Would you like to inspect the original subtitles? These are the user uploaded subtitles that are being translated:
1
00:00:05,300 --> 00:00:09,220
It was the biggest political
earthquake of the century.
2
00:00:09,220 --> 00:00:13,380
We will make America great again!
3
00:00:16,260 --> 00:00:18,540
But just how did Donald Trump
defy the predictions
4
00:00:18,540 --> 00:00:22,580
of political pundits
and pollsters?
5
00:00:22,580 --> 00:00:25,780
The secret lies here,
in San Antonio, Texas.
6
00:00:27,820 --> 00:00:30,100
This was our Project Alamo.
7
00:00:30,100 --> 00:00:36,420
This is where the digital arm of the
Trump campaign operation was held.
8
00:00:38,740 --> 00:00:41,140
This is the extraordinary story
of two men
9
00:00:41,140 --> 00:00:43,820
with two very different views
of the world.
10
00:00:44,860 --> 00:00:47,300
We will build the wall...
11
00:00:47,300 --> 00:00:50,700
The path forward
is to connect more, not less.
12
00:00:50,700 --> 00:00:55,060
And how Facebook's Mark Zuckerberg
inadvertently helped Donald Trump
13
00:00:55,060 --> 00:00:58,340
become the most powerful man
on the planet.
14
00:00:58,340 --> 00:01:01,500
Without Facebook,
we wouldn't have won.
15
00:01:01,500 --> 00:01:05,740
I mean, Facebook really and truly
put us over the edge.
16
00:01:05,740 --> 00:01:09,460
With their secret algorithms
and online tracking,
17
00:01:09,460 --> 00:01:12,980
social media companies
know more about us than anyone.
18
00:01:14,380 --> 00:01:17,820
So you can predict mine and
everybody else's personality
19
00:01:17,820 --> 00:01:19,900
based on the things that
they've liked?
20
00:01:19,900 --> 00:01:21,540
That's correct.
21
00:01:21,540 --> 00:01:24,940
A very accurate prediction
of your intimate traits such as
22
00:01:24,940 --> 00:01:29,020
religiosity, political views,
intelligence, sexual orientation.
23
00:01:29,020 --> 00:01:32,420
Now, this power is transforming
politics.
24
00:01:32,420 --> 00:01:34,060
Can you understand though why maybe
25
00:01:34,060 --> 00:01:36,900
some people find it a little bit
creepy?
26
00:01:36,900 --> 00:01:39,660
No, I can't - quite the opposite.
That is the way the world is moving.
27
00:01:39,660 --> 00:01:43,420
Whether you like it or not,
it's an inevitable fact.
28
00:01:43,420 --> 00:01:47,060
Social media can bring politics
closer to the people,
29
00:01:47,060 --> 00:01:48,820
but its destructive power
30
00:01:48,820 --> 00:01:52,060
is creating a new
and unpredictable world.
31
00:01:52,060 --> 00:01:56,580
This is the story of how
Silicon Valley's mission to connect
32
00:01:56,580 --> 00:01:59,900
all of us is disrupting politics,
33
00:01:59,900 --> 00:02:03,060
plunging us into a world of
political turbulence
34
00:02:03,060 --> 00:02:05,260
that no-one can control.
35
00:02:20,900 --> 00:02:25,260
It might not look like it, but anger
is building in Silicon Valley.
36
00:02:25,260 --> 00:02:28,260
It's usually pretty quiet
around here.
37
00:02:28,260 --> 00:02:30,660
But not today.
38
00:02:30,660 --> 00:02:34,780
Every day you wake up and you wonder
what's going to be today's grief,
39
00:02:34,780 --> 00:02:37,860
brought by certain politicians
and leaders in the world.
40
00:02:37,860 --> 00:02:42,500
This is a very unusual
demonstration.
41
00:02:42,500 --> 00:02:45,180
I've been to loads
of demonstrations,
42
00:02:45,180 --> 00:02:46,940
but these aren't the people
43
00:02:46,940 --> 00:02:49,580
that usually go to demonstrations.
44
00:02:49,580 --> 00:02:53,700
In some ways, these are
the winners of society.
45
00:02:53,700 --> 00:02:56,620
This is the tech community
in Silicon Valley.
46
00:02:56,620 --> 00:03:00,660
Some of the wealthiest
people in the world.
47
00:03:00,660 --> 00:03:05,940
And they're here protesting
and demonstrating...
48
00:03:05,940 --> 00:03:08,660
against what they see
as the kind of changing world
49
00:03:08,660 --> 00:03:10,740
that they don't like.
50
00:03:10,740 --> 00:03:13,700
The problem, of course,
is the election of Donald Trump.
51
00:03:16,620 --> 00:03:18,540
Every day, I'm sure, you think,
52
00:03:18,540 --> 00:03:21,100
"I could be a part of resisting
those efforts
53
00:03:21,100 --> 00:03:23,260
"to mess things up for
the rest of us."
54
00:03:23,260 --> 00:03:27,660
Trump came to power promising
to control immigration...
55
00:03:27,660 --> 00:03:31,540
We will build the wall, 100%.
56
00:03:31,540 --> 00:03:34,740
..and to disengage from the world.
57
00:03:34,740 --> 00:03:38,660
From this day forward,
it's going to be
58
00:03:38,660 --> 00:03:41,020
only America first.
59
00:03:43,020 --> 00:03:45,260
America first.
60
00:03:47,700 --> 00:03:52,180
Now, Silicon Valley
is mobilising against him.
61
00:03:52,180 --> 00:03:55,180
We are seeing this explosion of
political activism, you know,
62
00:03:55,180 --> 00:03:57,740
all through the US and in Europe.
63
00:03:57,740 --> 00:04:00,260
Before he became an activist,
64
00:04:00,260 --> 00:04:03,940
Dex Torricke-Barton was speech
writer to the chairman of Google
65
00:04:03,940 --> 00:04:06,020
and the founder of Facebook.
66
00:04:08,260 --> 00:04:11,580
It's a moment when people who
believe in this global vision,
67
00:04:11,580 --> 00:04:13,700
as opposed to the nationalist
vision of the world,
68
00:04:13,700 --> 00:04:15,980
who believe in a world that
isn't about protectionism,
69
00:04:15,980 --> 00:04:18,620
whether it's data or
whether it's about trade,
70
00:04:18,620 --> 00:04:21,660
they're coming to stand up and
to mobilise in response to that.
71
00:04:21,660 --> 00:04:23,900
Because it feels like the whole of
Silicon Valley
72
00:04:23,900 --> 00:04:26,820
has been slightly taken by surprise
by what's happening. Absolutely.
73
00:04:26,820 --> 00:04:28,940
But these are the
smartest minds in the world.
74
00:04:28,940 --> 00:04:31,620
Yeah. With the most amazing
data models and polling.
75
00:04:31,620 --> 00:04:34,020
The smartest minds in the world
often can be very,
76
00:04:34,020 --> 00:04:36,860
very ignorant of the things
that are going on in the world.
77
00:04:36,860 --> 00:04:42,380
The tech god with the most ambitious
global vision is Dex's old boss -
78
00:04:42,380 --> 00:04:44,860
Facebook founder Mark Zuckerberg.
79
00:04:44,860 --> 00:04:49,460
One word captures the world
Zuck, as he's known here,
80
00:04:49,460 --> 00:04:52,180
is trying to build.
81
00:04:52,180 --> 00:04:54,420
Connectivity and access.
If we connected them...
82
00:04:54,420 --> 00:04:56,660
You give people connectivity...
That's the mission.
83
00:04:56,660 --> 00:04:59,140
Connecting with their friends...
You connect people over time...
84
00:04:59,140 --> 00:05:01,340
Connect everyone in the world...
I'm really optimistic about that.
85
00:05:01,340 --> 00:05:04,020
Make the world more open and
connected, that's what I care about.
86
00:05:04,020 --> 00:05:06,660
This is part of the critical
enabling infrastructure
for the world.
87
00:05:06,660 --> 00:05:08,820
Thank you, guys.
88
00:05:08,820 --> 00:05:11,140
What's Mark Zuckerberg worried
about most?
89
00:05:11,140 --> 00:05:13,980
Well, you know, Mark has dedicated
his life to connecting, you know,
90
00:05:13,980 --> 00:05:16,740
the world. You know, this is
something that he really, you know,
91
00:05:16,740 --> 00:05:18,180
cares passionately about.
92
00:05:18,180 --> 00:05:20,500
And, you know, as I said, you know,
93
00:05:20,500 --> 00:05:23,660
the same worldview and set
of policies that, you know,
94
00:05:23,660 --> 00:05:25,340
we'll build walls here,
95
00:05:25,340 --> 00:05:27,780
we'll build walls against
the sharing of information
96
00:05:27,780 --> 00:05:29,940
and building those
kind of, you know, networks.
97
00:05:29,940 --> 00:05:31,260
It's bigger than just the tech.
98
00:05:31,260 --> 00:05:34,420
It is. It's about the society that
Silicon Valley also wants to create?
99
00:05:34,420 --> 00:05:35,500
Absolutely.
100
00:05:35,500 --> 00:05:37,540
The tech gods believe the election
101
00:05:37,540 --> 00:05:40,580
of Donald Trump threatens
their vision
102
00:05:40,580 --> 00:05:42,260
of a globalised world.
103
00:05:43,700 --> 00:05:45,260
But in a cruel twist,
104
00:05:45,260 --> 00:05:48,500
is it possible their mission
to connect the world
105
00:05:48,500 --> 00:05:50,740
actually helped
bring him to power?
106
00:05:50,740 --> 00:05:54,380
The question I have is whether
the revolution brought about
107
00:05:54,380 --> 00:05:57,660
by social media companies
like Facebook
108
00:05:57,660 --> 00:06:01,300
has actually led to the political
changes in the world
109
00:06:01,300 --> 00:06:03,700
that these guys are
so worried about.
110
00:06:07,580 --> 00:06:09,420
To answer that question,
111
00:06:09,420 --> 00:06:13,540
you have to understand how
the tech titans of Silicon Valley
112
00:06:13,540 --> 00:06:16,260
rose to power.
113
00:06:16,260 --> 00:06:18,780
For that, you have to go back
20 years
114
00:06:18,780 --> 00:06:23,220
to a time when the online world
was still in its infancy.
115
00:06:23,220 --> 00:06:26,060
MUSIC: Rock N Roll Star by Oasis
116
00:06:28,100 --> 00:06:32,220
There were fears the new internet
was like the Wild West,
117
00:06:32,220 --> 00:06:35,260
anarchic and potentially harmful.
118
00:06:35,260 --> 00:06:40,980
Today our world is being remade, yet
again, by an information revolution.
119
00:06:40,980 --> 00:06:44,340
Changing the way we work,
the way we live,
120
00:06:44,340 --> 00:06:46,780
the way we relate to each other.
121
00:06:46,780 --> 00:06:50,180
The Telecommunications Act of 1996
122
00:06:50,180 --> 00:06:52,700
was designed to civilise
the internet,
123
00:06:52,700 --> 00:06:56,860
including protect children
from pornography.
124
00:06:56,860 --> 00:07:00,940
Today with the stroke of a pen, our
laws will catch up with our future.
125
00:07:00,940 --> 00:07:06,460
But buried deep within the act was a
secret whose impact no-one foresaw.
126
00:07:14,700 --> 00:07:16,100
Jeremy?
127
00:07:16,100 --> 00:07:17,620
Jamie. How are you doing?
128
00:07:17,620 --> 00:07:19,980
Nice to meet you.
129
00:07:19,980 --> 00:07:22,140
VOICEOVER: Jeremy Malcolm
is an analyst
130
00:07:22,140 --> 00:07:24,220
at the Electronic Frontier
Foundation,
131
00:07:24,220 --> 00:07:27,180
a civil liberties group
for the digital age.
132
00:07:29,060 --> 00:07:32,980
Much of Silicon Valley's accelerated
growth in the last two decades
133
00:07:32,980 --> 00:07:37,140
has been enabled by one clause
in the legislation.
134
00:07:38,260 --> 00:07:41,140
Hidden away in the middle
of that is this Section 230.
135
00:07:41,140 --> 00:07:43,180
What's the key line?
It literally just says,
136
00:07:43,180 --> 00:07:45,860
no provider or user of an
interactive computer service
137
00:07:45,860 --> 00:07:49,340
shall be treated as
the publisher or speaker
138
00:07:49,340 --> 00:07:51,300
of any information provided
139
00:07:51,300 --> 00:07:54,060
by another information content
provider. That's it.
140
00:07:54,060 --> 00:07:56,860
So what that basically means is,
if you're an internet platform,
141
00:07:56,860 --> 00:07:59,580
you don't get treated as
the publisher or speaker
142
00:07:59,580 --> 00:08:02,580
of something that your users say
using your platform.
143
00:08:02,580 --> 00:08:06,260
If the user says something online
that is, say, defamatory,
144
00:08:06,260 --> 00:08:08,380
the platform that
they communicate on
145
00:08:08,380 --> 00:08:10,540
isn't going to be held
responsible for it.
146
00:08:10,540 --> 00:08:13,740
And the user, of course, can be held
directly responsible.
147
00:08:13,740 --> 00:08:20,460
How important is this line
for social media companies today?
148
00:08:20,460 --> 00:08:23,100
I think if we didn't have this,
we probably wouldn't have
149
00:08:23,100 --> 00:08:25,940
the same kind of social media
companies that we have today.
150
00:08:25,940 --> 00:08:29,300
They wouldn't be willing to take on
the risk of having so much
151
00:08:29,300 --> 00:08:33,860
unfettered discussion. It's key
to the internet's freedom, really?
152
00:08:33,860 --> 00:08:37,100
We wouldn't have the internet
of today without this.
153
00:08:37,100 --> 00:08:40,140
And so, if we are going to make
any changes to it,
154
00:08:40,140 --> 00:08:41,940
we have to be really,
really careful.
155
00:08:43,860 --> 00:08:47,220
These 26 words changed the world.
156
00:08:51,700 --> 00:08:54,580
They allowed a new kind
of business to spring up -
157
00:08:54,580 --> 00:08:58,820
online platforms that became
the internet giants of today.
158
00:08:58,820 --> 00:09:03,500
Facebook, Google, YouTube - they
encouraged users to upload content,
159
00:09:03,500 --> 00:09:07,340
often things about their lives
or moments that mattered to them,
160
00:09:07,340 --> 00:09:10,180
onto their sites for free.
161
00:09:10,180 --> 00:09:14,020
And in exchange,
they got to hoard all of that data
162
00:09:14,020 --> 00:09:16,300
but without any real responsibility
163
00:09:16,300 --> 00:09:19,940
for the effects of the content
that people were posting.
164
00:09:22,980 --> 00:09:26,100
Hundreds of millions of us
flocked to these new sites,
165
00:09:26,100 --> 00:09:29,980
putting more of our lives online.
166
00:09:29,980 --> 00:09:33,340
At first, the tech firms
couldn't figure out
167
00:09:33,340 --> 00:09:36,420
how to turn that data
into big money.
168
00:09:36,420 --> 00:09:41,340
But that changed when a secret
within that data was unlocked.
169
00:09:41,340 --> 00:09:43,420
Antonio, Jamie.
170
00:09:43,420 --> 00:09:47,180
'A secret Antonio Garcia Martinez
helped reveal at Facebook.'
171
00:09:50,220 --> 00:09:52,220
Tell me a bit about your time
at Facebook.
172
00:09:52,220 --> 00:09:53,860
Well, that was interesting.
173
00:09:53,860 --> 00:09:56,940
I was what's called
a product manager for ads targeting.
174
00:09:56,940 --> 00:09:59,580
That means basically taking your
data and using it to basically
175
00:09:59,580 --> 00:10:02,820
make money on Facebook,
to monetise Facebook's data.
176
00:10:02,820 --> 00:10:05,500
If you go browse the internet or
buy stuff in stores or whatever,
177
00:10:05,500 --> 00:10:07,340
and then you see ads related
to all that stuff
178
00:10:07,340 --> 00:10:08,780
inside Facebook - I created that.
179
00:10:10,900 --> 00:10:12,980
Facebook offers advertisers ways
180
00:10:12,980 --> 00:10:16,300
to target individual users
of the site with adverts.
181
00:10:16,300 --> 00:10:21,900
It can be driven by data about
how we use the platform.
182
00:10:21,900 --> 00:10:24,860
Here's some examples of what's data
for Facebook that makes money.
183
00:10:24,860 --> 00:10:27,620
What you've liked on Facebook,
links that you shared,
184
00:10:27,620 --> 00:10:29,740
who you happen to know on Facebook,
for example.
185
00:10:29,740 --> 00:10:32,900
Where you've used Facebook,
what devices, your iPad,
your work computer,
186
00:10:32,900 --> 00:10:34,940
your home computer.
In the case of Amazon,
187
00:10:34,940 --> 00:10:36,660
it's obviously what you've
purchased.
188
00:10:36,660 --> 00:10:38,780
In the case of Google,
it's what you searched for.
189
00:10:38,780 --> 00:10:40,940
How do they turn me...
I like something on Facebook,
190
00:10:40,940 --> 00:10:44,260
and I share a link on Facebook,
how could they turn that
191
00:10:44,260 --> 00:10:47,300
into something that another
company would care about?
192
00:10:47,300 --> 00:10:49,260
There is what's called
a targeting system,
193
00:10:49,260 --> 00:10:51,580
and so the advertiser can actually
go in and specify,
194
00:10:51,580 --> 00:10:55,580
I want people who are within
this city and who have liked BMW
or Burberry, for example.
195
00:10:55,580 --> 00:10:57,580
So an advertiser pays
Facebook and says,
196
00:10:57,580 --> 00:11:00,660
I want these sorts of people?
That's effectively it, that's right.
197
00:11:02,140 --> 00:11:06,220
The innovation that opened up bigger
profits was to allow Facebook users
198
00:11:06,220 --> 00:11:12,900
to be targeted using data about what
they do on the rest of the internet.
199
00:11:12,900 --> 00:11:14,940
The real key thing
that marketers want
200
00:11:14,940 --> 00:11:19,620
is the unique, immutable, flawless,
high fidelity ID
201
00:11:19,620 --> 00:11:22,060
for one person on the internet,
and Facebook provides that.
202
00:11:22,060 --> 00:11:23,860
It is your identity online.
203
00:11:23,860 --> 00:11:27,580
Facebook can tell an advertiser,
this is the real Jamie Barlow,
204
00:11:27,580 --> 00:11:29,580
this is what he's like? Yeah.
205
00:11:29,580 --> 00:11:33,220
A company like Walmart can literally
take your data, your e-mail,
206
00:11:33,220 --> 00:11:36,380
phone number, whatever you use
for their frequent shopper
programme, etc,
207
00:11:36,380 --> 00:11:39,780
and join that to Facebook
and literally target those people
based on that data.
208
00:11:39,780 --> 00:11:40,900
That's part of what I built.
209
00:11:46,180 --> 00:11:50,420
The tech gods suck in all this data
about how we use their technologies
210
00:11:50,420 --> 00:11:53,980
to build their vast fortunes.
211
00:11:55,820 --> 00:11:59,620
I mean, it sounds like data is like
oil, it's keeping the economy going?
212
00:11:59,620 --> 00:12:01,660
Right. I mean, the difference
is these companies,
213
00:12:01,660 --> 00:12:04,820
instead of drilling for this oil,
they generate this oil via,
214
00:12:04,820 --> 00:12:07,780
by getting users to actually use
their apps and then they actually
215
00:12:07,780 --> 00:12:10,180
monetise it. Usually via advertising
or other mechanisms.
216
00:12:10,180 --> 00:12:13,020
But, yeah, it is the new oil.
217
00:12:13,020 --> 00:12:16,660
Data about billions of us
is propelling Silicon Valley
218
00:12:16,660 --> 00:12:19,700
to the pinnacle of
the global economy.
219
00:12:22,380 --> 00:12:24,220
The world's largest hotel company,
Airbnb,
220
00:12:24,220 --> 00:12:26,020
doesn't own a single piece
of real estate.
221
00:12:26,020 --> 00:12:28,780
The world's largest taxi company,
Uber, doesn't own any cars.
222
00:12:28,780 --> 00:12:31,980
The world's largest media company,
Facebook, doesn't produce
any media, right?
223
00:12:31,980 --> 00:12:33,540
So what do they have?
224
00:12:33,540 --> 00:12:36,620
Well, they have the data around how
you use those resources and how you
225
00:12:36,620 --> 00:12:38,900
use those assets. And that's really
what they are.
226
00:12:42,660 --> 00:12:47,540
The secret of targeting us with
adverts is keeping us online
227
00:12:47,540 --> 00:12:50,500
for as long as possible.
228
00:12:50,500 --> 00:12:54,060
I thought I'd just see
how much time I spend on here.
229
00:12:54,060 --> 00:12:59,540
So I've got an app that counts
how often I pick this thing up.
230
00:12:59,540 --> 00:13:03,140
Our time is the Holy Grail
of Silicon Valley.
231
00:13:04,620 --> 00:13:07,220
Here's what my life looks like
on a typical day.
232
00:13:09,900 --> 00:13:13,340
Yeah, could I have a flat white,
please? Flat white? Yeah.
233
00:13:13,340 --> 00:13:18,700
Like more and more of us, my phone
is my gateway to the online world.
234
00:13:18,700 --> 00:13:21,500
It's how I check
my social media accounts.
235
00:13:21,500 --> 00:13:26,860
On average, Facebook users spend
50 minutes every day on the site.
236
00:13:32,300 --> 00:13:34,100
The longer we spend connected,
237
00:13:34,100 --> 00:13:38,340
the more Silicon Valley
can learn about us,
238
00:13:38,340 --> 00:13:42,340
and the more targeted and effective
their advertising can be.
239
00:14:06,540 --> 00:14:09,060
So apparently, today, I...
240
00:14:09,060 --> 00:14:13,660
I've checked my phone 117 times...
241
00:14:15,300 --> 00:14:20,820
..and I've been on this phone
for nearly five and a half hours.
242
00:14:20,820 --> 00:14:23,620
Well, I mean, that's a lot,
that's a lot of hours.
243
00:14:23,620 --> 00:14:26,660
I mean, it's kind of
nearly half the day,
244
00:14:26,660 --> 00:14:29,220
spent on this phone.
245
00:14:29,220 --> 00:14:32,540
But it's weird, because it doesn't
feel like I spend that long on it.
246
00:14:32,540 --> 00:14:34,380
The strange thing about it is
247
00:14:34,380 --> 00:14:36,820
that I don't even really know
what I'm doing
248
00:14:36,820 --> 00:14:39,460
for these five hours
that I'm spending on this phone.
249
00:14:41,100 --> 00:14:45,700
What is it that is keeping us hooked
to Silicon Valley's global network?
250
00:14:50,220 --> 00:14:55,060
I'm in Seattle to meet someone who
saw how the tech gods embraced
251
00:14:55,060 --> 00:14:59,660
new psychological insights
into how we all make decisions.
252
00:15:03,420 --> 00:15:05,500
I was a post-doc
with Stephen Hawking.
253
00:15:05,500 --> 00:15:08,340
Once Chief Technology Officer
at Microsoft,
254
00:15:08,340 --> 00:15:11,140
Nathan Myhrvold is the most
passionate technologist
255
00:15:11,140 --> 00:15:14,020
I have ever met.
256
00:15:14,020 --> 00:15:16,380
Some complicated-looking equations
in the background.
257
00:15:16,380 --> 00:15:18,700
Well, it turns out if you work
with Stephen Hawking,
258
00:15:18,700 --> 00:15:20,940
you do work with
complicated equations.
259
00:15:20,940 --> 00:15:23,340
It's kind of the nature
of the beast!
260
00:15:23,340 --> 00:15:24,580
Amazing picture.
261
00:15:27,820 --> 00:15:31,260
A decade ago, Nathan brought
together Daniel Kahneman,
262
00:15:31,260 --> 00:15:34,540
pioneer of the new science
of behavioural economics,
263
00:15:34,540 --> 00:15:38,340
and Silicon Valley's leaders,
for a series of meetings.
264
00:15:38,340 --> 00:15:41,020
I came and Jeff Bezos came.
265
00:15:41,020 --> 00:15:46,980
That's Jeff Bezos, the founder
of Amazon, worth $76 billion.
266
00:15:46,980 --> 00:15:49,140
Sean Parker. Sean was there.
267
00:15:49,140 --> 00:15:53,340
That's Sean Parker,
the first president of Facebook.
268
00:15:53,340 --> 00:15:55,420
And the Google founders were there.
269
00:15:57,500 --> 00:16:04,180
The proposition was, come to this
very nice resort in Napa,
270
00:16:04,180 --> 00:16:08,220
and for several days,
just have Kahneman
271
00:16:08,220 --> 00:16:10,060
and then also a couple of
272
00:16:10,060 --> 00:16:13,540
other behavioural economists
explain things.
273
00:16:13,540 --> 00:16:16,180
And ask questions
and see what happens.
274
00:16:16,180 --> 00:16:22,460
Kahneman had a simple but brilliant
theory on how we make decisions.
275
00:16:22,460 --> 00:16:27,740
He had found we use one of two
different systems of thinking.
276
00:16:28,940 --> 00:16:32,260
In this dichotomy, you have...
277
00:16:32,260 --> 00:16:34,820
over here is a hunch...
278
00:16:37,060 --> 00:16:38,500
..a guess,
279
00:16:38,500 --> 00:16:40,940
a gut feeling...
280
00:16:44,660 --> 00:16:47,860
..and "I just know".
281
00:16:47,860 --> 00:16:54,140
So this is sort of emotional
and more like, just instant stuff?
282
00:16:54,140 --> 00:16:56,900
That's the idea.
This set of things
283
00:16:56,900 --> 00:17:00,900
is not particularly good
at a different set of stuff
284
00:17:00,900 --> 00:17:02,900
that involves...
285
00:17:05,300 --> 00:17:06,500
..analysis...
286
00:17:09,380 --> 00:17:11,140
..numbers...
287
00:17:13,180 --> 00:17:14,220
..probability.
288
00:17:15,700 --> 00:17:18,940
The meetings in Napa didn't deal
with the basics
289
00:17:18,940 --> 00:17:21,900
of behavioural economics, but how
might the insights of the new
290
00:17:21,900 --> 00:17:24,140
science have helped the tech gods?
291
00:17:24,140 --> 00:17:27,660
A lot of advertising is about trying
to hook people
292
00:17:27,660 --> 00:17:31,340
in these type-one things to get
interested one way or the other.
293
00:17:31,340 --> 00:17:36,180
Technology companies undoubtedly
use that to one degree or another.
294
00:17:36,180 --> 00:17:37,860
You know, the term "clickbait",
295
00:17:37,860 --> 00:17:40,460
for things that look exciting
to click on.
296
00:17:40,460 --> 00:17:43,180
There's billions of dollars
change hands
297
00:17:43,180 --> 00:17:47,180
because we all get enticed
into clicking something.
298
00:17:47,180 --> 00:17:51,020
And there's a lot of things that
I click on and then you get there,
299
00:17:51,020 --> 00:17:54,100
you're like, OK, fine, you were just
messing with me.
300
00:17:54,100 --> 00:17:56,660
You're playing to
the type-one things,
301
00:17:56,660 --> 00:17:59,140
you're putting a set
of triggers out there
302
00:17:59,140 --> 00:18:02,220
that make me want to click on it,
303
00:18:02,220 --> 00:18:07,100
and even though, like, I'm aware
of that, I still sometimes click!
304
00:18:07,100 --> 00:18:09,820
Tech companies both
try to understand
305
00:18:09,820 --> 00:18:13,420
our behaviour by having smart humans
think about it and increasingly
306
00:18:13,420 --> 00:18:15,740
by having machines think about it.
307
00:18:15,740 --> 00:18:17,220
By having machines track us
308
00:18:17,220 --> 00:18:20,060
to see, what is the clickbait
Nathan falls for?
309
00:18:20,060 --> 00:18:22,500
What are the things he really likes
to spend time on?
310
00:18:22,500 --> 00:18:24,700
Let's show him more of that stuff!
311
00:18:26,580 --> 00:18:30,220
Trying to grab the attention
of the consumer is nothing new.
312
00:18:30,220 --> 00:18:32,900
That's what advertising
is all about.
313
00:18:32,900 --> 00:18:36,100
But insights into how we make
decisions helped Silicon Valley
314
00:18:36,100 --> 00:18:38,540
to shape the online world.
315
00:18:38,540 --> 00:18:43,820
And little wonder, their success
depends on keeping us engaged.
316
00:18:43,820 --> 00:18:48,220
From 1-Click buying on Amazon
to the Facebook like,
317
00:18:48,220 --> 00:18:52,300
the more they've hooked us,
the more the money has rolled in.
318
00:18:54,180 --> 00:18:56,260
As Silicon Valley became
more influential,
319
00:18:56,260 --> 00:19:01,300
it started attracting
powerful friends...in politics.
320
00:19:01,300 --> 00:19:07,820
In 2008, Barack Obama had pioneered
political campaigning on Facebook.
321
00:19:07,820 --> 00:19:13,220
As President, he was drawn
to Facebook's founder, Zuck.
322
00:19:13,220 --> 00:19:15,020
Sorry, I'm kind of nervous.
323
00:19:15,020 --> 00:19:18,620
We have the President
of the United States here!
324
00:19:18,620 --> 00:19:22,220
My name is Barack Obama
and I'm the guy who got Mark
325
00:19:22,220 --> 00:19:24,820
to wear a jacket and tie!
326
00:19:27,780 --> 00:19:29,140
How you doing?
327
00:19:29,140 --> 00:19:32,620
Great. I'll have
huevos rancheros, please.
328
00:19:32,620 --> 00:19:37,260
And if I could have an egg and
cheese sandwich on English muffin?
329
00:19:37,260 --> 00:19:41,460
Aneesh Chopra was Obama's
first Chief Technology Officer,
330
00:19:41,460 --> 00:19:43,460
and saw how close the relationship
331
00:19:43,460 --> 00:19:47,820
between the White House
and Silicon Valley became.
332
00:19:47,820 --> 00:19:51,540
The President's philosophy
and his approach to governing
333
00:19:51,540 --> 00:19:54,420
garnered a great deal
of personal interest
334
00:19:54,420 --> 00:19:57,140
among many executives
in Silicon Valley.
335
00:19:57,140 --> 00:20:00,620
They were donors to his campaign,
volunteers,
336
00:20:00,620 --> 00:20:02,660
active recruiters of
engineering talent
337
00:20:02,660 --> 00:20:04,260
to support the campaign
apparatus.
338
00:20:04,260 --> 00:20:06,500
He had struck a chord.
339
00:20:06,500 --> 00:20:07,860
Why? Because frankly,
340
00:20:07,860 --> 00:20:11,340
it was an inspiring voice
that really tapped into
341
00:20:11,340 --> 00:20:14,060
the hopefulness of the country.
342
00:20:14,060 --> 00:20:17,700
And a lot of Silicon Valley
shares this sense of hopefulness,
343
00:20:17,700 --> 00:20:20,100
this optimistic view that we can
solve problems
344
00:20:20,100 --> 00:20:21,380
if we would work together
345
00:20:21,380 --> 00:20:24,780
and take advantage of these new
capabilities that are coming online.
346
00:20:24,780 --> 00:20:28,660
So people in Silicon Valley
saw President Obama
347
00:20:28,660 --> 00:20:33,180
as a bit of a kindred spirit?
Oh, yeah. Oh, yeah.
348
00:20:33,180 --> 00:20:35,620
If you'd like, Mark,
we can take our jackets off.
349
00:20:35,620 --> 00:20:37,380
That's good!
350
00:20:37,380 --> 00:20:39,820
Facebook's mission
to connect the world
351
00:20:39,820 --> 00:20:44,900
went hand-in-hand with Obama's
policies promoting globalisation
352
00:20:44,900 --> 00:20:46,740
and free markets.
353
00:20:46,740 --> 00:20:49,140
And Facebook was seen
to be improving
354
00:20:49,140 --> 00:20:51,420
the political process itself.
355
00:20:51,420 --> 00:20:55,700
Part of what makes for
a healthy democracy
356
00:20:55,700 --> 00:21:00,020
is when you've got citizens
who are informed, who are engaged,
357
00:21:00,020 --> 00:21:03,300
and what Facebook allows us to do
358
00:21:03,300 --> 00:21:06,620
is make sure this isn't just
a one-way conversation.
359
00:21:06,620 --> 00:21:10,100
You have the tech mind-set and
governments increasingly share
360
00:21:10,100 --> 00:21:12,940
the same view of the world.
That's not a natural...
361
00:21:12,940 --> 00:21:14,340
It's a spirit of liberty.
362
00:21:14,340 --> 00:21:16,580
It's a spirit of freedom.
363
00:21:16,580 --> 00:21:20,180
That is manifest today
in these new technologies.
364
00:21:20,180 --> 00:21:23,740
It happens to be that freedom means
I can tweet something offensive.
365
00:21:23,740 --> 00:21:26,940
But it also means
that I have a voice.
366
00:21:26,940 --> 00:21:30,620
Please raise your right hand
and repeat after me...
367
00:21:30,620 --> 00:21:32,740
By the time Obama won
his second term,
368
00:21:32,740 --> 00:21:37,660
he was feted for his mastery of
social media's persuasive power.
369
00:21:37,660 --> 00:21:40,380
But across the political spectrum,
370
00:21:40,380 --> 00:21:44,620
the race was on to find new ways
to gain a digital edge.
371
00:21:44,620 --> 00:21:47,700
The world was about to change
for Facebook.
372
00:21:57,420 --> 00:22:01,900
Stanford University,
in the heart of Silicon Valley.
373
00:22:01,900 --> 00:22:06,500
Home to a psychologist
investigating just how revealing
374
00:22:06,500 --> 00:22:11,580
Facebook's hoard of information
about each of us could really be.
375
00:22:11,580 --> 00:22:14,500
How are you doing?
Nice to meet you.
376
00:22:14,500 --> 00:22:19,380
Dr Michal Kosinski specialises
in psychometrics -
377
00:22:19,380 --> 00:22:22,540
the science of predicting
psychological traits
378
00:22:22,540 --> 00:22:24,660
like personality.
379
00:22:24,660 --> 00:22:27,500
So in the past, when you wanted to
measure someone's personality
380
00:22:27,500 --> 00:22:31,340
or intelligence, you needed
to give them a question or a test,
381
00:22:31,340 --> 00:22:34,420
and they would have to answer
a bunch of questions.
382
00:22:34,420 --> 00:22:38,020
Now, many of those questions
would basically ask you
383
00:22:38,020 --> 00:22:40,060
about whether you like poetry,
384
00:22:40,060 --> 00:22:42,100
or you like hanging out
with other people,
385
00:22:42,100 --> 00:22:44,140
or you like the theatre,
and so on.
386
00:22:44,140 --> 00:22:47,780
But these days, you don't need
to ask these questions any more.
387
00:22:47,780 --> 00:22:51,700
Why? Because while going through
our lives,
388
00:22:51,700 --> 00:22:55,180
we are leaving behind
a lot of digital footprints
389
00:22:55,180 --> 00:22:57,940
that basically contain
the same information.
390
00:22:57,940 --> 00:23:02,220
So instead of asking you
whether you like poetry,
391
00:23:02,220 --> 00:23:06,540
I can just look at
your reading history on Amazon
392
00:23:06,540 --> 00:23:08,740
or your Facebook likes,
393
00:23:08,740 --> 00:23:11,340
and I would just get exactly
the same information.
394
00:23:13,100 --> 00:23:17,420
In 2011, Dr Kosinski and his team
at the University of Cambridge
395
00:23:17,420 --> 00:23:19,500
developed an online survey
396
00:23:19,500 --> 00:23:22,540
to measure volunteers'
personality traits.
397
00:23:22,540 --> 00:23:25,340
With their permission,
he matched their results
398
00:23:25,340 --> 00:23:27,340
with their Facebook data.
399
00:23:27,340 --> 00:23:30,740
More than 6 million people
took part.
400
00:23:30,740 --> 00:23:33,700
We have people's Facebook likes,
401
00:23:33,700 --> 00:23:36,980
people's status updates
and profile data,
402
00:23:36,980 --> 00:23:41,220
and this allows us to build those...
to gain better understanding
403
00:23:41,220 --> 00:23:44,020
of how psychological traits
are being expressed
404
00:23:44,020 --> 00:23:46,100
in the digital environment.
405
00:23:46,100 --> 00:23:50,780
How you can measure psychological
traits using digital footprints.
406
00:23:50,780 --> 00:23:53,420
An algorithm that can look
at millions of people
407
00:23:53,420 --> 00:23:55,820
and it can look at
hundreds of thousands
408
00:23:55,820 --> 00:23:57,700
or tens of thousands of your likes
409
00:23:57,700 --> 00:24:02,140
can extract and utilise even those
little pieces of information
410
00:24:02,140 --> 00:24:05,660
and combine it
into a very accurate profile.
411
00:24:05,660 --> 00:24:10,180
You can quite accurately predict
mine, and in fact, everybody else's
412
00:24:10,180 --> 00:24:14,860
personality, based on the things
that they've liked?
413
00:24:14,860 --> 00:24:16,340
That's correct.
414
00:24:16,340 --> 00:24:19,380
It can also be used to turn
your digital footprint
415
00:24:19,380 --> 00:24:21,860
into a very accurate prediction
416
00:24:21,860 --> 00:24:24,300
of your intimate traits,
such as religiosity,
417
00:24:24,300 --> 00:24:26,700
political views, personality,
intelligence,
418
00:24:26,700 --> 00:24:30,980
sexual orientation and a bunch of
other psychological traits.
419
00:24:30,980 --> 00:24:34,460
If I'm logged in, we can maybe see
how accurate this actually is.
420
00:24:34,460 --> 00:24:38,300
So I hit "make prediction"
and it's going to try and predict
421
00:24:38,300 --> 00:24:39,900
my personality from
my Facebook page.
422
00:24:39,900 --> 00:24:41,940
From your Facebook likes.
423
00:24:41,940 --> 00:24:43,700
According to your likes,
424
00:24:43,700 --> 00:24:47,020
you're open-minded,
liberal and artistic.
425
00:24:47,020 --> 00:24:49,900
Judges your intelligence
to be extremely high.
426
00:24:49,900 --> 00:24:54,140
Well done.
Yes. I'm extremely intelligent!
427
00:24:54,140 --> 00:24:57,180
You're not religious,
but if you are religious,
428
00:24:57,180 --> 00:25:00,420
most likely you'd be a Catholic.
I was raised a Catholic!
429
00:25:00,420 --> 00:25:02,860
I can't believe it knows that.
430
00:25:02,860 --> 00:25:06,460
Because I... I don't say anything
about being a Catholic anywhere,
431
00:25:06,460 --> 00:25:10,220
but I was raised as a Catholic,
but I'm not a practising Catholic.
432
00:25:10,220 --> 00:25:11,980
So it's like...
433
00:25:11,980 --> 00:25:14,060
It's absolutely spot on.
434
00:25:14,060 --> 00:25:18,060
Oh, my God! Journalism,
but also what did I study?
435
00:25:18,060 --> 00:25:21,820
Studied history, and I didn't put
anything about history in there.
436
00:25:21,820 --> 00:25:23,420
I think this is one of the things
437
00:25:23,420 --> 00:25:27,460
that people don't really get about
those predictions, that they think,
438
00:25:27,460 --> 00:25:30,140
look, if I like Lady Gaga
on Facebook,
439
00:25:30,140 --> 00:25:32,740
obviously people will know that
I like Lady Gaga,
440
00:25:32,740 --> 00:25:35,380
or the Government will know
that I like Lady Gaga.
441
00:25:35,380 --> 00:25:37,340
Look, you don't need
a rocket scientist
442
00:25:37,340 --> 00:25:40,460
to look at your Spotify playlist
or your Facebook likes
443
00:25:40,460 --> 00:25:42,900
to figure out that you like
Lady Gaga.
444
00:25:42,900 --> 00:25:48,580
What's really world-changing about
those algorithms is that they can
445
00:25:48,580 --> 00:25:52,660
take your music preferences
or your book preferences
446
00:25:52,660 --> 00:25:56,300
and extract from this
seemingly innocent information
447
00:25:56,300 --> 00:26:00,180
very accurate predictions about your
religiosity, leadership potential,
448
00:26:00,180 --> 00:26:03,500
political views,
personality and so on.
449
00:26:03,500 --> 00:26:06,460
Can you predict people's
political persuasions with this?
450
00:26:06,460 --> 00:26:09,100
In fact, the first dimension here,
openness to experience,
451
00:26:09,100 --> 00:26:11,940
is a very good predictor
of political views.
452
00:26:11,940 --> 00:26:15,620
People scoring high on openness,
they tend to be liberal,
453
00:26:15,620 --> 00:26:19,860
people who score low on openness,
we even call it conservative and
454
00:26:19,860 --> 00:26:23,300
traditional, they tend to vote
for conservative candidates.
455
00:26:23,300 --> 00:26:26,260
What about the potential
to manipulate people?
456
00:26:26,260 --> 00:26:30,020
So, obviously, if you now can use an
algorithm to get to know millions
457
00:26:30,020 --> 00:26:33,900
of people very intimately
and then use another algorithm
458
00:26:33,900 --> 00:26:38,500
to adjust the message
that you are sending to them,
459
00:26:38,500 --> 00:26:40,980
to make it most persuasive,
460
00:26:40,980 --> 00:26:43,820
obviously gives you a lot
of power.
461
00:26:52,780 --> 00:26:56,260
I'm quite surprised at just how
accurate this model could be.
462
00:26:56,260 --> 00:27:01,540
I mean, I cannot believe
that on the basis of a few things
463
00:27:01,540 --> 00:27:06,180
I just, you know, carelessly clicked
that I liked,
464
00:27:06,180 --> 00:27:10,900
the model was able to work out
that I could have been Catholic,
465
00:27:10,900 --> 00:27:13,700
or had a Catholic upbringing.
466
00:27:13,700 --> 00:27:19,140
And clearly, this is a very powerful
way of understanding people.
467
00:27:20,140 --> 00:27:23,460
Very exciting possibilities,
but I can't help
468
00:27:23,460 --> 00:27:29,500
fearing that there is that
potential, whoever has that power,
469
00:27:29,500 --> 00:27:32,140
whoever can control that model...
470
00:27:33,860 --> 00:27:38,660
..will have sort of unprecedented
possibilities of manipulating
471
00:27:38,660 --> 00:27:42,540
what people think, how they behave,
what they see,
472
00:27:42,540 --> 00:27:46,580
whether that's selling things
to people or how people vote,
473
00:27:46,580 --> 00:27:49,220
and that's pretty scary too.
474
00:27:52,700 --> 00:27:56,100
Our era is defined
by political shocks.
475
00:27:59,940 --> 00:28:03,180
None bigger than the rise
of Donald Trump,
476
00:28:03,180 --> 00:28:05,660
who defied pollsters
and the mainstream media
477
00:28:05,660 --> 00:28:08,100
to win the American presidency.
478
00:28:09,140 --> 00:28:12,980
Now questions swirl around
his use of the American affiliate
479
00:28:12,980 --> 00:28:15,220
of a British insights company,
480
00:28:15,220 --> 00:28:18,540
Cambridge Analytica,
who use psychographics.
481
00:28:26,260 --> 00:28:28,260
I'm in Texas to uncover how far
482
00:28:28,260 --> 00:28:32,540
Cambridge Analytica's expertise
in personality prediction
483
00:28:32,540 --> 00:28:36,940
played a part in Trump's
political triumph,
484
00:28:36,940 --> 00:28:39,420
and how his revolutionary campaign
485
00:28:39,420 --> 00:28:42,820
exploited Silicon Valley's
social networks.
486
00:28:44,660 --> 00:28:46,700
Everyone seems to agree
487
00:28:46,700 --> 00:28:50,140
that Trump ran an exceptional
election campaign
488
00:28:50,140 --> 00:28:53,460
using digital technologies.
489
00:28:53,460 --> 00:28:57,500
But no-one really knows
what they did,
490
00:28:57,500 --> 00:29:00,500
who they were working with,
who was helping them,
491
00:29:00,500 --> 00:29:03,980
what the techniques they used were.
492
00:29:03,980 --> 00:29:08,500
So I've come here to try to unravel
the mystery.
493
00:29:09,980 --> 00:29:13,740
The operation inside this unassuming
building in San Antonio
494
00:29:13,740 --> 00:29:18,820
was largely hidden from view,
but crucial to Trump's success.
495
00:29:18,820 --> 00:29:23,260
Since then,
I'm the only person to get in here
496
00:29:23,260 --> 00:29:25,660
to find out what really happened.
497
00:29:25,660 --> 00:29:27,780
This was our Project Alamo,
498
00:29:27,780 --> 00:29:31,140
so this was where the digital arm
499
00:29:31,140 --> 00:29:34,260
of the Trump campaign operation
was held.
500
00:29:34,260 --> 00:29:36,900
Theresa Hong is speaking publicly
501
00:29:36,900 --> 00:29:41,100
for the first time about her role
as digital content director
502
00:29:41,100 --> 00:29:42,940
for the Trump campaign.
503
00:29:44,140 --> 00:29:47,860
So, why is it called Project Alamo?
504
00:29:47,860 --> 00:29:52,140
It was called Project Alamo
based on the data, actually,
505
00:29:52,140 --> 00:29:54,660
that was Cambridge Analytica -
506
00:29:54,660 --> 00:29:57,260
they came up with the Alamo
data set, right?
507
00:29:57,260 --> 00:30:00,860
So, we just kind of adopted
the name Project Alamo.
508
00:30:00,860 --> 00:30:05,500
It does conjure up sort of images
of a battle of some sort.
509
00:30:05,500 --> 00:30:07,420
Yeah. It kind of was!
510
00:30:07,420 --> 00:30:10,380
In a sense, you know. Yeah, yeah.
511
00:30:14,260 --> 00:30:17,020
Project Alamo was so important,
512
00:30:17,020 --> 00:30:20,700
Donald Trump visited the hundred
or so workers based here
513
00:30:20,700 --> 00:30:23,220
during the campaign.
514
00:30:23,220 --> 00:30:26,660
Ever since he started tweeting
in 2009,
515
00:30:26,660 --> 00:30:30,300
Trump had grasped the power
of social media.
516
00:30:30,300 --> 00:30:33,380
Now, in the fight of his life,
517
00:30:33,380 --> 00:30:36,820
the campaign manipulated
his Facebook presence.
518
00:30:36,820 --> 00:30:41,460
Trump's Twitter account, that's
all his, he's the one that ran that,
519
00:30:41,460 --> 00:30:43,940
and I did a lot of his Facebook,
520
00:30:43,940 --> 00:30:49,540
so I wrote a lot for him, you know,
I kind of channelled Mr Trump.
521
00:30:49,540 --> 00:30:53,020
How do you possibly write a post
on Facebook like Donald Trump?
522
00:30:53,020 --> 00:30:54,700
"Believe me."
523
00:30:54,700 --> 00:30:58,740
A lot of believe me's,
a lot of alsos, a lot of...verys.
524
00:30:58,740 --> 00:31:02,340
Actually, he was really wonderful
to write for, just because
525
00:31:02,340 --> 00:31:06,820
it was so refreshing,
it was so authentic.
526
00:31:10,380 --> 00:31:12,340
We headed to the heart
of the operation.
527
00:31:15,180 --> 00:31:17,020
Cambridge Analytica was here.
528
00:31:17,020 --> 00:31:20,100
It was just a line of computers,
right?
529
00:31:20,100 --> 00:31:23,180
This is where their operation was
and this was kind of
530
00:31:23,180 --> 00:31:26,980
the brain of the data,
this was the data centre.
531
00:31:26,980 --> 00:31:30,660
This was the data centre, this was
the centre of the data centre.
532
00:31:30,660 --> 00:31:33,660
Exactly right, yes, it was.
Yes, it was.
533
00:31:33,660 --> 00:31:35,740
Cambridge Analytica were using data
534
00:31:35,740 --> 00:31:41,180
on around 220 million Americans to
target potential donors and voters.
535
00:31:41,180 --> 00:31:43,860
It was, you know,
536
00:31:43,860 --> 00:31:48,940
a bunch of card tables that sat here
and a bunch of computers and people
537
00:31:48,940 --> 00:31:52,380
that were behind the computers,
monitoring the data.
538
00:31:52,380 --> 00:31:56,060
We've got to target this state, that
state or this universe or whatever.
539
00:31:56,060 --> 00:31:58,980
So that's what they were doing,
they were gathering all the data.
540
00:31:58,980 --> 00:32:02,780
A "universe" was the name
given to a group of voters
541
00:32:02,780 --> 00:32:06,180
defined by Cambridge Analytica.
542
00:32:06,180 --> 00:32:08,060
What sort of attributes
did these people have
543
00:32:08,060 --> 00:32:10,180
that they had been able to work out?
544
00:32:10,180 --> 00:32:13,460
Some of the attributes would be,
when was the last time they voted?
545
00:32:13,460 --> 00:32:15,260
Who did they vote for?
546
00:32:15,260 --> 00:32:17,780
You know, what kind of car
do they drive?
547
00:32:17,780 --> 00:32:21,420
What kind of things do they
look at on the internet?
548
00:32:21,420 --> 00:32:22,860
What do they stand for?
549
00:32:22,860 --> 00:32:25,900
I mean, one person might really be
about job creation
550
00:32:25,900 --> 00:32:27,740
and keeping jobs in America,
you know.
551
00:32:27,740 --> 00:32:30,900
Another person, they might
resonate with, you know,
552
00:32:30,900 --> 00:32:33,780
Second Amendment
and gun rights, so...
553
00:32:33,780 --> 00:32:35,260
How would they know that?
554
00:32:35,260 --> 00:32:38,980
..within that... How would they know
that? That is their secret sauce.
555
00:32:41,100 --> 00:32:42,780
Did the "secret sauce"
556
00:32:42,780 --> 00:32:47,620
contain predictions of personality
and political leanings?
557
00:32:47,620 --> 00:32:51,140
Were they able to kind of
understand people's personalities?
558
00:32:51,140 --> 00:32:52,900
Yes, I mean, you know,
559
00:32:52,900 --> 00:32:56,900
they do specialise
in psychographics, right?
560
00:32:56,900 --> 00:33:02,020
But based on personal interests and
based on what a person cares for
561
00:33:02,020 --> 00:33:04,100
and what means something to them,
562
00:33:04,100 --> 00:33:07,460
they were able to extract
and then we were able to target.
563
00:33:07,460 --> 00:33:11,380
So, the psychographic stuff,
were they using that here?
564
00:33:11,380 --> 00:33:14,220
Was that part of the model
you were working on?
565
00:33:14,220 --> 00:33:17,820
Well, I mean, towards the end
with the persuasion, absolutely.
566
00:33:17,820 --> 00:33:19,820
I mean, we really were targeting
567
00:33:19,820 --> 00:33:22,780
on these universes
that they had collected.
568
00:33:22,780 --> 00:33:26,100
I mean, did some of the attributes
that you were able to use
569
00:33:26,100 --> 00:33:29,900
from Cambridge Analytica have a sort
of emotional effect, you know,
570
00:33:29,900 --> 00:33:33,460
happy, sad, anxious, worried -
moods, that kind of thing?
571
00:33:33,460 --> 00:33:34,780
Right, yeah.
572
00:33:34,780 --> 00:33:39,620
I do know Cambridge Analytica
follows something called Ocean
573
00:33:39,620 --> 00:33:42,460
and it's based on, you know,
whether or not, like,
574
00:33:42,460 --> 00:33:44,500
are you an extrovert?
575
00:33:44,500 --> 00:33:46,500
Are you more of an introvert?
576
00:33:46,500 --> 00:33:51,380
Are you fearful? Are you positive?
577
00:33:51,380 --> 00:33:53,420
So, they did use that.
578
00:33:55,580 --> 00:33:58,940
Armed with Cambridge Analytica's
revolutionary insights,
579
00:33:58,940 --> 00:34:02,820
the next step in the battle
to win over millions of Americans
580
00:34:02,820 --> 00:34:06,740
was to shape the online messages
they would see.
581
00:34:06,740 --> 00:34:11,500
Now we're going to go into the big
kind of bull pen where a lot of
582
00:34:11,500 --> 00:34:15,580
the creatives were, and this is
where I was as well. Right.
583
00:34:15,580 --> 00:34:18,140
For today's modern
working-class families,
584
00:34:18,140 --> 00:34:20,340
the challenge
is very, very real.
585
00:34:20,340 --> 00:34:21,900
With childcare costs...
586
00:34:21,900 --> 00:34:25,780
Adverts were tailored to particular
audiences, defined by data.
587
00:34:25,780 --> 00:34:29,220
Donald Trump wants to give families
the break they deserve.
588
00:34:29,220 --> 00:34:31,860
This universe right here
that Cambridge Analytica,
589
00:34:31,860 --> 00:34:35,500
they've collected data and they
have identified as working mothers
590
00:34:35,500 --> 00:34:37,940
that are concerned about childcare,
591
00:34:37,940 --> 00:34:41,220
and childcare, obviously,
that's not going to be, like,
592
00:34:41,220 --> 00:34:44,220
a war-ridden, you know,
destructive ad, right?
593
00:34:44,220 --> 00:34:45,580
That's more warm and fuzzy,
594
00:34:45,580 --> 00:34:47,500
and this is what Trump is going
to do for you.
595
00:34:47,500 --> 00:34:50,420
But if you notice,
Trump's not speaking.
596
00:34:50,420 --> 00:34:52,420
He wasn't speaking,
he wasn't in it at all.
597
00:34:52,420 --> 00:34:54,180
Right, Trump wasn't speaking in it.
598
00:34:54,180 --> 00:34:57,460
That audience there, we wanted
a softer approach,
599
00:34:57,460 --> 00:35:00,380
so this is the type of approach
that we would take.
600
00:35:00,380 --> 00:35:04,540
The campaign made thousands
of different versions
601
00:35:04,540 --> 00:35:06,700
of the same fundraising adverts.
602
00:35:06,700 --> 00:35:11,860
The design was constantly tweaked
to see which version performed best.
603
00:35:11,860 --> 00:35:16,340
It wasn't uncommon to have
about 35-45,000 iterations
604
00:35:16,340 --> 00:35:19,140
of these types of ads
every day, right?
605
00:35:19,140 --> 00:35:23,460
So, you know, it could be
606
00:35:23,460 --> 00:35:28,740
as subtle and just, you know,
people wouldn't even notice,
607
00:35:28,740 --> 00:35:31,180
where you have the green button,
608
00:35:31,180 --> 00:35:33,620
sometimes a red button
would work better
609
00:35:33,620 --> 00:35:35,060
or a blue button would
work better.
610
00:35:35,060 --> 00:35:38,860
Now, the voters Cambridge Analytica
had targeted
611
00:35:38,860 --> 00:35:40,900
were bombarded with adverts.
612
00:35:46,980 --> 00:35:48,260
I may have short-circuited...
613
00:35:49,820 --> 00:35:52,260
I'm Donald Trump
and I approve this message.
614
00:35:53,500 --> 00:35:57,980
On a typical day, the campaign would
run more than 100 different adverts.
615
00:35:57,980 --> 00:36:02,820
The delivery system was Silicon
Valley's vast social networks.
616
00:36:02,820 --> 00:36:06,260
We had the Facebook
and YouTube and Google people,
617
00:36:06,260 --> 00:36:08,340
they would congregate here.
618
00:36:08,340 --> 00:36:13,500
Almost all of America's voters could
now be reached online in an instant.
619
00:36:13,500 --> 00:36:17,380
I mean, what were Facebook and
Google and YouTube people actually
620
00:36:17,380 --> 00:36:20,500
doing here, why were they here?
They were helping us, you know.
621
00:36:20,500 --> 00:36:25,580
They were basically our kind of
hands-on partners
622
00:36:25,580 --> 00:36:28,100
as far as being able to utilise
the platform
623
00:36:28,100 --> 00:36:30,260
as effectively as possible.
624
00:36:30,260 --> 00:36:34,860
The Trump campaign spent the lion's
share of its advertising budget,
625
00:36:34,860 --> 00:36:39,220
around $85 million, on Facebook.
626
00:36:39,220 --> 00:36:43,020
When you're pumping in millions
and millions of dollars
627
00:36:43,020 --> 00:36:44,900
to these social platforms,
628
00:36:44,900 --> 00:36:47,020
you're going to get
white-glove treatment,
629
00:36:47,020 --> 00:36:49,260
so, they would send people,
you know,
630
00:36:49,260 --> 00:36:51,900
representatives to, you know,
631
00:36:51,900 --> 00:36:56,900
Project Alamo to ensure that all
of our needs were being met.
632
00:36:56,900 --> 00:37:00,140
The success of Trump's digital
strategy was built on
633
00:37:00,140 --> 00:37:03,780
the effectiveness of Facebook
as an advertising medium.
634
00:37:03,780 --> 00:37:07,020
It's become a very powerful
political tool
635
00:37:07,020 --> 00:37:09,500
that's largely unregulated.
636
00:37:09,500 --> 00:37:12,780
Without Facebook,
we wouldn't have won.
637
00:37:12,780 --> 00:37:16,700
I mean, Facebook really
and truly put us over the edge,
638
00:37:16,700 --> 00:37:19,260
I mean, Facebook was the medium
639
00:37:19,260 --> 00:37:22,620
that proved most successful
for this campaign.
640
00:37:26,420 --> 00:37:29,420
Facebook didn't want to meet me,
but made it clear that like all
641
00:37:29,420 --> 00:37:33,260
advertisers on Facebook,
political campaigns must ensure
642
00:37:33,260 --> 00:37:37,220
their ads comply with all applicable
laws and regulations.
643
00:37:37,220 --> 00:37:40,740
The company also said
no personally identifiable
644
00:37:40,740 --> 00:37:43,500
information can be shared
with advertising,
645
00:37:43,500 --> 00:37:47,180
measurement or analytics partners
unless people give permission.
646
00:37:52,100 --> 00:37:53,740
In London, I'm on the trail
647
00:37:53,740 --> 00:37:57,060
of the data company
Cambridge Analytica.
648
00:38:00,500 --> 00:38:04,020
Ever since the American election,
the firm has been under pressure
649
00:38:04,020 --> 00:38:09,100
to come clean over its use
of personality prediction.
650
00:38:09,100 --> 00:38:12,820
I want to know exactly how the firm
used psychographics
651
00:38:12,820 --> 00:38:17,060
to target voters for
the Trump campaign.
652
00:38:17,060 --> 00:38:18,700
Alexander. Hi.
653
00:38:18,700 --> 00:38:20,780
How do you do? Pleasure.
654
00:38:20,780 --> 00:38:24,420
VOICEOVER: Alexander Nix's firm
first used data to target voters
655
00:38:24,420 --> 00:38:28,220
in the American presidential
elections while working on
656
00:38:28,220 --> 00:38:32,460
the Ted Cruz campaign
for the Republican nomination.
657
00:38:32,460 --> 00:38:35,940
When Cruz lost,
they went to work for Trump.
658
00:38:35,940 --> 00:38:38,020
I want to start with
the Trump campaign.
659
00:38:38,020 --> 00:38:40,780
Did Cambridge Analytica ever use
660
00:38:40,780 --> 00:38:44,860
psychometric or psychographic
methods in this campaign?
661
00:38:44,860 --> 00:38:48,580
We left the Cruz campaign in April
after the nomination was over.
662
00:38:48,580 --> 00:38:51,020
We pivoted right across
onto the Trump campaign.
663
00:38:51,020 --> 00:38:54,300
It was about five and a half
months before polling.
664
00:38:54,300 --> 00:38:59,460
And whilst on the Cruz campaign,
we were able to invest
665
00:38:59,460 --> 00:39:02,020
a lot more time
into building psychographic models,
666
00:39:02,020 --> 00:39:05,380
into profiling, using behavioural
profiling to understand different
667
00:39:05,380 --> 00:39:08,300
personality groups
and different personality drivers,
668
00:39:08,300 --> 00:39:11,540
in order to inform
our messaging and our creative.
669
00:39:11,540 --> 00:39:15,220
We simply didn't have the time
to employ this level of
670
00:39:15,220 --> 00:39:17,740
rigorous methodology for Trump.
671
00:39:17,740 --> 00:39:22,300
For Cruz, Cambridge Analytica built
computer models which could crunch
672
00:39:22,300 --> 00:39:27,340
huge amounts of data on each voter,
including psychographic data
673
00:39:27,340 --> 00:39:31,180
predicting voters'
personality types.
674
00:39:31,180 --> 00:39:33,260
When the firm transferred
to the Trump campaign,
675
00:39:33,260 --> 00:39:35,540
they took data with them.
676
00:39:35,540 --> 00:39:41,180
Now, there is clearly some legacy
psychographics in the data,
677
00:39:41,180 --> 00:39:44,860
because the data is, um, model data,
678
00:39:44,860 --> 00:39:47,500
or a lot of it is model data
that we had used across
679
00:39:47,500 --> 00:39:51,180
the last 14, 15 months
of campaigning through the midterms,
680
00:39:51,180 --> 00:39:53,380
and then through the primaries.
681
00:39:53,380 --> 00:39:58,460
But specifically, did we build
specific psychographic models
682
00:39:58,460 --> 00:40:00,500
for the Trump campaign?
No, we didn't.
683
00:40:00,500 --> 00:40:03,940
So you didn't build specific models
for this campaign,
684
00:40:03,940 --> 00:40:08,420
but it sounds like you did use some
element of psychographic modelling,
685
00:40:08,420 --> 00:40:11,660
as an approach
in the Trump campaign?
686
00:40:11,660 --> 00:40:14,700
Only as a result
of legacy data models.
687
00:40:14,700 --> 00:40:17,980
So, the answer... The answer
you are looking for is no.
688
00:40:17,980 --> 00:40:21,060
The answer I am looking for
is the extent to which it was used.
689
00:40:21,060 --> 00:40:24,260
I mean, legacy... I don't know what
that means, legacy data modelling.
690
00:40:24,260 --> 00:40:26,700
What does that mean
for the Trump campaign?
691
00:40:26,700 --> 00:40:29,300
Well, so we were able to take models
we had made previously
692
00:40:29,300 --> 00:40:30,980
over the last two or three years,
693
00:40:30,980 --> 00:40:33,820
and integrate those into some of
the work we were doing.
694
00:40:33,820 --> 00:40:35,660
Where did all the information
695
00:40:35,660 --> 00:40:39,620
to predict voters' personalities
come from?
696
00:40:39,620 --> 00:40:43,540
Very originally, we used a
combination of telephone surveys
697
00:40:43,540 --> 00:40:46,580
and then we used a number of...
698
00:40:47,620 --> 00:40:50,780
..online platforms...
699
00:40:50,780 --> 00:40:53,300
for gathering questions.
700
00:40:53,300 --> 00:40:56,180
As we started to gather more data,
701
00:40:56,180 --> 00:40:58,540
we started to look
at other platforms.
702
00:40:58,540 --> 00:41:01,020
Such as Facebook, for instance.
703
00:41:03,060 --> 00:41:06,220
Predicting personality
is just one element in the big data
704
00:41:06,220 --> 00:41:10,340
companies like Cambridge Analytica
are using to revolutionise
705
00:41:10,340 --> 00:41:12,460
the way democracy works.
706
00:41:12,460 --> 00:41:14,740
Can you understand, though,
why maybe
707
00:41:14,740 --> 00:41:16,820
some people find it
a little bit creepy?
708
00:41:16,820 --> 00:41:18,340
No, I can't. Quite the opposite.
709
00:41:18,340 --> 00:41:21,740
I think that the move away
from blanket advertising,
710
00:41:21,740 --> 00:41:25,180
the move towards ever more
personalised communication,
711
00:41:25,180 --> 00:41:26,620
is a very natural progression.
712
00:41:26,620 --> 00:41:28,500
I think it is only going
to increase.
713
00:41:28,500 --> 00:41:31,700
I find it a little bit weird,
if I had a very sort of detailed
714
00:41:31,700 --> 00:41:35,780
personality assessment of me,
based on all sorts of different data
715
00:41:35,780 --> 00:41:38,220
that I had put all over
the internet,
716
00:41:38,220 --> 00:41:41,580
and that as a result of that,
some profile had been made of me
717
00:41:41,580 --> 00:41:44,100
that was then used to target me
for adverts,
718
00:41:44,100 --> 00:41:46,660
and I didn't really know
that any of that had happened.
719
00:41:46,660 --> 00:41:49,580
That's why some people might find it
a little bit sinister.
720
00:41:49,580 --> 00:41:51,020
Well, you have just said yourself,
721
00:41:51,020 --> 00:41:53,460
you are putting this data out
into the public domain.
722
00:41:53,460 --> 00:41:57,420
I'm sure that you have
a supermarket loyalty card.
723
00:41:57,420 --> 00:42:02,300
I'm sure you understand the
reciprocity that is going on there -
724
00:42:02,300 --> 00:42:05,780
you get points, and in return,
they gather your data
725
00:42:05,780 --> 00:42:08,060
on your consumer behaviour.
726
00:42:08,060 --> 00:42:09,700
I mean, we are talking about
politics
727
00:42:09,700 --> 00:42:12,780
and we're talking about shopping.
Are they really the same thing?
728
00:42:12,780 --> 00:42:14,940
The technology is the same.
729
00:42:14,940 --> 00:42:16,380
In the next ten years,
730
00:42:16,380 --> 00:42:19,260
the sheer volumes of data
that are going to be available,
731
00:42:19,260 --> 00:42:21,580
that are going to be driving
all sorts of things
732
00:42:21,580 --> 00:42:23,900
including marketing and
communications,
733
00:42:23,900 --> 00:42:26,700
is going to be a paradigm shift
from where we are now
734
00:42:26,700 --> 00:42:28,220
and it's going to be a revolution,
735
00:42:28,220 --> 00:42:30,180
and that is the way the world
is moving.
736
00:42:30,180 --> 00:42:32,180
And, you know, I think,
737
00:42:32,180 --> 00:42:35,660
whether you like it or not,
it, it...
738
00:42:35,660 --> 00:42:38,460
it is an inevitable fact.
739
00:42:39,820 --> 00:42:43,620
To the new data barons,
this is all just business.
740
00:42:43,620 --> 00:42:46,700
But to the rest of us,
it's more than that.
741
00:42:50,020 --> 00:42:52,740
By the time
of Donald Trump's inauguration,
742
00:42:52,740 --> 00:42:56,500
it was accepted that his mastery
of data and social media
743
00:42:56,500 --> 00:43:00,540
had made him the most powerful man
in the world.
744
00:43:00,540 --> 00:43:05,060
We will make America
great again.
745
00:43:05,060 --> 00:43:07,180
The election of Donald Trump
746
00:43:07,180 --> 00:43:11,020
was greeted with barely concealed
fury in Silicon Valley.
747
00:43:11,020 --> 00:43:14,060
But Facebook and other
tech companies had made
748
00:43:14,060 --> 00:43:17,340
millions of dollars by helping
to make it happen.
749
00:43:17,340 --> 00:43:19,780
Their power as
advertising platforms
750
00:43:19,780 --> 00:43:22,620
had been exploited
by a politician
751
00:43:22,620 --> 00:43:25,460
with a very different view
of the world.
752
00:43:28,340 --> 00:43:31,020
But Facebook's problems
were only just beginning.
753
00:43:33,140 --> 00:43:34,820
Another phenomenon of the election
754
00:43:34,820 --> 00:43:38,220
was plunging the tech titan
into crisis.
755
00:43:44,140 --> 00:43:47,260
Fake news, often targeting
Hillary Clinton,
756
00:43:47,260 --> 00:43:50,420
had dominated the election campaign.
757
00:43:52,820 --> 00:43:56,260
Now, the departing President
turned on the social media giant
758
00:43:56,260 --> 00:43:58,700
he had once embraced.
759
00:43:59,980 --> 00:44:02,540
In an age where, uh...
760
00:44:02,540 --> 00:44:07,460
there is so much
active misinformation,
761
00:44:07,460 --> 00:44:09,180
and it is packaged very well,
762
00:44:09,180 --> 00:44:12,540
and it looks the same
when you see it on a Facebook page,
763
00:44:12,540 --> 00:44:14,660
or you turn on your television,
764
00:44:14,660 --> 00:44:17,940
if everything, uh...
765
00:44:17,940 --> 00:44:20,700
seems to be the same,
766
00:44:20,700 --> 00:44:22,940
and no distinctions are made,
767
00:44:22,940 --> 00:44:26,140
then we won't know what to protect.
768
00:44:26,140 --> 00:44:28,500
We won't know what to fight for.
769
00:44:34,140 --> 00:44:37,140
Fake news had provoked
a storm of criticism
770
00:44:37,140 --> 00:44:40,180
over Facebook's impact on democracy.
771
00:44:40,180 --> 00:44:43,500
Its founder, Zuck, claimed it was
extremely unlikely
772
00:44:43,500 --> 00:44:46,900
fake news had changed
the election's outcome.
773
00:44:46,900 --> 00:44:49,780
But he didn't address why
it had spread like wildfire
774
00:44:49,780 --> 00:44:52,140
across the platform.
775
00:44:52,140 --> 00:44:53,980
Jamie. Hi, Jamie.
776
00:44:53,980 --> 00:44:55,860
VOICEOVER: Meet Jeff Hancock,
777
00:44:55,860 --> 00:44:59,380
a psychologist who has investigated
a hidden aspect of Facebook
778
00:44:59,380 --> 00:45:04,100
that helps explain how the platform
became weaponised in this way.
779
00:45:07,220 --> 00:45:12,060
It turns out the power of Facebook
to affect our emotions is key,
780
00:45:12,060 --> 00:45:14,340
something that had been uncovered
781
00:45:14,340 --> 00:45:18,780
in an experiment the company itself
had run in 2012.
782
00:45:18,780 --> 00:45:20,460
It was one of the earlier, you know,
783
00:45:20,460 --> 00:45:23,460
what we would call big data,
social science-type studies.
784
00:45:26,940 --> 00:45:31,620
The newsfeeds of nearly 700,000
users were secretly manipulated
785
00:45:31,620 --> 00:45:36,620
so they would see
fewer positive or negative posts.
786
00:45:36,620 --> 00:45:39,100
Jeff helped interpret the results.
787
00:45:39,100 --> 00:45:41,100
So, what did you actually find?
788
00:45:41,100 --> 00:45:43,780
We found that if you were one
of those people
789
00:45:43,780 --> 00:45:47,020
that were seeing less negative
emotion words in their posts,
790
00:45:47,020 --> 00:45:50,620
then you would write with less
negative emotion in your own posts,
791
00:45:50,620 --> 00:45:52,540
and more positive emotion.
792
00:45:52,540 --> 00:45:56,380
This is emotional contagion.
And what about positive posts?
793
00:45:56,380 --> 00:45:58,380
Did that have the same kind
of effect?
794
00:45:58,380 --> 00:46:00,620
Yeah, we saw the same effect
795
00:46:00,620 --> 00:46:04,300
when positive emotion worded posts
were decreased.
796
00:46:04,300 --> 00:46:07,140
We saw the same thing. So I would
produce fewer positive
797
00:46:07,140 --> 00:46:09,580
emotion words, and more negative
emotion words.
798
00:46:09,580 --> 00:46:12,260
And that is consistent with
emotional contagion theory.
799
00:46:12,260 --> 00:46:16,300
Basically, we were showing that
people were writing in a way
800
00:46:16,300 --> 00:46:19,180
that was matching the emotion
that they were seeing
801
00:46:19,180 --> 00:46:21,980
in the Facebook news feed.
802
00:46:21,980 --> 00:46:28,260
Emotion draws people to fake news,
and then supercharges its spread.
803
00:46:28,260 --> 00:46:29,900
So the more emotional the content,
804
00:46:29,900 --> 00:46:32,540
the more likely it is
to spread online.
805
00:46:32,540 --> 00:46:35,380
Is that true? Yeah, the more intense
the emotion in content,
806
00:46:35,380 --> 00:46:39,020
the more likely it is to spread,
to go viral.
807
00:46:39,020 --> 00:46:42,900
It doesn't matter whether it is sad
or happy, like negative or positive,
808
00:46:42,900 --> 00:46:45,580
the more important thing
is how intense the emotion is.
809
00:46:45,580 --> 00:46:47,780
The process of emotional contagion
810
00:46:47,780 --> 00:46:50,780
helps explain why fake news
has spread
811
00:46:50,780 --> 00:46:53,620
so far across social media.
812
00:46:55,820 --> 00:46:58,700
Advertisers have been well aware
of how emotion can be used
813
00:46:58,700 --> 00:47:00,660
to manipulate people's attention.
814
00:47:00,660 --> 00:47:03,180
But now we are seeing this with
a whole host of other actors,
815
00:47:03,180 --> 00:47:04,700
some of them nefarious.
816
00:47:04,700 --> 00:47:07,180
So, other state actors trying
to influence your election,
817
00:47:07,180 --> 00:47:09,140
people trying
to manipulate the media
818
00:47:09,140 --> 00:47:11,460
are using emotional contagion,
819
00:47:11,460 --> 00:47:14,980
and also using those original
platforms like Facebook
820
00:47:14,980 --> 00:47:19,820
to accomplish other objectives,
like sowing distrust,
821
00:47:19,820 --> 00:47:21,820
or creating false beliefs.
822
00:47:21,820 --> 00:47:24,380
You see it sort of,
maybe weaponised
823
00:47:24,380 --> 00:47:26,940
and being used at scale.
824
00:47:36,380 --> 00:47:37,780
I'm in Germany,
825
00:47:37,780 --> 00:47:40,380
where the presence of more
than a million new refugees
826
00:47:40,380 --> 00:47:43,660
has caused tension
across the political spectrum.
827
00:47:43,660 --> 00:47:45,660
Earlier this year,
828
00:47:45,660 --> 00:47:49,980
a story appeared on the American
alt-right news site Breitbart
829
00:47:49,980 --> 00:47:54,820
about the torching of a church
in Dortmund by a North African mob.
830
00:47:54,820 --> 00:47:58,220
It was widely shared
on social media -
831
00:47:58,220 --> 00:48:00,060
but it wasn't true.
832
00:48:05,500 --> 00:48:07,180
The problem with social networks
833
00:48:07,180 --> 00:48:09,420
is that all information
is treated equally.
834
00:48:09,420 --> 00:48:13,260
So you have good, honest,
accurate information
835
00:48:13,260 --> 00:48:16,500
sitting alongside
and treated equally to
836
00:48:16,500 --> 00:48:19,100
lies and propaganda.
837
00:48:19,100 --> 00:48:22,220
And the difficulty for citizens
is that it can be very hard
838
00:48:22,220 --> 00:48:25,060
to tell the difference
between the two.
839
00:48:25,060 --> 00:48:30,980
But one data scientist here has
discovered an even darker side
840
00:48:30,980 --> 00:48:33,220
to the way Facebook
is being manipulated.
841
00:48:34,620 --> 00:48:37,740
We created a network
of all the likes
842
00:48:37,740 --> 00:48:40,860
in the refugee debate on Facebook.
843
00:48:43,100 --> 00:48:46,780
Professor Simon Hegelich has
found evidence the debate
844
00:48:46,780 --> 00:48:48,620
about refugees on Facebook
845
00:48:48,620 --> 00:48:53,460
is being skewed by anonymous
political forces.
846
00:48:53,460 --> 00:48:57,500
So you are trying to understand
who is liking pages?
847
00:48:57,500 --> 00:48:58,780
Exactly.
848
00:48:58,780 --> 00:49:01,900
One statistic among many used
by Facebook to rank stories in
849
00:49:01,900 --> 00:49:04,900
your news feed is the number
of likes they get.
850
00:49:04,900 --> 00:49:09,660
The red area of this chart shows
a network of people liking
851
00:49:09,660 --> 00:49:12,900
anti-refugee posts on Facebook.
852
00:49:12,900 --> 00:49:19,100
Most of the likes are sent by just a
handful of people - 25 people are...
853
00:49:20,860 --> 00:49:24,780
..responsible for more than 90%
of all these likes.
854
00:49:24,780 --> 00:49:27,220
25 Facebook accounts each liked
855
00:49:27,220 --> 00:49:31,100
more than 30,000 comments
over six months.
856
00:49:31,100 --> 00:49:37,180
These hyperactive accounts could be
run by real people, or software.
857
00:49:37,180 --> 00:49:39,100
I think the rationale behind this
858
00:49:39,100 --> 00:49:43,060
is that they tried to make
their content viral.
859
00:49:43,060 --> 00:49:46,660
They think if we like all
this stuff, then Facebook,
860
00:49:46,660 --> 00:49:48,540
the algorithm of Facebook,
861
00:49:48,540 --> 00:49:52,220
will pick up our content
and show it to other users,
862
00:49:52,220 --> 00:49:55,660
and then the whole world sees
that refugees are bad
863
00:49:55,660 --> 00:49:58,220
and that they shouldn't come
to Germany.
864
00:49:58,220 --> 00:50:02,860
This is evidence the number of likes
on Facebook can be easily gamed
865
00:50:02,860 --> 00:50:05,900
as part of an effort to try
to influence the prominence
866
00:50:05,900 --> 00:50:09,140
of anti-refugee content on the site.
867
00:50:09,140 --> 00:50:11,260
Does this worry you, though?
868
00:50:11,260 --> 00:50:15,380
It's definitely changing
structure of public opinion.
869
00:50:15,380 --> 00:50:19,220
Democracy is built
on public opinion,
870
00:50:19,220 --> 00:50:24,260
so such a change definitely has
to change the way democracy works.
871
00:50:27,340 --> 00:50:30,820
Facebook told us they are working
to disrupt the economic incentives
872
00:50:30,820 --> 00:50:35,020
behind false news, removing tens
of thousands of fake accounts,
873
00:50:35,020 --> 00:50:36,900
and building new products
874
00:50:36,900 --> 00:50:40,380
to identify and limit the spread
of false news.
875
00:50:40,380 --> 00:50:44,820
Zuck is trying to hold the line that
his company is not a publisher,
876
00:50:44,820 --> 00:50:50,260
based on that obscure legal clause
from the 1990s.
877
00:50:50,260 --> 00:50:52,340
You know, Facebook is a new kind
of platform.
878
00:50:52,340 --> 00:50:55,540
You know, it's not a traditional
technology company,
879
00:50:55,540 --> 00:50:58,420
it's not a traditional
media company.
880
00:50:58,420 --> 00:51:00,540
Um, we don't write the news
that people...
881
00:51:00,540 --> 00:51:02,620
that people read on the platform.
882
00:51:04,740 --> 00:51:09,100
In Germany, one man is taking on the
tech gods over their responsibility
883
00:51:09,100 --> 00:51:11,580
for what appears on their sites.
884
00:51:13,420 --> 00:51:17,740
Ulrich Kelber is a minister
in the Justice Department.
885
00:51:17,740 --> 00:51:21,340
He was once called a "Jewish pig"
in a post on Facebook.
886
00:51:21,340 --> 00:51:25,380
He reported it to the company,
who refused to delete it.
887
00:51:25,380 --> 00:51:27,140
IN GERMAN:
888
00:51:44,420 --> 00:51:45,980
Under a new law,
889
00:51:45,980 --> 00:51:48,380
Facebook and other
social networking sites
890
00:51:48,380 --> 00:51:51,100
could be fined
up to 50 million euros
891
00:51:51,100 --> 00:51:56,020
if they fail to take down
hate speech posts that are illegal.
892
00:51:56,020 --> 00:51:57,420
Is this too much?
893
00:51:57,420 --> 00:51:59,940
Is this a little bit Draconian?
894
00:52:20,740 --> 00:52:24,340
This is the first time a government
has challenged
895
00:52:24,340 --> 00:52:27,780
the principles underlying
a Silicon Valley platform.
896
00:52:27,780 --> 00:52:30,300
Once this thread has been pulled,
897
00:52:30,300 --> 00:52:33,460
their whole world
could start to unravel.
898
00:52:33,460 --> 00:52:35,740
They must find you a pain.
899
00:52:35,740 --> 00:52:37,820
I mean, this must be annoying
for them.
900
00:52:37,820 --> 00:52:39,340
It must be.
901
00:52:48,140 --> 00:52:52,020
Facebook told us they share
the goal of fighting hate speech,
902
00:52:52,020 --> 00:52:55,380
and they have made
substantial progress
903
00:52:55,380 --> 00:52:57,260
in removing illegal content,
904
00:52:57,260 --> 00:53:01,100
adding 3,000 people to their
community operations team.
905
00:53:05,180 --> 00:53:09,260
Facebook now connects more than
two billion people around the world,
906
00:53:09,260 --> 00:53:12,940
including more and more voters
in the West.
907
00:53:12,940 --> 00:53:15,140
When you think of what
Facebook has become,
908
00:53:15,140 --> 00:53:19,020
in such a short space of time,
it's actually pretty bizarre.
909
00:53:19,020 --> 00:53:20,820
I mean, this was just a platform
910
00:53:20,820 --> 00:53:23,940
for sharing photos
or chatting to friends,
911
00:53:23,940 --> 00:53:27,740
but in less than a decade,
it has become a platform that has
912
00:53:27,740 --> 00:53:31,340
dramatic implications for how
our democracy works.
913
00:53:33,220 --> 00:53:36,060
Old structures of power
are falling away.
914
00:53:36,060 --> 00:53:40,740
Social media is giving ordinary
people access to huge audiences.
915
00:53:40,740 --> 00:53:43,740
And politics is changing
as a result.
916
00:53:46,780 --> 00:53:48,740
Significant numbers were motivated
917
00:53:48,740 --> 00:53:51,060
through social media
to vote for Brexit.
918
00:53:53,060 --> 00:53:55,420
Jeremy Corbyn lost
the general election,
919
00:53:55,420 --> 00:53:58,700
but enjoyed unexpected gains,
920
00:53:58,700 --> 00:54:03,820
an achievement he put down in part
to the power of social media.
921
00:54:07,540 --> 00:54:12,820
One influential force in this new
world is found here in Bristol.
922
00:54:12,820 --> 00:54:16,420
The Canary is an online
political news outlet.
923
00:54:16,420 --> 00:54:18,900
During the election campaign,
924
00:54:18,900 --> 00:54:23,380
their stories got more than
25 million hits on a tiny budget.
925
00:54:24,820 --> 00:54:28,420
So, how much of your readership
comes through Facebook?
926
00:54:28,420 --> 00:54:30,900
It's really high, it's about 80%.
927
00:54:30,900 --> 00:54:34,260
So it is an enormously important
distribution mechanism.
928
00:54:34,260 --> 00:54:35,900
And free.
929
00:54:35,900 --> 00:54:40,380
The Canary's presentation
of its pro-Corbyn news
930
00:54:40,380 --> 00:54:44,060
is tailored to social media.
931
00:54:44,060 --> 00:54:46,540
I have noticed that a lot of your
headlines, or your pictures,
932
00:54:46,540 --> 00:54:48,380
they are quite emotional.
933
00:54:48,380 --> 00:54:51,180
They are sort of
pictures of sad Theresa May,
934
00:54:51,180 --> 00:54:52,740
or delighted Jeremy Corbyn.
935
00:54:52,740 --> 00:54:54,860
Is that part of the purpose
of it all?
936
00:54:54,860 --> 00:54:58,340
Yeah, it has to be. We are out there
trying to have a conversation
937
00:54:58,340 --> 00:55:01,580
with a lot of people, so it is on us
to be compelling.
938
00:55:01,580 --> 00:55:04,940
Human beings work on facts,
but they also work on gut instinct,
939
00:55:04,940 --> 00:55:09,860
they work on emotions, feelings,
and fidelity and community.
940
00:55:09,860 --> 00:55:11,500
All of these issues.
941
00:55:11,500 --> 00:55:16,220
Social media enables those
with few resources to compete
942
00:55:16,220 --> 00:55:21,500
with the mainstream media for
the attention of millions of us.
943
00:55:21,500 --> 00:55:24,540
You put up quite sort of
clickbait-y stories, you know,
944
00:55:24,540 --> 00:55:26,740
the headlines are there
to get clicks.
945
00:55:26,740 --> 00:55:29,220
Yeah. Um... Is that a fair
criticism?
946
00:55:29,220 --> 00:55:30,780
Of course they're there
to get clicks.
947
00:55:30,780 --> 00:55:32,900
We don't want to have a conversation
with ten people.
948
00:55:32,900 --> 00:55:34,860
You can't change the world
talking to ten people.
949
00:55:43,540 --> 00:55:48,700
The tech gods are giving all of us
the power to influence the world.
950
00:55:48,700 --> 00:55:51,140
Connectivity and access...
Connect everyone in the world...
951
00:55:51,140 --> 00:55:52,580
Make the world more open
and connected...
952
00:55:52,580 --> 00:55:54,780
You connect people over time...
You get people connectivity...
953
00:55:54,780 --> 00:55:57,980
That's the mission.
That's what I care about.
954
00:55:57,980 --> 00:56:01,860
Social media's unparalleled
power to persuade,
955
00:56:01,860 --> 00:56:03,940
first developed for advertisers,
956
00:56:03,940 --> 00:56:07,620
is now being exploited by
political forces of all kinds.
957
00:56:07,620 --> 00:56:11,820
Grassroots movements are regaining
their power,
958
00:56:11,820 --> 00:56:14,300
challenging political elites.
959
00:56:14,300 --> 00:56:19,780
Extremists are discovering new ways
to stoke hatred and spread lies.
960
00:56:19,780 --> 00:56:22,220
And wealthy political parties
are developing the ability
961
00:56:22,220 --> 00:56:25,060
to manipulate our thoughts
and feelings
962
00:56:25,060 --> 00:56:27,820
using powerful
psychological tools,
963
00:56:27,820 --> 00:56:32,740
which is leading to a world of
unexpected political opportunity,
964
00:56:32,740 --> 00:56:34,620
and turbulence.
965
00:56:34,620 --> 00:56:38,460
I think the people that connected
the world really believed that
966
00:56:38,460 --> 00:56:43,540
somehow, just by us being connected,
our politics would be better.
967
00:56:43,540 --> 00:56:49,660
But the world is changing
in ways that they never imagined,
968
00:56:49,660 --> 00:56:52,100
and they are probably
not happy about any more.
969
00:56:52,100 --> 00:56:53,860
But in truth,
970
00:56:53,860 --> 00:56:57,580
they are no more in charge of this
technology than any of us are now.
971
00:56:57,580 --> 00:57:01,380
Silicon Valley's philosophy
is called disruption.
972
00:57:01,380 --> 00:57:03,460
Breaking down the way we do things
973
00:57:03,460 --> 00:57:07,100
and using technology
to improve the world.
974
00:57:07,100 --> 00:57:08,940
In this series, I have seen how
975
00:57:08,940 --> 00:57:12,420
sharing platforms
like Uber and Airbnb
976
00:57:12,420 --> 00:57:14,980
are transforming our cities.
977
00:57:18,900 --> 00:57:21,980
And how automation and
artificial intelligence
978
00:57:21,980 --> 00:57:24,980
threaten to destroy
millions of jobs.
979
00:57:24,980 --> 00:57:28,220
Within 30 years, half
of humanity won't have a job.
980
00:57:28,220 --> 00:57:30,660
It could get ugly,
there could be revolution.
981
00:57:30,660 --> 00:57:35,540
Now, the technology to connect the
world unleashed by a few billionaire
982
00:57:35,540 --> 00:57:40,140
entrepreneurs is having a dramatic
influence on our politics.
983
00:57:40,140 --> 00:57:43,700
The people who are responsible
for building this technology,
984
00:57:43,700 --> 00:57:48,540
for unleashing this disruption
onto all of us,
985
00:57:48,540 --> 00:57:52,180
don't ever feel like they are
responsible for the consequences
986
00:57:52,180 --> 00:57:54,060
of any of that.
987
00:57:54,060 --> 00:57:59,340
They retain this absolute religious
faith that technology and
988
00:57:59,340 --> 00:58:03,980
connectivity is always going to make
things turn out for the best.
989
00:58:03,980 --> 00:58:05,660
And it doesn't matter what happens,
990
00:58:05,660 --> 00:58:08,700
it doesn't matter how much
that's proven not to be the case,
991
00:58:08,700 --> 00:58:11,060
they still believe.
992
00:58:23,300 --> 00:58:25,940
How did Silicon Valley
become so influential?
993
00:58:25,940 --> 00:58:28,740
The Open University has produced
an interactive timeline
994
00:58:28,740 --> 00:58:31,460
exploring the history of this place.
995
00:58:31,460 --> 00:58:33,460
To find out more, visit...
996
00:58:37,700 --> 00:58:40,700
..and follow the links
to the Open University.
85722
Can't find what you're looking for?
Get subtitles in any language from opensubtitles.com, and translate them here.