the raw HTTP[br]source. If we, if it's not an account page
0:30:39.890,0:30:45.070
then everything is OK. If it's not, it's[br]not a public account and we are not able
0:30:45.070,0:30:52.580
to save any data. And if we call, if we[br]do, if we do not get redirected in the
0:30:52.580,0:31:01.630
first step, then we call the second[br]domain, facebook.com/profile.php?id=FID
0:31:01.630,0:31:09.289
with the mobile user agent. And if we get[br]redirected then, then again, it is a
0:31:09.289,0:31:14.990
nonpublic profile and we cannot save[br]anything. But, and if we get not
0:31:14.990,0:31:22.710
redirected, it is an invalid profile and[br]it is most often a deleted account. Yeah.
0:31:22.710,0:31:29.390
And yeah, that's the high level overview[br]of our scraper. And Phillip will now give
0:31:29.390,0:31:32.340
some more information on interesting[br]results.
0:31:32.340,0:31:38.820
Phillip: So the most interesting result of[br]this scraping of the sample of Facebook
0:31:38.820,0:31:47.070
IDs was that one in four Facebook IDs[br]corresponds to a valid account. And you
0:31:47.070,0:31:53.559
can do the math. There are 40 billion[br]Facebook IDs, so there must be 10 billion
0:31:53.559,0:32:00.170
registered users on Facebook. And this[br]means that there are more registered users
0:32:00.170,0:32:08.140
on Facebook than there are humans on[br]Earth. And also, it means that it's even
0:32:08.140,0:32:12.460
worse than that because not everybody on[br]Earth can have a Facebook account because
0:32:12.460,0:32:17.370
not everybody, you need a smartphone for[br]that. And many people don't have those. So
0:32:17.370,0:32:22.270
this is actually a pretty high number and[br]it's very unexpected. So in July 2019,
0:32:22.270,0:32:29.059
there were more than ten billion Facebook[br]accounts. Also, we did another research on
0:32:29.059,0:32:36.429
the timeframe between October 2018 and[br]today, or this month. And we found that in
0:32:36.429,0:32:43.140
this timeframe there were 2 billion new[br]registered Facebook accounts. So this is
0:32:43.140,0:32:48.679
like the timeframe of one year, more or[br]less. And in a similar timeframe, the
0:32:48.679,0:32:58.899
monthly active user base rose by only 187[br]million. Facebook deleted 150 million
0:32:58.899,0:33:05.419
older accounts between October 2018 and[br]July 2019. And we know that because we
0:33:05.419,0:33:11.460
pulled the same sample over a longer[br]period of time. And then we watched for
0:33:11.460,0:33:16.230
accounts that got deleted in the sample.[br]And that enables us to estimate this
0:33:16.230,0:33:23.400
number of 150 million accounts that got[br]deleted that are basically older than our
0:33:23.400,0:33:31.890
sample. So I made some nice graphs for[br]your viewing pleasure. So, again, the
0:33:31.890,0:33:40.919
older accounts were, just 150 million were[br]deleted since October 2018. These are
0:33:40.919,0:33:46.350
accounts that are older than last year.[br]And Facebook claims that since then, about
0:33:46.350,0:33:52.789
7 billion accounts got deleted from their[br]platform, which is vastly more than these
0:33:52.789,0:33:58.370
older accounts. And that that's why we[br]think that Facebook mostly deleted these
0:33:58.370,0:34:06.770
newer accounts. And if an account is older[br]than a certain age, then it is very
0:34:06.770,0:34:13.069
unlikely that it gets deleted. And also, I[br]think you can see the scales here. So, of
0:34:13.069,0:34:17.960
course, the registered users are not the[br]same thing as active users, but you can
0:34:17.960,0:34:23.290
still see that there are much more[br]registrations of, of new users than there
0:34:23.290,0:34:30.139
are active users. And there are new active[br]users during the last year. So what does
0:34:30.139,0:34:37.909
this all mean? Does it mean that Facebook[br]gets flooded by fake accounts? We don't
0:34:37.909,0:34:42.980
really know. We only know these numbers.[br]What Facebook is telling us is that they
0:34:42.980,0:34:50.409
only count and publish active users, as I[br]already said, that there is a disconnect
0:34:50.409,0:34:56.759
between this record, registered users and[br]active users and Facebook only reports on
0:34:56.759,0:35:04.289
the active users. Also, they say that[br]users register accounts, but they don't
0:35:04.289,0:35:10.519
verify them or they don't use them, and[br]that's how this number gets so high. But I
0:35:10.519,0:35:19.319
think that that's not really explaining[br]these high numbers and because that's just
0:35:19.319,0:35:26.469
by orders of magnitude larger than[br]anything that this could cause. Also, they
0:35:26.469,0:35:31.819
say that they regularly delete fake[br]accounts. But we have seen that these are
0:35:31.819,0:35:37.519
mostly accounts that get deleted directly[br]after their creation. And if they survive
0:35:37.519,0:35:46.170
long enough, then they are getting[br]through. So what does this all mean?
0:35:46.170,0:35:55.390
Svea: Okay, so you got the full load,[br]which I had like over two or three months.
0:35:55.390,0:36:02.869
And what for me was, was a one very big[br]conclusion was that we have some kind of
0:36:02.869,0:36:08.530
broken metric here, that all the likes and[br]all the hearts on Instagram and the
0:36:08.530,0:36:13.650
followers that they can so easily be[br]manipulated. And then it's it's so hard to
0:36:13.650,0:36:19.029
tell in some cases, it's so hard to tell[br]if they are real or not real. And this
0:36:19.029,0:36:26.160
opens the gate for manipulation and yes,[br]untrueness. And for economic losses, if
0:36:26.160,0:36:33.109
you think as somebody who is investing[br]money and or as an advertiser, for
0:36:33.109,0:36:40.170
example. And in the very end, it is a case[br]of eroding trust, which means that we
0:36:40.170,0:36:45.739
cannot trust these numbers anymore. These[br]numbers are, you know, they are so easily
0:36:45.739,0:36:53.799
manipulated. And why should we trust this?[br]And this has a severe consequence for all
0:36:53.799,0:36:59.420
the social networks. If you are still in[br]them. So what can be a solution? And
0:36:59.420,0:37:05.150
Philip, you thought about that.[br]Phillip: So basically we have two
0:37:05.150,0:37:11.410
problems. One is click workers and one is[br]fakes. Click workers are basically just
0:37:11.410,0:37:18.420
hyper active users and they are selling[br]their hyper activity. And so what social
0:37:18.420,0:37:23.660
networks could do is just make[br]interactions scarce, so just lower the
0:37:23.660,0:37:29.180
value of more interactions. If you are a[br]hyper active users, then your interaction
0:37:29.180,0:37:34.240
should count less than the interactions of[br]a less active user.
0:37:34.240,0:37:39.229
Mumbling[br]That's kind of solvable, I think. The real
0:37:39.229,0:37:46.890
problem is the authenticity. So if you if[br]you get stopped from posting or liking
0:37:46.890,0:37:52.640
hundreds of pages a day, then maybe you[br]just create multiple accounts and operate
0:37:52.640,0:37:58.599
them simultaneously. And this can only be[br]solved by authenticity. So this can only
0:37:58.599,0:38:04.990
be solved if you know that the person who[br]is operating the account is just one
0:38:04.990,0:38:10.569
person, is operating one account. And this[br]is really hard to do, because Facebook
0:38:10.569,0:38:14.940
doesn't know who is clicking. Is it a bot?[br]Is it a clickworrker, or is it one
0:38:14.940,0:38:20.410
clickworker for ten accounts? How does[br]this work? And so this is really hard for
0:38:20.410,0:38:27.609
the, for the social media companies to do.[br]And you could say, OK, let's send in the
0:38:27.609,0:38:32.359
passport or something like that to prove[br]authenticity. But that's actually not a
0:38:32.359,0:38:37.109
good idea because nobody wants to send[br]their passport to Facebook. And so this is
0:38:37.109,0:38:42.359
really a hard problem that has to be[br]solved. If we want to use social, social
0:38:42.359,0:38:49.750
media in a meaningful way. And so this is[br]what, what companies could do. And now...
0:38:49.750,0:38:53.200
Svea: But what do what you[br]could do. Okay. Of course, you can delete
0:38:53.200,0:38:56.469
your Facebook account or your Instagram[br]account and stop.
0:38:56.469,0:39:01.299
Slight Applause, Lauthing[br]Svea: Yeah! Stay away from social media.
0:39:01.299,0:39:08.959
But this maybe is not for all of us a[br]solution. So I think be aware, of course.
0:39:08.959,0:39:17.499
Spread the word, tell others. And if, if[br]you, if you like, then and you get more
0:39:17.499,0:39:24.019
intelligence about that, we are really[br]happy to dig deeper in these networks. And
0:39:24.019,0:39:30.180
and we will go on investigating and so at[br]last but not least, it's to say thank you
0:39:30.180,0:39:33.349
to you guys. Thank you very much for[br]listening.
0:39:33.349,0:39:40.089
Applause[br]Svea: And we did not do this alone. We are
0:39:40.089,0:39:44.849
not three people. There are many more[br]standing behind and doing this, this
0:39:44.849,0:39:50.709
beautiful research. And we are opening now[br]for questions, please.
0:39:50.719,0:39:55.429
Herald: Yes. Please, thank Svea, Phil and[br]Dennis again.
0:39:55.429,0:40:05.519
Applause[br]And we have microphones out
0:40:05.519,0:40:09.680
here in the room, about nine of them,[br]actually. If you line up behind them to
0:40:09.680,0:40:15.780
ask a question, remember that a question[br]is a sentence with a question mark behind
0:40:15.780,0:40:20.500
it. And I think I see somebody at number[br]three. So let's start with that.
0:40:20.500,0:40:25.979
Question: Hi. I, I just have a little[br]question. Wouldn't a dislike button, the
0:40:25.979,0:40:30.749
concept of a dislike button, wouldn't that[br]be a solution to all the problems?
0:40:30.749,0:40:38.039
Phillip: So we thought about recommending[br]that Facebook ditches the like button
0:40:38.039,0:40:42.299
altogether. I think that would be a better[br]solution than a dislike button, because a
0:40:42.299,0:40:47.079
dislike button could also be manipulated[br]and it would be even worse because you
0:40:47.079,0:40:54.119
could actually manipulate the network into[br]down ranking posts or kind of not showing
0:40:54.119,0:41:00.670
posts to somebody. And that, I think would[br]be even worse. I imagine what dictators
0:41:00.670,0:41:08.209
would do with that. And so I think the[br]best option would be to actually not show
0:41:08.209,0:41:18.029
off like, like counts anymore and to this,[br]to actually make people not invest into
0:41:18.029,0:41:25.199
these counts if they become meaningless.[br]Herald: I think I see a microphone 7, up
0:41:25.199,0:41:28.109
there.[br]Question: Hello. So one question I had is
0:41:28.109,0:41:37.210
you are signed creation dates to IDs. How[br]did you do this?
0:41:37.210,0:41:52.489
Phillip: So, we actually knew the creation[br]date of some accounts. And then we kind of
0:41:52.489,0:41:58.210
interpolated between the creation dates[br]and the IDs. So you see this black line
0:41:58.210,0:42:04.109
there. That's actually our, our[br]interpolation. And with this black line,
0:42:04.109,0:42:10.910
we can then estimate the creation dates[br]for IDs that we do not yet know because
0:42:10.910,0:42:17.430
they did, kind of fill in the gaps.[br]Q: Follow up question, do you know why
0:42:17.430,0:42:20.310
there are some points outside of this[br]graph?
0:42:20.310,0:42:23.999
Phillip: No.[br]Q: No? Thank you.
0:42:23.999,0:42:26.400
Herald: So there was a question from the[br]Internet.
0:42:26.400,0:42:33.723
Question: Did you report your findings to[br]Facebook? And did they do anything?
0:42:33.723,0:42:41.509
Svea: Because this research is very new,[br]we, we just recently approached them and
0:42:41.509,0:42:47.190
showed them the research and we got an[br]answer. But I think we also already showed
0:42:47.190,0:42:54.480
the answer. It was that they, I think that[br]they only count and publish active users.
0:42:54.480,0:42:59.680
They could, they did not want to tell us[br]how many registered users they have, that
0:42:59.680,0:43:03.859
they say, oh, sometimes users register[br]accounts, but don't use them or verify
0:43:03.859,0:43:08.930
them. And that they regularly delete fake[br]accounts. But we hope that we get into a
0:43:08.930,0:43:12.469
closer discussion with them soon about[br]this.
0:43:12.469,0:43:19.469
Herald: Microphone two.[br]Question: When hunting down the bias of
0:43:19.469,0:43:26.740
the campaigns, did you dig out your own[br]campaign line, Line below the line? No,
0:43:26.740,0:43:34.039
because they stopped scraping in August.[br]And I, you stopped scraping in August. And
0:43:34.039,0:43:39.449
then I started, you know, the whole[br]project started with them coming to us
0:43:39.449,0:43:44.599
with the list. And then we thought, oh,[br]this is very interesting. And then the
0:43:44.599,0:43:50.729
whole journalistic research started. And,[br]but I think if we, I think if we would do
0:43:50.729,0:43:56.200
it again, of course, I think we would find[br]us. We all also found there was another
0:43:56.200,0:44:01.650
magazine, and they did, also a test, paid[br]test a couple of years ago. And we found
0:44:01.650,0:44:04.920
their campaign.[br]Phillip: So, so we we actually did another
0:44:04.920,0:44:11.480
test. And for the other test, I noted we[br]also got like this ID, I think. And it
0:44:11.480,0:44:20.329
worked to plug it into the URL and then we[br]also got to redirected to our own page. So
0:44:20.329,0:44:22.569
that worked.[br]Q: Thank you.
0:44:22.569,0:44:26.379
Herald: Microphone three.[br]Question: Hi. I'm Farhan, I'm a Pakistani
0:44:26.379,0:44:30.759
journalist. And first of all, I would like[br]to say that you were right when you said
0:44:30.759,0:44:34.910
that there might be people sitting in[br]Pakistan clicking on the likes. That does
0:44:34.910,0:44:41.329
happen. But my question would be that[br]Facebook does have its own ad program that
0:44:41.329,0:44:47.470
it aggressively pushes. And in that ad[br]program, there is also options whereby
0:44:47.470,0:44:53.701
people can buy likes and comments and[br]impressions and reactions. Did you, would
0:44:53.701,0:44:59.670
you also consider those as a fake? I mean,[br]that they're not fake, per se, but they're
0:44:59.670,0:45:05.799
still bought likes. So what's your view on[br]those? Thank you.
0:45:05.799,0:45:14.349
Phillip: So, when you buy ads on Facebook,[br]then, so, what you what you actually want
0:45:14.349,0:45:19.489
to have is fans for your page that are[br]actually interested in your page. So
0:45:19.489,0:45:25.460
that's kind of the difference, I think to[br]the, to the paid likes system where the
0:45:25.460,0:45:30.119
people themselves, they get paid for[br]liking stuff that they wouldn't normally
0:45:30.119,0:45:35.599
like. So I think that's the fundamental[br]difference between the two programs. And
0:45:35.599,0:45:40.529
that's why I think that one is unethical.[br]And one is not really that unethical.
0:45:40.529,0:45:47.749
Svea: The very problem is if you, if you[br]buy these click workers, then you have
0:45:47.749,0:45:52.789
many people in your fan page. They are not[br]interested in you. They don't care about
0:45:52.789,0:45:57.410
you. They don't look at your products.[br]They don't look at your political party.
0:45:57.410,0:46:03.539
And then often the people, they[br]additionally, they make Facebook ads, and
0:46:03.539,0:46:08.229
these ads, they are shown, again, the[br]click workers and they don't look at them.
0:46:08.229,0:46:13.410
So, you know, people, they are burning[br]money and money and money with this whole
0:46:13.410,0:46:18.069
corrupt system.[br]Herald: So, microphone two.
0:46:18.069,0:46:22.039
Question: Hi. Thanks. Thanks for the talk[br]and thanks for the effort of going through
0:46:22.039,0:46:27.709
all of this project. From my[br]understanding, this whole finding
0:46:27.709,0:46:35.209
basically undermines the trust in[br]Facebook's likes in general, per se. So I
0:46:35.209,0:46:42.369
would expect now the price of likes to[br]drop and the pay for click workers to drop
0:46:42.369,0:46:49.250
as well. Do you have any metrics on that?[br]Svea: The research just went public. I
0:46:49.250,0:46:56.180
think one week ago. So, so what we have[br]seen as an effect is that Facebook, they
0:46:56.180,0:47:02.940
excluded paid likes for, for a moment. So,[br]yes, of course, one platform is down. But
0:47:02.940,0:47:08.010
I think there are so many outside. There[br]are so many. So I think...
0:47:08.010,0:47:14.229
Q: I meant the phenomenon of paid likes,[br]not the company itself. Like the value of
0:47:14.229,0:47:19.319
a like as a measure of credibility...[br]Phillip: We didn't...
0:47:19.319,0:47:22.829
Q: ...is declining now. That's my, that's[br]my...
0:47:22.829,0:47:27.869
Svea: Yes. That's why many people are[br]buying Instagram hearts now. So, so, yes,
0:47:27.869,0:47:32.900
that's true. The like is not the fancy hot[br]shit anymore. Yes. And we also saw in the
0:47:32.900,0:47:40.670
data that the likes for the fan pages,[br]they rapidly went down and the likes for
0:47:40.670,0:47:45.229
the posts and the comments, they went up.[br]So I think, yes, there is a shift. And
0:47:45.229,0:47:51.809
what we also saw in that data was that the[br]Facebook likes, they, they went down from
0:47:51.809,0:47:57.839
2016. They are rapidly down. And what is[br]growing and rising is YouTube and
0:47:57.839,0:48:01.609
Instagram. Now, everything is about,[br]today, everything is about Instagram.
0:48:01.609,0:48:05.270
Q: Thanks.[br]Herald: So let's go to number one.
0:48:05.270,0:48:09.630
Question: Hello and thank you very much[br]for this fascinating talk, because I've
0:48:09.630,0:48:15.400
been following this whole topic for a[br]while. And I was wondering if you were
0:48:15.400,0:48:20.849
looking also into the demographics, in[br]terms of age groups and social class, not
0:48:20.849,0:48:25.619
of the people who were doing the actual[br]liking, but actually, you know, buying
0:48:25.619,0:48:31.249
these likes. Because I think that what is[br]changing is an entire social discourse on
0:48:31.249,0:48:36.709
social capital and, the bold U.S. kind of[br]term, because it can now be quantified. As
0:48:36.709,0:48:43.650
a teacher, I hear of kids who buy likes to[br]be more popular than their other
0:48:43.650,0:48:47.880
schoolmates. So I'm wondering if you're[br]looking into that, because I think that's
0:48:47.880,0:48:52.559
fascinating, fascinating area to actually[br]come up with numbers about it.
0:48:52.559,0:48:59.229
Svea: It definitely is. And we were all so[br]fascinated by this data set of 90,000 data
0:48:59.229,0:49:05.479
points. And what we did was, and this was[br]very hard, and was that we tried it, first
0:49:05.479,0:49:11.869
of all, to look who is buying likes, like[br]automotives, you know, to to, this some,
0:49:11.869,0:49:18.910
you know, what, what kind of branches? Who[br]is in that? And so this was this was
0:49:18.910,0:49:24.769
doable. But to get more into demographics,[br]you would have liked to, to crawl, to
0:49:24.769,0:49:33.699
click every page. And so we we did not do[br]this. What we did was, of course, that we
0:49:33.699,0:49:38.489
that we were a team of three to ten people[br]and manually looking into it. And what we,
0:49:38.489,0:49:43.739
of course, saw that on Instagram and on[br]YouTube, you have many of these very young
0:49:43.739,0:49:47.219
people. Some of them, I actually called[br]them and they were like, Yes, I bought
0:49:47.219,0:49:54.089
likes. Very bad idea. So I think yes, I[br]think there is a demographic shift away
0:49:54.089,0:49:59.890
from the companies and the automotive and[br]industries buying Facebook fan page likes
0:49:59.890,0:50:04.390
to Instagram and YouTube wannabe-[br]influencers.
0:50:04.390,0:50:06.430
Q: Influencers, influencer culture is[br]obviously...
0:50:06.430,0:50:12.670
Svea: Yes. And I have to admit here we, we[br]showed you the political side, but we have
0:50:12.670,0:50:19.849
to admit that the political likes, they[br]were like this small in the numbers. And
0:50:19.849,0:50:25.640
the very, very vast majority of this data[br]set, it's about wedding planners,
0:50:25.640,0:50:31.440
photography, tattoo studios and[br]influencers, influencers, influencers and
0:50:31.440,0:50:34.479
YouTubers, of course.[br]Q: Yes. Thank you so much.
0:50:34.479,0:50:37.439
Herald: So we have a lot of questions in[br]the room. I'm going to get to you as soon
0:50:37.439,0:50:40.009
as we can. I'd like to go to the Internet[br]first.
0:50:40.009,0:50:44.680
Signal Angel: Do you think this will get[br]bit better or worse if people move to more
0:50:44.680,0:50:48.319
decentralized platforms?[br]Phillip: To more what?
0:50:48.319,0:50:54.910
Svea: If it get better or worse.[br]Dennis: Can you repeat that, please?
0:50:54.910,0:50:58.880
Herald: Would this issue get better or[br]worse if people move to a more
0:50:58.880,0:51:01.239
decentralized platform?[br]Phillip: Decentralized. decentralized,
0:51:01.239,0:51:12.160
okay. So, I mean, we can look at, at the,[br]this slide, I think, and think about
0:51:12.160,0:51:18.249
whether decentralized platforms would[br]change any of these, any of these two
0:51:18.249,0:51:25.999
points here. And I fear, I don't think so,[br]because they cannot solve the interactions
0:51:25.999,0:51:30.210
problem that people can be hyperactive.[br]Actually, that's kind of a normal thing
0:51:30.210,0:51:34.299
with social media. A small portion of[br]social media users is much more active
0:51:34.299,0:51:39.880
than everybody else. That's kind of. You[br]have that without paying for it. So
0:51:39.880,0:51:44.720
without even having paid likes, you will[br]have to consider if social media is really
0:51:44.720,0:51:51.189
kind of representative of the society.[br]But, and the other thing is authenticity.
0:51:51.189,0:51:57.170
And also in a decentralized platform, you[br]could have multiple accounts run by the
0:51:57.170,0:52:01.199
same person.[br]Herald: So, microphone seven, all the way
0:52:01.199,0:52:06.779
back there.[br]Question: Hi. Do you know if Facebook even
0:52:06.779,0:52:10.220
removes the likes when they delete fake[br]accounts?
0:52:10.220,0:52:17.319
Svea: Do you know that?[br]Phillip: No, we don't know that. No, we
0:52:17.319,0:52:21.259
don't. We don't know. We know they delete[br]fake accounts, but we don't know if they
0:52:21.259,0:52:27.619
also delete the likes. I know from our[br]research that the people we approached,
0:52:27.619,0:52:31.329
they did not delete the click workers.[br]They get...
0:52:31.329,0:52:35.839
Herald: Microphone two.[br]Question: Yeah. Hi. So I have a question
0:52:35.839,0:52:41.359
with respect to this, one out of four[br]Facebook accounts are active in your, in
0:52:41.359,0:52:46.949
your test. Did you see any difference with[br]respect to age of the accounts? So is it
0:52:46.949,0:52:52.489
always one out the four to the entire[br]sample? Or does it maybe change over the,
0:52:52.489,0:52:57.730
over the like going from a zero ID to,[br]well, 10 billion or 40 billion?
0:52:57.730,0:53:02.189
Phillip: So you're talking about the[br]density of accounts in our ID?
0:53:02.189,0:53:05.989
Q: Kind of.[br]Phillip: So, so there are changes over
0:53:05.989,0:53:12.150
time. Yeah. So I guess I think now it's[br]less than it was before. So now they are
0:53:12.150,0:53:19.089
less than for then, and before it was more[br]and so I think it was. Yeah. I don't know.
0:53:19.089,0:53:23.660
Q: But you don't see anything specific[br]that now, only in the new accounts, only
0:53:23.660,0:53:28.229
one out of 10 is active or valid and[br]before it was one out of two or something
0:53:28.229,0:53:31.259
like that.[br]Phillip: It's not that extreme. So it's
0:53:31.259,0:53:34.859
less than that. It's kind of...[br]Dennis: We have to say we did not check
0:53:34.859,0:53:41.239
this, but there were no special cases.[br]Phillip: But it changed over time? So
0:53:41.239,0:53:47.200
before it was less and, before it was more[br]and now it is less. And so what we checked
0:53:47.200,0:53:54.710
was whether an ID actually corresponds to[br]an account. And so this metric, yeah. And
0:53:54.710,0:53:57.299
it changed a little bit over time, but not[br]much.
0:53:57.299,0:54:02.239
Herald: So, so number three, please.[br]Question: Yeah. Thank you for a very
0:54:02.239,0:54:06.989
interesting talk. At the end, you gave[br]some recommendations, how to fix the
0:54:06.989,0:54:11.769
metrics, right? And it's always nice to[br]have some metrics because then, well, we
0:54:11.769,0:54:15.220
are the people who deal with the numbers.[br]So we want the metrics. But I want to
0:54:15.220,0:54:20.309
raise the issue whether quantitative[br]measure is actually the right thing to do.
0:54:20.309,0:54:26.449
So would you buy your furniture from store[br]A with 300 likes against store B with 200
0:54:26.449,0:54:32.049
likes? Or would it not be better to have a[br]more qualitative thing? And to what extent
0:54:32.049,0:54:38.259
is a quantitative measure maybe also the[br]source of a lot of bad developments we see
0:54:38.259,0:54:43.390
in social media to begin with, even not[br]with bot firms and anything, but just
0:54:43.390,0:54:48.339
people who go for the quick like and say[br]Hooray for Trump and then get, whatever,
0:54:48.339,0:54:52.479
all the Trumpists is liking that and the[br]others say Fuck Trump and you get all the
0:54:52.479,0:54:57.229
non Trumpists like that and you get all[br]the polarization, right? So, Instagram, I
0:54:57.229,0:55:02.650
think they just don't just display their[br]like equivalent anymore in order to
0:55:02.650,0:55:04.929
prevent that, so could you maybe comment[br]on that?
0:55:04.929,0:55:12.299
Svea: I think this is a good idea, to, to[br]hide the likes. Yes. But I you know, we
0:55:12.299,0:55:17.799
talked to many clickworkers and they do a[br]lot of stuff. And what they also do is
0:55:17.799,0:55:23.309
taking comments and doing copy paste for[br]comments section or for Amazon reviews.
0:55:23.309,0:55:29.789
So, you know, I think it's really hard to[br]get them out of the system because maybe
0:55:29.789,0:55:34.390
if the likes are not shown and if and when[br]the comments are counting, then you will
0:55:34.390,0:55:41.069
have people who are copy pasting comments[br]in the comments section. So I really think
0:55:41.069,0:55:44.519
that the networks, that they really have[br]an issue here.
0:55:44.519,0:55:49.829
Herald: So let's try to squeeze the last[br]three questions now. First, number seven,
0:55:49.829,0:55:52.950
really quick.[br]Question: Very quick. Thank you for the
0:55:52.950,0:55:58.799
nice insights. And I have a question about[br]the location of the users. So you made
0:55:58.799,0:56:03.289
your point that you can analyze by the[br]metadata where, uh, when the account was
0:56:03.289,0:56:08.650
made. But how about the location of the[br]followers? Is there any way to analyze
0:56:08.650,0:56:12.339
that as well?[br]Phillip: So we can only analyze that if
0:56:12.339,0:56:21.049
the users agreed to share it publicly and[br]not all of them do that, I think often a
0:56:21.049,0:56:26.460
name check is often a very good way to[br]check where somebody is from. For these
0:56:26.460,0:56:32.190
fake likes, for example. But as I said, it[br]always depends on what the user himself is
0:56:32.190,0:56:36.130
willing to share.[br]Herald: Internet?
0:56:36.130,0:56:41.039
Signal Angel: Isn't this just the western[br]version of the Chinese social credit
0:56:41.039,0:56:43.999
system? Where do we go from here? What is[br]the future of all this?
0:56:43.999,0:56:54.089
Svea: Yeah, it's dystopian, right? Oh,[br]yeah, I don't, after this research, you
0:56:54.089,0:57:01.109
know, for me, I deleted my Facebook[br]account like one or two years ago. So this
0:57:01.109,0:57:07.279
does you know, this did not matter to me[br]so much. But I stayed on Instagram and
0:57:07.279,0:57:13.359
when I saw all this bought likes and[br]abonnents and followers and also YouTube,
0:57:13.359,0:57:16.999
all this views, this, because the click[br]workers, they also watch YouTube videos.
0:57:16.999,0:57:20.859
They have to stay on them like 40 seconds,[br]it's really funny because they hate
0:57:20.859,0:57:27.239
hearing like techno music, rap music, all[br]40 seconds and then they go on. But when I
0:57:27.239,0:57:34.589
sit next to Herald for two hour, three[br]hours, I was so desillusionated about all
0:57:34.589,0:57:40.960
the social network things. And and I[br]thought, OK, don't count on anything. Just
0:57:40.960,0:57:46.119
if you like the content, follow them and[br]look at them. But don't believe anything.
0:57:46.119,0:57:50.479
That was my personal take away from this[br]research.
0:57:50.479,0:57:53.970
Herald: So very last question, microphone[br]two.
0:57:53.970,0:57:59.150
Question: A couple of days ago, The[br]Independent reported that Facebook, the
0:57:59.150,0:58:06.839
Facebook App was activating the camera[br]when reading a news feed. Could this be in
0:58:06.839,0:58:10.779
use in the context of detecting fake[br]accounts?
0:58:10.779,0:58:18.400
Svea: I don't know.[br]Phillip: So, I think that that in this
0:58:18.400,0:58:26.799
particular instance that it was probably a[br]bug. So, I don't know, but I mean that the
0:58:26.799,0:58:30.679
people who work at Facebook are, not all[br]of them are like crooks or anything that
0:58:30.679,0:58:35.130
they will deliberately program this kind[br]of stuff. So they said that it was kind of
0:58:35.130,0:58:41.189
a bug from from an update that they did.[br]And the question is whether we can
0:58:41.189,0:58:49.430
actually detect fake accounts with the[br]camera. And the problem is that current, I
0:58:49.430,0:58:57.469
don't think that current face recognition[br]technology is enough to detect that you
0:58:57.469,0:59:02.940
are a unique person. So there are so many[br]people on the planet that probably another
0:59:02.940,0:59:08.959
person who has the same face. And I think[br]the new iPhone, they also have this much
0:59:08.959,0:59:14.579
more sophisticated version of this[br]technology. And even they say, OK, there's
0:59:14.579,0:59:19.079
a chance of one in, I don't know, that[br]there is somebody who can unlock your
0:59:19.079,0:59:23.829
phone. So I think it's really hard to do[br]that with, do that with recording
0:59:23.829,0:59:29.299
technology, to actually prove that[br]somebody is just one person.
0:59:29.299,0:59:38.059
Herald: So with that, would you please[br]help me thank Svea, Dennis and Philip
0:59:38.059,0:59:41.160
one more time for this fantastic[br]presentation! Very interesting and very,
0:59:41.160,0:59:48.099
very disturbing. Thank you very much. [br]Applause
0:59:48.099,0:59:52.099
postroll music
0:59:52.099,1:00:16.000
Subtitles created by c3subtitles.de[br]in the year 2020. Join, and help us!