00:00:00
If I told you uh sincerely that Hillary
00:00:04
Clinton ran a child pornography
00:00:07
web out of a pizza parlor in Virginia
00:00:10
and had dozens of kids chained to the
00:00:13
wall, you would either find it amusing
00:00:18
cuz you know it's false or you would
00:00:20
find it terrifying because you know
00:00:22
there are people who would believe it.
00:00:24
Welcome to the Ripple Effect, the
00:00:26
podcast that takes you on a journey
00:00:28
through the minds of Wharton faculty.
00:00:30
I'm your host, Dan Looney, and in each
00:00:32
episode, we'll be diving deep into the
00:00:34
inspiration [music]
00:00:35
behind the groundbreaking research that
00:00:37
Wharton professors have conducted and
00:00:39
exploring how their findings resonate
00:00:41
with the world today. Well, there is
00:00:44
really no doubt that we have seen social
00:00:46
media platforms [music] have to deal
00:00:48
with growing levels of disinformation in
00:00:51
recent years and on a wide range of
00:00:54
topics. The reasons why this is
00:00:55
occurring can obviously be political,
00:00:58
can be profit centered and there are so
00:01:00
many other reasons. But how can these
00:01:02
issues of diff disinformation be dealt
00:01:05
with? Pleasure to be joined by Eric
00:01:07
Clemens who's professor of operations,
00:01:09
information, and decisions here at the
00:01:11
Wharton School and he was part on new
00:01:13
research on this topic earlier this
00:01:15
year. Eric, always great to chat. How
00:01:17
are you, sir?
00:01:18
>> Uh, I'm okay. I'm missing a front tooth
00:01:21
so I hope I don't look too goofy for
00:01:24
your audience.
00:01:25
>> You're good. You're good, Eric.
00:01:26
>> I'll try to avoid smiling. This is
00:01:29
obviously such an important topic right
00:01:30
now, but let me start with something
00:01:33
that is noted right at the top of the
00:01:36
research and you talk about the
00:01:38
component of campaigns and how much the
00:01:41
the the concept of a campaign is really
00:01:45
involved in much of these issues around
00:01:49
disinformation.
00:01:50
>> Yeah. Uh one of the things that's
00:01:52
interesting is how effective it is. uh
00:01:57
disinformation about climate change is
00:02:00
easy. You tell a bunch of coal miners
00:02:03
that it's a Chinese plot to uh
00:02:07
the American coal industry, increase the
00:02:10
cost of American production, and
00:02:12
increase uh Chinese welfare, Chinese
00:02:14
well-being. You tell someone who lives
00:02:17
in a uh coastal community that uh just
00:02:22
because somebody missed seven foul
00:02:24
shots, uh you don't trade them,
00:02:26
>> right?
00:02:27
>> Uh bad bad stuff happens and a couple of
00:02:30
bad hurricanes in a row doesn't mean you
00:02:32
should sell the house or get rid of your
00:02:34
SUV. That's easy to say. The explanation
00:02:38
for climate change is so complicated. Uh
00:02:43
we all know how a microwave works. The
00:02:46
radiation in a microwave makes water
00:02:49
vibrate, which is the definition of
00:02:51
heat. Radiation from the earth makes
00:02:53
carbon dioxide vibrate, which is the
00:02:56
definition of heat. Why we don't melt.
00:02:59
Uh it's called black body radiation.
00:03:01
Eventually, you reach a new equilibrium.
00:03:04
I gave my class uh an experiment which
00:03:07
was I told them a couple of climate
00:03:10
change lies
00:03:12
which were easy to repeat and I told
00:03:15
them some climate change truth
00:03:18
which is hedged and incredibly
00:03:20
complicated
00:03:22
and none of them could repeat it. I was
00:03:25
stuck on an airplane and the poor
00:03:27
passenger stuck between me and the aisle
00:03:30
had just bought a house uh off the Outer
00:03:32
Banks. And at the end of two hours, he
00:03:35
said, "Let me see if I understand. If
00:03:38
I'm very lucky and every bit of data we
00:03:41
have on carbon dioxide is false, and if
00:03:44
cloud cover cools the Earth, I'm fine.
00:03:47
Otherwise, my house is going to be
00:03:49
underwater in 10 years." And I said,
00:03:52
"Yeah, it could be okay." But he said,
00:03:54
"I'm selling. Why doesn't every American
00:03:57
in in why doesn't every American know
00:03:59
this?" And I said, "You were stuck for
00:04:02
two hours with nothing to do but listen
00:04:04
to me."
00:04:05
>> Right?
00:04:05
>> I don't have that opportunity, and I
00:04:07
don't have the time to talk to 20
00:04:09
million voters. But the lie is easy and
00:04:13
effective. The truth is long, slow, and
00:04:16
complicated. Makes it a very effective
00:04:18
way to change opinions. And so when you
00:04:21
put it in a campaign form, it becomes
00:04:23
that much more effective.
00:04:26
>> It it it does. Now, a lot of uh the
00:04:29
things that are banned in uh
00:04:33
disinformation campaign are sort of
00:04:36
banned in old marketing law. But this is
00:04:40
so complicated. It's moved so quickly.
00:04:43
uh platforms have so many exemptions and
00:04:46
the profit motive to be paid to
00:04:48
facilitate a campaign is so high that
00:04:51
it's really been almost impossible to
00:04:53
control.
00:04:53
>> And so one of the other things you
00:04:55
mention is that uh social media content
00:04:58
this disinformation in many cases is
00:05:01
quote unquote highly processed similar
00:05:04
to many of the foods that we eat.
00:05:06
Correct.
00:05:07
>> I love the analogy. We are finally
00:05:09
understanding that highly processed food
00:05:12
isn't the same as as a balanced diet and
00:05:15
highly processed news isn't the same as
00:05:17
an informed uh electorate. Uh that's
00:05:20
it's exactly right and I I should
00:05:23
explain a little bit what highly
00:05:25
processed means in the context of of
00:05:28
news. The first part is I know what
00:05:32
you're inclined to believe. I know what
00:05:34
you know. The amount of information
00:05:36
available on every social media users is
00:05:40
incredible. Recent studies have shown
00:05:43
that social media models can predict the
00:05:46
behavior of a user more accurately than
00:05:50
their spouse. I mean it's it's we we
00:05:52
know enough and if we share that
00:05:55
information with a highly motivated
00:05:58
liar, let's call it leave us out of it
00:06:00
and call it a North Korean troll
00:06:02
factory, they can write stories that
00:06:05
will appeal very closely to individual
00:06:09
segments of the market without actually
00:06:12
needing to violate any individual's
00:06:15
privacy.
00:06:16
So I now know enough to have highly
00:06:19
processed lying, very effective targeted
00:06:22
lying. If I send that to everybody, if I
00:06:26
told you u sincerely that Hillary
00:06:29
Clinton ran a child pornography
00:06:32
web out of a pizza parlor in Virginia
00:06:36
and had dozens of kids chained to the
00:06:39
wall,
00:06:40
>> you would either find it amusing because
00:06:44
you know it's false or you would find it
00:06:46
terrifying because you know there are
00:06:48
people who would believe it. Highly
00:06:51
processed news also goes to the people
00:06:53
who want to believe it.
00:06:54
>> So then how different then is news that
00:06:59
people see on a social media platform
00:07:02
from what we see on a network newscast
00:07:05
or if we read it in a newspaper
00:07:09
>> changing a little bit since the Bush
00:07:12
administration
00:07:13
basically got rid of the FCC's fairness
00:07:16
requirement.
00:07:17
But but still if the New York Times runs
00:07:21
a story, it has to be plausible to all
00:07:26
of its readers. If readers find it
00:07:29
preposterous,
00:07:30
it loses its credibility, which destroys
00:07:33
its value as a brand. Uh with the in the
00:07:37
case of social media,
00:07:39
uh the social media's brand is
00:07:41
entertainment.
00:07:43
And as long as I show stuff that people
00:07:46
find entertaining,
00:07:48
I'm protecting and enhancing the brand.
00:07:50
I'm not weakening it. So there's no
00:07:52
penalty
00:07:54
uh to Facebook or no penalty to Twitter
00:07:57
for highly focused, highly targeted
00:07:59
lying, whereas this would destroy the
00:08:02
brand of the New York Times or for that
00:08:05
matter NPR.
00:08:07
So most broadcast uh channels
00:08:11
attempt to have at least some element of
00:08:15
universal plausibility.
00:08:17
Even Fox News, which is about as biased
00:08:21
as a network can be, even Fox News on
00:08:25
their news programs, news segments,
00:08:28
attempts a certain amount of uh
00:08:31
reliability. Their editorials are a
00:08:34
different story. When I wanted to get
00:08:36
unprocessed news about the Middle East,
00:08:40
I actually got it from Al Jazer because
00:08:42
Al Al Jazer is so vulnerable to claims
00:08:46
of being false that their news is is as
00:08:50
reliable as the uh as the Financial
00:08:53
Times. Their editorials are a different
00:08:55
story. I don't spend a lot of time with
00:08:58
uh Alazer editorials, but their news has
00:09:01
to be accurate or they they have no
00:09:04
credit. they they lose whatever
00:09:06
credibility they have.
00:09:07
>> You make a distinction in the in the
00:09:09
paper that you wrote about this topic
00:09:12
between misinformation and
00:09:14
disinformation. Please take a moment to
00:09:16
explain the difference.
00:09:18
>> Sure. If I actually believed that um oh
00:09:24
that the election was stolen and I
00:09:26
repeat that or I actually believe that
00:09:29
uh vaccines cause autism
00:09:34
uh or that um
00:09:37
uh Joe Biden was fully competent even
00:09:41
going into the debate. If I truly
00:09:43
believe that
00:09:45
and I repeat that, I post that, I repost
00:09:48
that, that's misinformation.
00:09:51
>> Mhm.
00:09:51
>> If I tell stuff that I I believe for the
00:09:55
purpose of sharing my view of truth,
00:09:58
again, that's misinformation.
00:10:00
But if I know something to be false, I'm
00:10:04
actually working in a troll factory or
00:10:07
I'm working for a political campaign and
00:10:10
I design really effective arguments to
00:10:13
spread falsehood.
00:10:15
That's disinformation.
00:10:18
Misinformation is protected by the first
00:10:24
amendment.
00:10:25
Disinformation is not. So why then is
00:10:28
disinformation so hard to control?
00:10:31
>> A variety of reasons. One is
00:10:35
uh one man's um disinformation is
00:10:38
another man's uh political campaign.
00:10:42
Right. Uh we I've got to assume that
00:10:46
most people in the Trump White House
00:10:48
know that he lost the previous election
00:10:52
and yet they went after um companies for
00:10:57
v voting machine companies for fraud. Uh
00:11:00
they're fighting to block um mail-in
00:11:03
voting,
00:11:04
>> right? Be
00:11:05
>> because it's effective. So there are
00:11:08
there are times you do stuff I've I've
00:11:10
said this before. This is not a Democrat
00:11:13
or a Republican issue. It's a liberal or
00:11:17
conservative issue. Liberals by
00:11:19
definition consider almost everything
00:11:22
discussible. I don't mean inifat. I
00:11:25
don't mean u radical anarchists.
00:11:28
By definition, a political liberal in
00:11:31
the definition from the late 1700s
00:11:35
believes everything is discussable. Uh
00:11:37
for the truly committed, you're doing
00:11:39
God's work.
00:11:40
>> Right?
00:11:41
>> So if you're doing go the guy who said,
00:11:44
"I will I disagree with what you say,
00:11:46
but I will defend to the death your
00:11:48
right to say it." He wasn't part of the
00:11:50
Inquisition.
00:11:51
>> Right.
00:11:52
>> Right. I mean, he he wasn't doing God's
00:11:54
work. Well, and as you bring up, the
00:11:56
First Amendment obviously plays a very
00:11:58
important role in this process of
00:12:00
bringing a lot of this information to
00:12:03
the forefront to begin with.
00:12:05
>> It it does and the first amendment is a
00:12:08
very good thing. The first amendment is
00:12:11
based on the idea that truth wins in the
00:12:15
free market for ideas,
00:12:17
>> right?
00:12:17
>> Similarly, trial by jury isn't perfect.
00:12:21
The idea is both parties do their best.
00:12:24
They hire the best attorneys. They they
00:12:27
don't lie, but they manipulate
00:12:29
perception as best they can. And the
00:12:31
idea is in the end, the best idea wins.
00:12:35
Uh it doesn't work when technology
00:12:40
allows highly processed news. It doesn't
00:12:42
work when I can craft a story
00:12:45
specifically for you to believe and then
00:12:48
make sure you're the only one who sees
00:12:49
it. are are traditional
00:12:53
news outlets
00:12:55
that we have known for decades,
00:12:58
are they impacted negatively because of
00:13:02
what I think is the mindset around
00:13:04
bringing information forward, news
00:13:07
information forward on social media in
00:13:09
this day and age. I mean, this is
00:13:11
something that is really geared for
00:13:13
where we are right now.
00:13:15
The countries that have the strictest
00:13:17
rules on the use of information, uh,
00:13:20
nations in the EU would be the nations
00:13:23
that you would expect to be least
00:13:25
affected by highly processed news. And
00:13:29
four or five years ago, I met with the
00:13:31
head of strategy for a traditional
00:13:35
Danish broadcaster
00:13:37
and she showed that her market share in
00:13:41
almost all areas had been reduced by
00:13:43
about 75%.
00:13:46
And she said her stories have to be
00:13:49
true. They have to be the same for
00:13:51
everyone. As soon as she starts varying
00:13:54
what you see based on what you want to
00:13:56
believe, she loses her license to
00:13:59
broadcast.
00:14:00
And she pointed out that social media is
00:14:04
is much more entertaining. There's been
00:14:07
some research that shows that when we
00:14:11
watch TV, we want to be entertained,
00:14:13
>> right? And we view social media as
00:14:16
entertainment. Uh there's a work by um
00:14:22
uh two famous Nobel laurate behavioral
00:14:25
uh economists conne
00:14:28
uh conneeded out that we have type one
00:14:31
and type two thinking right
00:14:32
>> fast and slow fast is something's wrong
00:14:35
I have to get out of here slow is
00:14:38
complex and analytical he said we we
00:14:41
watch the the fake news researcher said
00:14:45
we watch fake news with fast thinking.
00:14:48
Is it entertaining? We don't say is it
00:14:51
correct? It's one of the reasons why
00:14:55
labeling demonstrably false news as
00:14:58
false and originating in Russia, China,
00:15:01
or North Korea has very little impact on
00:15:04
people's reposting it because it's fun.
00:15:08
>> Yeah.
00:15:09
>> So fun fun Trump's real media. I don't
00:15:13
want to use the word Trump in that
00:15:15
context. fun beats traditional media on
00:15:20
on reposting. It beats traditional media
00:15:23
on market share. It's just compelling.
00:15:25
It wins. But then is there I think a lot
00:15:28
of people would also say that um you
00:15:32
have to look at the importance of what
00:15:34
social media has become and that it's in
00:15:37
our culture and knowing that it's going
00:15:38
to be around our culture. But how do you
00:15:41
deal with the level of disinformation
00:15:44
that we have there
00:15:46
but still allow
00:15:49
social media platforms to be fun for
00:15:53
people who would like them to be fun?
00:15:54
you know, to kind of take the the
00:15:56
negative and the angst out of social
00:15:58
media.
00:16:00
>> I have an example of a social media
00:16:03
website that is so trivial that it was
00:16:07
able to police itself until it was
00:16:10
bought by Anheiser Bush. And it was a
00:16:14
craft beer rating site.
00:16:15
>> Okay.
00:16:16
>> And I could say, "Who's got the profile
00:16:19
most like mine?" And it turned out to be
00:16:22
a guy named Ratman.
00:16:24
Uh, and I would say, "What did Ratman
00:16:27
think of this beer?" And he would say,
00:16:29
"Even for me, too bitter." Well, all
00:16:32
right. I would, if he didn't like it,
00:16:33
it's certainly too much of an IPA for
00:16:35
me. [snorts] There was one beer
00:16:38
introduced by a brewery
00:16:40
and they obviously told their employees
00:16:42
to love it and suddenly there were
00:16:45
thousands of positive reviews for a
00:16:48
beer, all coming from the same location,
00:16:51
the location of the brewery.
00:16:52
>> Right. I I had it pulled because we
00:16:55
could we could police it. It really
00:16:57
didn't it didn't matter. Um how do you
00:17:00
police social media? Well, the first
00:17:03
thing you do uh is you prohibit the
00:17:08
social media from sharing aggregate
00:17:10
statistics. This is one of the reasons
00:17:12
why Tik Tok's claim to be safe is
00:17:15
patently false. Tik Tok doesn't have to
00:17:18
move any data on any individual out of
00:17:21
the US. It simply needs to categorize
00:17:24
individuals into buckets of what they
00:17:26
believe.
00:17:28
And that data doesn't have to leave the
00:17:30
states. Uh China can craft stories and
00:17:34
send them to Tik Tok. Tik Tok can post
00:17:37
them and direct them to the people who
00:17:41
want to read them. And none of this
00:17:43
violates privacy law.
00:17:44
>> Right? So if you want to um limit the
00:17:49
effectiveness of highly processed fake
00:17:52
news, you take away the processing
00:17:54
factory. You don't let Facebook share
00:17:57
data, even aggregate data on people's
00:18:01
beliefs. You don't let Facebook
00:18:04
uh post stories to individuals who are
00:18:08
responsive to them.
00:18:10
That's hard to do because aggregate data
00:18:13
doesn't violate privacy law, right? And
00:18:16
Facebook has successfully argued in
00:18:18
front of the EU that sending you a story
00:18:22
that you love is part of their
00:18:24
commitment to be a good entertainment
00:18:27
company. So it's it's not easy to do
00:18:29
this, but if you take away the raw
00:18:32
material, you shut down the processing
00:18:34
factory, then social media becomes just
00:18:37
a just a blackboard.
00:18:41
So then realistically, as we're moving
00:18:43
forward, you know, further down the road
00:18:45
here, the concept of disinformation is
00:18:48
something that we will probably see stay
00:18:50
around for the foreseeable future.
00:18:54
If you remember 1984,
00:18:58
which was
00:18:59
>> the book or the year?
00:19:00
>> The book.
00:19:01
>> Okay.
00:19:02
>> If I said there's this old joke, do you
00:19:04
remember 1968? And the answer is if you
00:19:07
were there, you don't remember it.
00:19:08
>> Right. Right. Right.
00:19:10
>> But but uh the book was originally
00:19:12
titled 1948 and the expectation was
00:19:16
technology would enable this degree of
00:19:19
societal manipulation very quickly.
00:19:22
>> Right?
00:19:23
>> It didn't. But it we're we're we're
00:19:25
getting there. So, social media will be
00:19:27
around forever. Lying has been a tool of
00:19:32
statecraft and a tool of Warcraft
00:19:35
uh forever. Absolutely forever. The
00:19:39
difference is between traditional oneoff
00:19:44
lying which is highly effective only
00:19:47
when a society is falling apart
00:19:50
and highly targeted lying which works
00:19:53
most of the time.
00:19:54
>> Right?
00:19:55
>> So we've we've got a society where there
00:19:58
is a strong motivation to win. There are
00:20:00
people who truly believe they're doing
00:20:02
God's work
00:20:04
>> and they have the tools to manipulate.
00:20:09
What we have to do if if we want to
00:20:11
limit it without limiting free speech is
00:20:15
deprive the the tools of the data needed
00:20:18
to process, create, and direct the news.
00:20:22
>> Eric, great conversation. Thanks very
00:20:24
much for your time and obviously
00:20:26
continued success with your work. All
00:20:27
the best. Thanks so much,
00:20:29
>> Eric Clemens, professor of operations,
00:20:32
information, [music] and decisions here
00:20:34
at the Wharton School. Thank you for
00:20:36
listening to The Ripple Effect. We hope
00:20:37
you found this episode informative and
00:20:39
engaging. Don't forget to subscribe and
00:20:41
[music] leave us a review so that we can
00:20:43
continue to bring you the best insight
00:20:45
from the Wharton School.
