Russia is by far the most advanced foreign adversary

sending manipulative messages over social media.

They're even more nuanced today than they were in 2016,

and we are no more prepared today than we were in 2016.

[dramatic music]

Hey, Wired, I'm Sinan Aral.

I'm the David Austin Professor of Management at MIT,

Director of MIT's Initiative on the Digital Economy,

and author of The Hype Machine,

about how social media disrupts our world.

I'm here to debunk some myths about the role

of social media in elections,

and specifically in the upcoming election of 2020.

[dramatic music]

Social media sways elections.

Did Russia sway the election of 2016?

We know they sent manipulative messages

to 126 million people on Facebook,

20 million on Instagram,

and 10 million Tweets.

Do these things sway the election?

There's really three things to know.

Does it change vote choice?

Does it change voter turnout?

And is the reach, scope, and targeting of misinformation

or campaign information enough to sway an election?

Voter choice is simply who you choose to vote for,

given that you're voting.

Do you vote for Republicans or Democrats?

That's a vote choice.

The evidence on vote choice is relatively clear.

Social media messaging and digital advertising

in general has a very small to negligible

to zero effect on vote choice.

Voter turnout on the other hand,

is do you choose to vote at all

and the number of people who vote in an election.

And there, the evidence is a little bit scarier

in the sense that large scale experiments have shown

that social media messages,

digital advertising can have

statistically significant effects on voter turnout.

Facebook ran an experiment

with 61 million people in it in 2010,

which showed that with a simple message,

they could create votes in congressional elections

that wouldn't have happened without their message.

They replicated that experiment in 2012

and demonstrated again the ability

for social media to create voter turnout.

Many studies indicate how digital messaging

can get out the vote and that's an important part

of changing or swaying elections.

Targeting is who you direct

digital social media messages to.

Which populations in which regions,

in which districts in the voting electorate.

The evidence in 2016 indicated

that Russian interference was targeted at swing states

and that the reach and scope of it

was large enough to affect voters in a way

that could change the election through voter turnout.

In addition,

we know that a lot of the manipulative messages sent

by Russia in 2016 were about voter suppression.

Voter suppression memes tend to be targeted

at specific communities.

So for instance,

we know that in 2016 on Instagram in particular,

African American voters were targeted

with voter suppression memes.

Indicating for example,

Hillary Clinton is not a fan of the Black voter

and therefore we should stay home

or there's really no one to vote for in this election.

There's no reason to vote.

Those types of memes were targeted through

at mentions at communities that were African American

or in this year's election,

follow the Black Lives Matter movement and so on,

trying to suppress specific communities

of voters in key swing states.

The number one culprit in spreading misinformation

and voter suppression memes in 2016

and likely in 2020 is Russia.

While Russian misinformation was scary in the 2016 election,

they are much more sophisticated today

than they were four years ago.

In addition,

this is happening during a global pandemic

with civil unrest in the streets arising

from the justifiable social movements

against police brutality in the United States.

With all of this uncertainty,

we are at a dramatic risk for foreign interference

in our election in 2020.

And of course,

social media is nowhere near

the only factor affecting elections.

Certainly the candidates, their charisma, their policies,

advertising their ability to connect with voters,

as well as news of the day.

What is hitting the pocketbooks and the homes and families

of every day voters obviously has the largest effect.

Social media spreads fake news faster than the truth.

That's true.

We did a 10-year longitudinal study

of all of the verified true and false news

that spread on Twitter between 2006 and 2017.

In fact we found that false news

was 70% more likely to be retweeted

and false news traveled about six times faster

than true news online.

Fake news is not a new term.

It was not invented by Donald Trump.

In fact,

it first appeared I believe

in a Harper's Magazine news story

and we've had the concept of falsity in journalism

for many years and decades prior to today.

The thing that makes today different however,

is the speed and breadth and depth

with which social media can spread fake news

so much faster than the truth online

and how that can be targeted at specific individuals

and communities creating separate realities

for people who are seeing one type of news in one community

and a different type of news in a different community.

So when we found these results in our Twitter data,

the natural next question for us was why.

Why does fake news spread so much farther,

faster, deeper, and more broadly than the truth?

What we came up with was

what we call the Novelty Hypothesis.

So if you read the cognitive science literature,

you know that human attention is drawn to novelty.

New things in the environment.

If you read the sociology literature,

you know that we gain in status

when we share novel information

because it makes us look like we're in the know

or that we have inside information

that other people don't have.

So these two factors make it more likely

that we share novelty.

So when we've checked the novelty of true

and false news compared to everything

that a given individual on Twitter had seen

in the two months prior,

we found that indeed false news

was way more novel than the truth,

and when we checked the replies to true and false Tweets

to see how people were expressing sentiment

about what they were reading,

we found that indeed in reply to false news,

people express surprise, anger, and disgust,

while in reply to true news

they expressed anticipation, joy, and trust.

So the surprise confirmed our novelty hypothesis that yes,

false news is more novel.

People spread more novel information more often

than less novel information,

and people were genuinely surprised by fake news.

Voting booths are not hackable.

False.

Voting booths can be hacked,

have been hacked,

and abnormal ballot results in a number

of examples over the last decade or so in the United States.

Many people believe that because the United States

has a federalist system in which each state tallies

its votes in its own way.

Uses different computer systems,

that there's no centralized tallying of ballots,

that this somehow protects the American voting system

from hacking at the voting booth itself.

But that's not true.

It just means that there are 50 different types

of systems that a hacker can attack.

So for instance, we know that in 2016,

a hacker named CyberZeist hacked the Alaska voting systems

and claimed that he could change the voting tallies

in any direction that he wanted in Alaska.

There are also a number of myths surrounding voting

that are spreading on social media.

The major one is that there is widespread voter fraud.

There is no real evidence of a systematic voter fraud

at the level of ballots or other types of voter fraud.

People voting twice, dead people voting, and so on.

Although there have been a very, very small handful

of incidents that may have happened

where there is an error on a ballot,

there has been no evidence of systematic voter fraud

since we can remember about elections in the United States.

Which means that despite all

of the myths floating around social media,

we as citizens can be confident

in the integrity of our elections.

So my advice to all of us is that we vote

and vote as quickly as possible ahead of November 3rd.

Social media algorithms are dividing our society.

There is evidence that recommendation algorithms

that social media uses does tend to give us more

of what we want and therefore lock us into narrower

and narrower sets of information.

Filter bubbles refers to the fact

that in an algorithmic world,

we are each living in our own information bubble.

Meaning that what I see on social media is not what you see

and not what your friends see,

because everything that you see is tailored to you.

And it's tailored to you by algorithms

that are designed to give you more

of what you want to keep you engaged.

That creates these filter bubbles of information

that are unique to every individual.

Echo chambers are groups or communities of people

that are sharing the same information over

and over again with each other,

and that that information stays locked in that community

and doesn't cross over for instance,

to the other side of the aisle

where different information is being constantly shared

amongst a different set of people.

So there are certain algorithms.

For instance, the YouTube algorithm

that tends to recommend more and more

of the type of content that you seem engaged with

and interested in.

Studies have shown that these types

of algorithms can tend to lead

to more extreme content being shown to the viewer.

These algorithms are designed to be bottomless or endless,

meaning they keep you engaged

in a constantly updating reel of new videos.

While the jury is out on whether this can radicalize someone

or the degree to which there

are systematic extremism outcomes

that are created by these algorithms,

the fact that they are sending you down rabbit holes

of more and more content similar to what you like

and engage with is troubling,

given the notion of the filter bubble.

In order to fight the filter bubble,

we have to seek out diverse content.

We have to follow people whose opinions

are different than our own.

We have to do searches for content

that is contrary to what we believe.

We have to demonstrate to the hype machine,

to the social media industrial complex,

that we're interested in diversity,

and that we are seeking diversity

or opinions that are different from our own.

That will help us break out of the filter bubbles

that we find ourselves in with these algorithms.

You can easily spot a deepfake.

Deepfakes are synthetic video that are generated

by machine learning algorithms called

generative adversarial networks.

These networks have a generator and a discriminator,

where the discriminator's job is to tell real

from fake videos and the generator tries

to generate more and more convincing synthetic video

'til it fools the discriminator

into believing that it's true.

Now, the problem with deepfakes is

that they're more difficult to spot every day that goes by.

There are instances of audio deepfakes,

where companies have been defrauded out

of millions of dollars.

Where the CFO will be called by a synthetic attacker

that is using the voice of the CEO requesting

that large sums of money be transferred

before the end of the quarter or to close a deal.

The reason why deepfakes are so troubling

is because seeing is believing

and a picture is worth a thousand words.

I have seen some incredibly professionally created

and convincing deepfakes for instance,

of President Barack Obama,

or Mark Zuckerberg,

or Prime Minister Boris Johnson,

or Kim Jong-Un that really kind of skate the line

between is this convincing or is this not?

As deepfakes become more commonplace

as the technology used to create them

becomes more democratized

and more people have access to it,

I think we're going to see a rise of wave of synthetic audio

and video that could become very dangerous

in a political environment or in a commercial environment,

either through fraud or through political manipulation.

I think the most effective way to spot a deepfake is

to distinguish the content of what's being said in the film.

If you can't imagine those words coming out of the mouth

of the person that you're viewing,

that is a good sign that this is a deepfake.

Social media can bring about positive change.

Most recently we've been focused on the potential disasters

that social media can create in our world,

but it's important not

to forget about the tremendous potential

for promise that social media can also bring.

We know for instance,

that when Nepal experienced the greatest earthquake

that it's seen in 100 years,

Facebook spun up a donate now button

and raised $15.5 million

from 770,000 people in over 100 countries,

which just shows you the mobilization potential

of this technology.

It's certainly played a catalyzing and accelerant role

in important social movements around the world,

like Black Lives Matter,

the Arab Spring, the Snow Revolution in Russia,

social mobilization in Japan and Hong Kong.

These kinds of social movements

can really be accelerated by social media.

Research at MIT and at Stanford shows

that Facebook creates $370 billion a year

in consumer surplus in the United States alone.

Imagine that for the entire world.

That's economic opportunity,

that's the ability to find jobs,

access to life saving health information,

and real human connection.

In some countries around the world,

Facebook is the internet.

It's the way that people conduct any number

of human activities,

from market transactions,

to running their businesses,

to staying in touch with their friends and family,

or finding out about where to vote or how to get healthcare.

These types of benefits are actually tremendous.

Social media is a very powerful tool

for creating such change in society.

The real question is what are we gonna use it for?

Are we gonna use it for the nefarious purposes

that we've seen it be used for recently

or are we gonna use it to bring about a better world?