Skip to main content
Search form
  • 00:05


  • 00:08

    PABLO BARBERA: Thank you.Thank you all for coming, and thank youfor such a kind invitation.So today, I'm going to be presenting and discussingsome of that research that I've done over the past few yearswith colleagues at NYU, University of SouthernCalifornia, and also at the London School of Economicswhere I'm currently based.So until recently, the way in which most individuals consume

  • 00:28

    PABLO BARBERA [continued]: political information-- the way in which most of us learn aboutwhat's happening in the world--looks something like this.So it would be at the end of the day,we would sit in front of the TV, we would watch the eveningnewscasts, or maybe we would read the newspaperslike magazines.We would talk with our friends and our familyabout what's happening in that world.And that's how we would consume political information.

  • 00:52

    PABLO BARBERA [continued]: But nowadays, that picture looks slightly different.It looks more like this.So we don't read print newspapers.We don't watch as much TV as before.We do a lot of things through the screens of our laptops,of our smartphones, of our computers.Digital technologies are revolutionizing

  • 01:13

    PABLO BARBERA [continued]: the way in which we consume political information.And that is having a very direct and identifiable impactalso in our behavior.So we've seen over the past few years the Arab Springand this wave of massive protest and demonstrations and rallieshappening all around the world.And social media, digital technologies,

  • 01:34

    PABLO BARBERA [continued]: probably had something to do with that.We're now seeing how politicians, public officials,parties, rely on digital technologiesto broadcast messages to large audiences.Now that I'm talking about this, Idon't think I need to justify this as much anymore with Trumpas president in the US.

  • 01:55

    PABLO BARBERA [continued]: But that's something that is really changing,and it has a massive impact in political accountability,communications strategies, election campaigns,and many other things.But today, I want to talk about one specific transformationthat has received a lot of attention,and that's the impact that social mediaand digital technologies and this transformation in how

  • 02:16

    PABLO BARBERA [continued]: we consume political information may be havingon political polarization.So there's this common idea--this conventional wisdom that maybe digital technologiesare creating cyberbalkanization, echo chambers, filter bubbles,is drawing us apart, is destroyingdemocratic societies.

  • 02:37

    PABLO BARBERA [continued]: So I want to talk about some of the researchthat we've done in that area.I'll start by presenting what I thinkis the standard conventional argument that a lot of peoplemake in this space.The idea is social media, digital technologies arecreating spaces where like-minded people can find one

  • 02:59

    PABLO BARBERA [continued]: another--these communities of people that share things in common.And this could be an interest in send pictures of the sunset,or an interest in publishing, or books, or things like that.But it could also be white nationalism, alt right,alt left, all these type of things that are notas benign for democracy.

  • 03:22

    PABLO BARBERA [continued]: It could be a variety of things.It could be a variety of processesthat explain why we end up in this situationwith these communities of like-minded individuals.It could be this pattern social behavior that'scalled "homophily," and that's like the propensityof individuals that share things in common to flocktogether like birds of a feather flocking together.It could be there's some kind of influence process, some kind

  • 03:42

    PABLO BARBERA [continued]: of like foreign interference creatingthese enclaves of people that sharean extreme political ideology.And the empirical evidence that is oftenused to justify this view looks something like this.So here on the left, we have a representationof blogs in the 2004 presidential electionin the US.So blogs are colored according to whether they

  • 04:05

    PABLO BARBERA [continued]: favor the Republican Party or the Democratic Party.Each of these nodes here-- each of these circles is a blog.And then you can see that the connections are hyperlinks.So pro-Republican blogs are only linkingto the pro-Republican on blogs, and vice versa for Democrats.Here on the right, we have a similar network diagram,but this time, it's Twitter data.

  • 04:26

    PABLO BARBERA [continued]: So these are Twitter users.This is the 2010 congressional election in the US.And here are the connections indicatingwho is re-tweeting one another.And you can see that if you're a Democrat,you're mostly re-tweeting other pro-Democrat people,and the same with Republican.So this leads to these communities of people

  • 04:46

    PABLO BARBERA [continued]: that just like spreading information among themselves,and they themselves happen to sharethe same political ideology.Why is this a problem?Well, the problem is, according to this standard argument,we're now only exposed to information thatconfirms our political views.So if you're a Democrat, you only

  • 05:08

    PABLO BARBERA [continued]: see information that is favorableto the Democratic candidate, for example, if it's an election.Or you might be only exposed to one side of a particular issueaccording to your pre-existing opinion on that issue.There could be some specific featuresof the platforms that are contributing to this change.So all these ranking algorithms that try

  • 05:30

    PABLO BARBERA [continued]: to predict what you're going to like and then try to show youwhat you're going to like.And this is what Eli [inaudible] and others have--they coined the term "filter bubble,"but we'll see in a moment that, actually, it's like this term--similar ideas are present in the literature before.And the outcome of all of this--and this is where we end up, the end of the argument--

  • 05:51

    PABLO BARBERA [continued]: is this is going to increase political polarization.This is eliminating the possibilityof finding common ground between people that think differently.We're in a situation, according to this argument,in which we're never exposed to something that may challengeour views-- something that may actuallymake us change our mind about a particular issue.Everything we see aligns with our pre-existing political

  • 06:13

    PABLO BARBERA [continued]: ideas.And again, I want to emphasize, thisseems to be a very new argument that now we see in the mediaall the time, but it's not that new.One of my favorite books of all time,Bowling Alone by Robert Putnam.This classic in sociology, political science,about the decline of social capital in the US.And really it's very prescient.

  • 06:35

    PABLO BARBERA [continued]: I was re-reading it preparing for this talk,and it just surprise to realize this is the year 2000,and a lot of the arguments that we're still making todayare already in this book.So Robert Putnam was saying, real world interactions oftenforce us to deal with diversity, whereas the virtual world maybemore homogeneous, not in demographic terms,but in terms of interest and outlook.

  • 06:56

    PABLO BARBERA [continued]: Internet technologies allows and encourages white supremaciststo narrow their circle to like minded intimates,and new filtering technologies thatautomate the screening of irrelevant messagesmake the problem worse.So the year 2000-- and we really see this argument.And again, this is an old argument, but in recent years,it has become widespread.

  • 07:17

    PABLO BARBERA [continued]: Everybody is making this argument includingPresident Barack Obama himself.So in this interview in Netflix, he said,if you're getting all your information of algorithms beingsent through phone, and it's justreinforcing whatever biases you have,which is the pattern that develops at a certain point,you just live in a bubble.So the filter bubble idea pops up here again,

  • 07:38

    PABLO BARBERA [continued]: and that's part of why our politics is so polarized rightnow.I think it's a solvable problem, but Ithink it's one that we have to spenda lot of time thinking about.A very popular argument.And today, I want to say, also wrong.When we look at the empirical evidence--and hopefully by the end of this talk,

  • 07:58

    PABLO BARBERA [continued]: I will convince you that actually there'snot a lot of empirical evidence suggesting this is true.So the argument that I'm going to try to make,which is based on the research that I've doneand also empirical evidence from other existing studies,is the following.So the echo chambers are actually empty.So while it might be true that most

  • 08:22

    PABLO BARBERA [continued]: of the political interactions that take place on social mediasites are exchanges between peoplethat share political inclinations,cross-scanning interactions are actually more frequentthan commonly believed.So interactions between people on the left and on the rightand between the right and the leftare actually way more frequent than what

  • 08:43

    PABLO BARBERA [continued]: you could expect based on these earlier studiesof this conventional wisdom.We're actually much more exposed to diverse ideason social media than through the consumption of newspapers,watching TV, offline interactions with people.So if anything, social media's increasing the extentto which we see new ideas.

  • 09:05

    PABLO BARBERA [continued]: And finally, contrary to this idea that these algorithms arecreating this filter bubbles, with all the empirical evidencesuggests that it's is not so bad.Sure, there's some filtering process taking place,but there's other factors that aremore to blame for any potential cyberbalkanizationother than these algorithms.And we'll come back to that in a second.

  • 09:27

    PABLO BARBERA [continued]: Now, that doesn't mean that social media may notbe polarizing us.So social media could still be polarizing, but probablyin ways that are not the ways most people think.So while it may be true that we're being more exposedto information from the other side,

  • 09:47

    PABLO BARBERA [continued]: maybe there's other types of polarizationthat are increasing as a result of being on social mediaor like digital technologies more generally.And I will talk about the conceptof effective polarization, which is notso much about our ideas are becoming more extreme,it's just we tend to dislike the other side more than before.

  • 10:08

    PABLO BARBERA [continued]: So we're not getting apart from one another.It's just we don't want to hang outwith the ones from the other ideology anymore.It is true there might be a feedbackloop also in terms of thinking about misinformation,incivility.So maybe we see the messages from the other side,but we see them with a negative connotation.And that might make us more polarized

  • 10:30

    PABLO BARBERA [continued]: and from this effective definition of the term.And finally, even if for most peoplesocial media activities may not be terribly polarizing,it could still be the case for some minority of people.And we found evidence of that, for some minorities,social media activities-- consumptionof political information might stillbe having polarizing effects.

  • 10:52

    PABLO BARBERA [continued]: So I will show you evidence for all these,but I wanted to give you first in a nutshell where we're goingover the next 30 minutes or so.So before I show you some of the evidence over here,I want to first introduce some new methodthat I've been working on over the past few years thatis necessary in order to understand everything that'sgoing to come later.

  • 11:12

    PABLO BARBERA [continued]: So one problem if we want to understand consumptionof political information through social media,first we need to measure the political ideologyof social media people, like people on social media.And here, social media-- due to data availability,it will be mostly Twitter data, but Ithink most of what I will tell you todayalso applies to Facebook and other social media platforms.

  • 11:34

    PABLO BARBERA [continued]: So we need to measure political ideology in orderto, for example, map the flows of informationin the Twittersphere.So how do we do that?So I'll show you with one exampleto give you the intuition of how it works, and then I'llexplain why this works.So let's say we have this one Twitteruser, whose name is Brian Patrik, has been a fake user,

  • 11:57

    PABLO BARBERA [continued]: but the data is real.And this is his network.So this is what called an eagle network.And here, each of the circles that we seeis someone that Ryan follows on Twitter.And then the connections are whether those peoplefollow one another in turn.So we see very tightly interconnected networks.This is quite common when we go to social media networksbecause most of the people that we follow

  • 12:19

    PABLO BARBERA [continued]: are people that we know in the real world,so they're probably going to be followingone another as well because they also know one another.So from this network, we're going to zero-inon political accounts.So we're going to be looking at the political accountsthat this particular user follows on Twitter.Political accounts, meaning membersof Congress, other politicians, media outlets, interest groups

  • 12:44

    PABLO BARBERA [continued]: and think tanks, these type of things.So in this case, this particular user is from Ohio in the US,follows Barack Obama, Rachel Maddow, Senator Rob Portmanfrom Ohio who's a Republican.You cannot really guess but just based on the accounts that hefollows, he's probably going to be leaning left.And this is something that I'll show you

  • 13:05

    PABLO BARBERA [continued]: in a moment empirical evidence that this actually worksin practice, but turns out, just lookingat who you follow on Twitter, the political accountsand media that you consume, that's actuallyhighly predictive of your political ideology.We can measure with a very high degree of accuracywhether you lean left or right and to whatextent just by looking at that.

  • 13:26

    PABLO BARBERA [continued]: So we start with this, and then we'regoing to have a list of different political accountsthat you could follow which is marked with a one or a zerowhether you follow each other's accounts.And then, this is one user, but we can dothis for any user on Twitter.So we can build what in the network science literatureis called an adjacency matrix.

  • 13:46

    PABLO BARBERA [continued]: It's a matrix that indicates whether the user is connectedor not to all these political accounts.And then we take these metrics and weapply some dimensionality reduction techniques.So we take these metrics and we tryto scale those users and the political accountson a latent dimension, and there is

  • 14:06

    PABLO BARBERA [continued]: going to be ideology to compute the number that we'regoing to use to quantify your ideology.So this is a number there's goingto be approximately minus 3, 2 plus 3.Low numbers correspond to someone who is on the left.High numbers means that someone is on the right.This person we estimate that the ideologies minus 1.You can read more if you're interested

  • 14:27

    PABLO BARBERA [continued]: about the mathematical details.You can read more about this in the papersthat we have published.Does this work?It terms out it actually works.You can measure, as I mentioned earlier,with very high accuracy whether someoneis liberal or conservative in the US.So how do we evaluate that?So we match Twitter accounts in the US with publicly available

  • 14:51

    PABLO BARBERA [continued]: voting registration records.So this is something that might be shocking to usas Europeans, particularly in the GDPR world.But in the US, in many states, youcan get a list of everybody who is registered to vote,and also you can tell if someone is registeredas a Republican or a Democrat.Why?Because they need to have that information so that

  • 15:11

    PABLO BARBERA [continued]: you can vote in the primary elections.And in some states, that's also available for research,and it's totally fair game to use it.Again, quite shocking when I discover that.I'm like, yay, data.But then, ouch, data privacy.So what we did is we looked on Twitterwhether there was someone tweeting

  • 15:32

    PABLO BARBERA [continued]: from a specific location with a given name.Let's say, Pablo Barbera tweeting from central London,and then we would go to their voter fileand see is there any Pablo Barbera thatlives in central London.If there is, we would say, OK, we match it.And then for those people, we computedtheir political ideology using this method that I described,and then we looked at whether those people were registered

  • 15:52

    PABLO BARBERA [continued]: Democrat or Republican.Turns out, just looking at who they follow on Twitter,we can predicts that information with 83% accuracy,which is quite surprising.So be careful who you follow on Twitteror if you don't want people like me to bepredicting things about you.This is something that is useful,and I'll come back to the polarization in a moment.

  • 16:12

    PABLO BARBERA [continued]: But this is something that you use fornot only for this, but also for other applications.So one example of that is, in this blog postthat I wrote for the Washington Post a couple of yearsago where I was interested in combiningthe political ideology of different candidates runningin the 2016 presidential election.So we can look at their Twitter account,

  • 16:33

    PABLO BARBERA [continued]: and we use this method to measuretheir political leanings, and we cananswer questions like, who is the mostconservative Republican candidate thatis running for president?Ted Cruz?And on the left, the most left wing,no surprise here, Bernie Sanders.But what's interesting is that wecan compare these estimates of ideologyfor different types of people.

  • 16:54

    PABLO BARBERA [continued]: We can look at media outlets.We can look-- so here, this density in red,and in blue, those are members of Congress.The gray area is just like a sample of Twitter users.So we can say things like, oh, Bernie Sandersis on the fifth percentile of the distribution of ideologyin the US, which maybe explains why he didn't get elected.

  • 17:15

    PABLO BARBERA [continued]: There were not a lot of people here on the left.Although, of course, like democratic primary electionsis a different thing.But anyway, there's just a side note on stuffthat you can do with these techniques that we develop.But going back to political polarization,so the first thing that I want to show you

  • 17:35

    PABLO BARBERA [continued]: is, is it the case that informationstays within the echo chamber?Is it the case that it doesn't travel far and wideto different parts of the network?And the way we looked at this was,we tried to recreate some of those network visualizationsthat I show you in the beginning that appear to have these veryhigh levels of segregation.

  • 17:56

    PABLO BARBERA [continued]: But this time, we did it with our new measure of ideologythat can be applied to a much larger sample of people.So it's not only blogs.It's not only in the case of Twitter users like peopleusing a very partisan hashtag.Now we can do this with almost everybody on Twitter.And once you look beyond the most partisan people,it turns out that the picture looks slightly different.

  • 18:16

    PABLO BARBERA [continued]: So here on the left, we have, again, similar visualizationof the network of people tweeting about the 2012election on Twitter.And these circles-- and they're like toosmall to see because there's a lot of people.There's like 200,000 users in this plot.The circles I recall are based on their political ideology

  • 18:36

    PABLO BARBERA [continued]: with this method that we develop.And then the connections indicatewhether they are re-tweeting one another or not.So what do we find?So in this case of these very partisan topic,we do find relatively high levels of segregation.So it's not surprising.People who are on the left, they tendto re-tweet mostly people on the left, and vice versa.

  • 18:58

    PABLO BARBERA [continued]: But we also see very high levels of inter-group exchanges.So we see a lot of people on the left re-tweeting peopleon the right, and people on the right re-tweeting peopleon the left.What this means, in practice, is that the information was notstaying within the enclaves on the left or on that right.

  • 19:18

    PABLO BARBERA [continued]: It was flowing through the entire network.So if someone was tweeting something with a liberal slant,that message would not stay only within liberals.It had the potential to reach everybody on Twitter.And this is a very partisan topic,but what happens when we look at other things thatare not partisan at all.So let's say discussion of like a major sports event

  • 19:39

    PABLO BARBERA [continued]: like the Super Bowl in 2013.In those cases, we don't see this segregated pattern at all.Then you might be thinking, well,why would we expect to see segregationwhen it comes to sports?Well, if you think about this, the main argumentthat we're all constantly not interactinganymore because of social media, then youwould assume, well, that should apply

  • 19:59

    PABLO BARBERA [continued]: to all types of discussions.And we don't see that in other sports events.And in fact, this is, in a way, the extremeof the distribution.Of course, this is one of the most partisan discussionsthat you could have on Twitter.But even in this extreme, again, as I mentioned earlier,we do see that the information does notstay within these communities of like-minded individuals.

  • 20:21

    PABLO BARBERA [continued]: We looked not only at these two examples,but we looked at the variety of different topicsor different issues that people couldbe discussing on social media.And we came up with a new way of visualizing these networks.Because one of the problems that we were having is,how do you put hundreds of thousands of dotsin just one single graph.

  • 20:43

    PABLO BARBERA [continued]: So what we did is, OK, let's try to come up with these heat mapsthat show you the micro structure of conversationson Twitter and how information spreads across peopleof different ideology.So let me walk you through what we found here,and then I'll explain our main take-away point.So here, on the x-axis, we have the ideology

  • 21:05

    PABLO BARBERA [continued]: of who is creating the information.So here's the person writing the messagefor the first time on Twitter.And then on the y-axis, we have the ideology of whoeveris spreading that message.So again, the author, and then, here'slike the ideology of [inaudible] Twitter.And the shade of each of these cells

  • 21:26

    PABLO BARBERA [continued]: indicates how many tweets are in each of those cells.So this is these two-dimensional representationof this network diagram.And again, what we find is all these peopleon the left re-tweeting people on the left.People on the right re-tweeting people on the right.So again, we see there's more.

  • 21:46

    PABLO BARBERA [continued]: If you see in the published papersthere's a little bit more shades of gray.Not like the book, but, yeah.But there's a little bit more here.And something that was interesting--I'll come back to this in a second--is that we did find systematicallythat there was a little bit more of segregation on the rightthan on the left.I'll come back to this in a second.

  • 22:07

    PABLO BARBERA [continued]: But anyway, so if we were to find that information is onlystaying within people of the same ideology,this would look like this or like a 45 degree line.Because it would mean only very liberal peopleare only re-tweeting other liberals, et cetera.But when we look at these range of topics,we see examples of information that is created by liberals.

  • 22:28

    PABLO BARBERA [continued]: For example, these are tweets about the minimum wagein the US, which was a major issue.It still is today a major political issue in the US.You see a lot of information from liberalsthat makes it all the way to moderate conservatives.Or here, State of the Union, theseare going to be a lot of tweets by Barack Obama.This 2014, 2015.

  • 22:50

    PABLO BARBERA [continued]: Those were the days.And you see how it's like people from all types of ideologyre-tweeting those messages.Again, these are very partisan discussions,but when we look at other things that are notas partisan like the election or the new minimum wage,we do see that this pattern completely disappears.

  • 23:10

    PABLO BARBERA [continued]: People are interacting and spreadinginformation regardless of what their political views are.And we see things like the Olympics or thisis like a terrorist attack that happenedat the end of the Boston Marathon in 2013.And we do see all this informationspreading throughout the entire network

  • 23:31

    PABLO BARBERA [continued]: without ideological similarity playing a role in that.We also found something that was interesting,which is there was some dynamic patterns in howthese structure of the information diffusionevolves over time.So it's not that these topics are static and fixed,but there's some changes over time.

  • 23:53

    PABLO BARBERA [continued]: So something that we found is that, looking at these Newtownshooting in this elementary school in Connecticut,and this is the days since we started collecting data.And in that case, we started collecting datawhen it happened.And initially, it was not a polarized topic at all.But it was after 15, 20 days, it became polarized

  • 24:13

    PABLO BARBERA [continued]: because it was like framed in terms of gun control versus gunrights.So you can see how some topics initially might notbe as polarizing, but eventually, we do seesome polarization happening.But overall, again, a lot of these sports events,stuff like that, we don't see any polarization at all.And the last small finding that I want to show you here

  • 24:36

    PABLO BARBERA [continued]: that I think is important when we want to understand someof the patterns that we're seeing today on social mediais this potential asymmetry between people on the leftand people on the right.So something that we found is, systematically,people on the right, so conservatives,were less likely to engage with peopleof the other ideology than the other way around, at least

  • 24:58

    PABLO BARBERA [continued]: in the US.So the way we measure this is, webuild a model that predicts whether each person ismore or less likely to re-tweet someone from the other side.And what we find is that after controllingfor a series of factors, people who are conservative--again, this is a US data.It might be different in other countries.

  • 25:19

    PABLO BARBERA [continued]: Conservatives were less likely to engagein this inter-ideological dissemination of informationthan the other way around.So just to give you an idea, so this is our estimated rateof cross-ideological re-tweeting, which means--so for example, 0.25 means for every re-tweet thatis across ideologies, there are fourthat are within ideologies.

  • 25:40

    PABLO BARBERA [continued]: And we see, systematically, peoplewho are liberals on the left in the USare more likely to retreat across ideological linesthan the other way around.And this is consistent with some theories and empirical evidencein political psychology that showsthat, again, at least in the US, liberals exhibithigher degrees of openness to change whereas conservatives

  • 26:01

    PABLO BARBERA [continued]: exhibit higher degrees of system justificationand some of these other personality traitsthat people have done research on in political psychology.So what is the conclusion of all of this?That at least, when it comes to the diffusionof political information on social media,it is not the case that the informationstays within the echo chamber.

  • 26:22

    PABLO BARBERA [continued]: There is no such a thing as an echo chamber in that sense.But one limitation of all of this is the following.So we've only looked so far at how information spreads,but you could be re-tweeting stuff because you disagreewith it, or you could be re-tweeting something[inaudible] do you actually see it?Are you actually exposed to it?

  • 26:44

    PABLO BARBERA [continued]: So what we tried to do next is like,OK, let's try to see if it is the case that people are onlyexposed to that information that confirms their prior belief.So it's not only about information diffusion anymore.Now, it's like what you see on social media.It's a bit tough to measure that,but this is how we try to do it.So let's go back to Ryan Patrick.

  • 27:06

    PABLO BARBERA [continued]: So we already use the political accountsthat he follows on Twitter to measure his political ideology,but now, let's look at everybody else.And a lot of these people are goingto be his friends, family, high school colleagues, co-workers,et cetera.And again, we can predict political ideologywith very high accuracy for almost everybody on Twitter.

  • 27:28

    PABLO BARBERA [continued]: So then, we can take that network and justcount how many people are of the opposite ideology.So Ryan Patrick is on the left, and then welook at the network.Not surprisingly, we see a majorityof people who are also on the left,but we also see 32% of people that hefollows other than these political accounts that

  • 27:50

    PABLO BARBERA [continued]: are conservative.So that's the number that we want to measure.So it's not exactly perfect because we don't know exactlywhat people see, but we can make the assumption that,if 32% of the people that you follow of your social ties areof the opposite ideology, probably your degreeof cross-cutting exposure-- so the extent to which you see

  • 28:11

    PABLO BARBERA [continued]: information that does not align with your political ideology--is going to be around that number.So we built this index.We call it the index of potential exposureto disagreement, to be like very precise and very academicallyto make sure that this goes through peer review.So we count-- so here we have--this is for the number of users in the network of a given user

  • 28:35

    PABLO BARBERA [continued]: I that are conservative or liberal.So we just count the proportion.It's actually a very simple proportion.So this is just the proportion of individualsin a user's network that disagreewith their ideological positions.So if the user is conservative-- of allof the individuals for which we have an ideological estimate,how many of those are liberal?

  • 28:56

    PABLO BARBERA [continued]: And again, we have that number for one user,but we can look at a large sample of people on Twitterto measure the extent to which people are indeedexposed to diversity of views.And this is the histogram of whatthat distribution looks like.So let me walk you through this.So here is just this index of potential exposure

  • 29:16

    PABLO BARBERA [continued]: to disagreement.So 0% here would mean you never seeanything that challenge you.Everything you potentially see is somethingthat aligns with your views.And this would be the world of perfect homopholy.You're only connected of other people thatare exactly the same as you.50% is the other extreme.So just like following people at random.

  • 29:39

    PABLO BARBERA [continued]: Your network is like 50-50 liberal conservative.And we cannot really see that most of the mass of thesehistograms is to the left of this.So what does this mean?It means that for most people, a majority of their network has--[inaudible] most of them share the same ideologyas themselves.

  • 29:59

    PABLO BARBERA [continued]: So this is not surprising.It's not specific to social media.If you think about your circle of friends, of your family,for most people, most of your friends, most of your familywill have very similar political ideas as yourself.Again, not specific to social media.It's just like a pattern in social lifein our social networks.And the median of this distribution is here--

  • 30:22

    PABLO BARBERA [continued]: so 32%.So this is actually quite similar to this examplethat I was showing you earlier.So what this means is that, for the average person on Twitter,32% of what they see potentially actuallychallenges their political views.It's something that is coming from the other side.So again, way above this idea of, oh, nothing that we see

  • 30:47

    PABLO BARBERA [continued]: is against our views.So like not at all a situation of the echo chamber.It is indeed the case-- and I'll come back to this in a second.For some people, they are indeed in an echo chamber,but it's a minority.There's very few people here.Most people are here.They still see some degree of diversity on social media.

  • 31:08

    PABLO BARBERA [continued]: And this is Twitter, but we also have some evidencefrom Facebook data.So there's a paper that was published in Science 2015.Actually, they tried to measure exactly the same thing that wewere trying to measure here.They had this index of cross-cutting contentin the percent of scanning content that they see.And I will get through this because I think it's actuallyquite interesting too.

  • 31:28

    PABLO BARBERA [continued]: The comparisons are like remarkably similarwhat they found on Facebook with what we found on Twitter.So this is if people were just like following other peopleat random.So again, kind of similar to the 50% that I had earlier.This was their index potential from the network.So this is just for people on Facebook.

  • 31:49

    PABLO BARBERA [continued]: If you just count what their friends look like,the distribution, that's what they see.And this is the number that we had in our paper.Because we cannot measure exposure or what they clickon, we can just see the potential exposureto cross-cutting content from their network.And here, we have what they actually see.

  • 32:11

    PABLO BARBERA [continued]: So the difference between here and hereis the effect of the algorithm.And you can see that the effect of the algorithmis not that large.It reduces a little bit the diversity of the contentthat people see, but not that much.The biggest selection process is from everybodythat people could be friends with to who they actuallybecome friends with on Facebook.

  • 32:32

    PABLO BARBERA [continued]: And this is finally what they actually click on.So again, you can see that far from that 0%--and again, maybe it's a straw-man argument,but there's 0% in the echo chamber.Between 25% and 30% of what people see on Facebookis actually cross-cutting is somethingthat might be different from what they actually think.

  • 32:53

    PABLO BARBERA [continued]: And again, this is some of the research that we've done,some research from Facebook.There's a growing body of evidence in political scienceand other fields--you want to take a picture?A growing body of evidence in political scienceand other fields showing that-- so for example,there's a paper in Public Opinion Quarterly, again,using search data in the US finding that social media users

  • 33:16

    PABLO BARBERA [continued]: have higher levels of cross-cutting exposurethan those who are just visiting political websites directly.So this is comparing internet with social media.There's some research using survey data thatcompares face-to-face interaction,consumption of newspapers, and TV with social media.And again, there's more political disagreementon social media than in other places.

  • 33:38

    PABLO BARBERA [continued]: And we don't have a lot of data outside of the US,which is rather unfortunate.But the little data that we do have--and this is some survey data that peopleat Oxford University, the ReutersInstitute of Journalism, which they do a fantasticwork developing this comparative and longitudinal measuresof consumption of news.

  • 33:58

    PABLO BARBERA [continued]: They also found that, when you look at the diversity of whatpeople see in terms of the news, the newscoming from social media is actually the most diverse.So hopefully by now, all these overwhelming body of evidencehas convinced you that, at least,it has made you doubt that we are

  • 34:19

    PABLO BARBERA [continued]: in echo chambers on social media.Again, for most people, we're not in micro chambers.But there might be things that could go wrong,and that's also what I want to talk about.So ways in which social media couldbe polarizing or depolarizing.So now we go from what you see on social mediato how that may change your political attitudes.

  • 34:40

    PABLO BARBERA [continued]: So that social media usage increasepolitical polarization.So let's try to answer that question.So to me, the most important changein the consumption of social media comparedto other types of consumption of news is the following.So we can think of this in terms of the social ties that are

  • 35:03

    PABLO BARBERA [continued]: sharing information with us.We could think about people that wefeel very close to like our close familymembers, close friends, and peoplethat we feel less close to like our acquaintances high schoolfriends, co-workers.And we tend to find when you ask, how often do youdisagree with these types of peopleis that, you tend to disagree more often with your co-workers

  • 35:25

    PABLO BARBERA [continued]: than with your family.There could be different explanationsfor why this happens because with your family, you disagree.Well, you might have some trouble with co-workers.You can just avoid politics at all.So more disagreement with our acquaintances or there'smore differences than with family members.But this is all fine because we regularly

  • 35:47

    PABLO BARBERA [continued]: talk more with our family than with our acquaintancesabout politics.So it's fine.So most of the time, we are discussing politicswith people with whom we agree.So this is all the offline world.I mean, the weird thing about thisis what social media is changing is this graph over here.So social media is increasing the extent

  • 36:10

    PABLO BARBERA [continued]: to which we discuss politics with our acquaintances.So these are what we would call that weak ties.So weak ties, there's a lot of research in sociologythat's showing that they help spread novel information.For example, the classic example is,we tend to find a job through these weak tiesbecause these are more like extended networks,and they're connected to different parts

  • 36:31

    PABLO BARBERA [continued]: of the network, et cetera.And what social media changes is, nowwe see all these messages from our crazy high school friends,crazy relatives like in some rural area, and all that stuff,that we didn't see before.So it doesn't change how we communicate with our familybecause you can still talk to them offline.You don't see more stuff from your very close family,

  • 36:53

    PABLO BARBERA [continued]: very close friends, because you already discussed politicswith them offline.What changes is that it used to be only once per Christmasduring the Christmas dinner that youhave these awkward conversations about politics.Now, you can have them all-year longon Facebook and social media.And so what happens with that?So now we see all these ideas that are new to us.

  • 37:17

    PABLO BARBERA [continued]: And to me, that's the key of all of this.So how would we expect that to affect our levelsof political polarization?So there's a lot of research in the social sciences about whathappens when you're exposed to a new ideas.So something that could happen is,this could be some kind of cognitive mechanism.So you see new ideas that you didn't think of before,

  • 37:39

    PABLO BARBERA [continued]: and maybe you just converge towards those ideas.If you're like some who's very liberaland you never thought about some conservative ideasthat you think makes sense, maybe once you hear themfor the first time, you're like, oh,wow, maybe that makes sense.Maybe I'm not as liberal as I thought I was.And there's a lot of research on this.Another way to think about this isthat it could be some kind of effective connection.

  • 38:01

    PABLO BARBERA [continued]: So this is, again, research on political tolerance and allthat stuff.So we train to think that someone whois in the out group that is different from us,their ideas are not legitimate.But now that we see those ideas, and we see themin a context in which social cues are emphasized,maybe we're like, oh, well, maybe conservatives or maybe

  • 38:21

    PABLO BARBERA [continued]: liberals are not as bad.Maybe there's something about what they'resaying that I should listen to.But the opposite could also be true.So there could also be all this boomerang and backlash effects.So we can think about in two different ways.So one could be, you see these new ideas,and you're like, no, no, no, no, no.I'm absolutely right.

  • 38:42

    PABLO BARBERA [continued]: And then you become more convincedthat what you thought before is the truth.And the other side of this is that, you could see those ideasin a context in which they are presented in a negative way.So on social media, there's a lotof emphasis on confrontation.We know negative stuff on social mediatends to be re-tweeted more often.Ironic, sarcastic content, again, tends to spread,

  • 39:06

    PABLO BARBERA [continued]: become more viral.So maybe if you see these new ideas, but people are saying,oh, this makes no sense, then you might actuallybecome more like a staunch believer in your actual ideas.So what does the evidence show?And I'm sorry because here, I don't havea very clear answer for you.Spoiler alert.So the research that I've done, personally,

  • 39:27

    PABLO BARBERA [continued]: and that we've published shows that most people,when they see new ideas on social media,that they actually become a little bit morepolitically moderate.So the way we tested this is, we used that methodto measure political ideology that I introduced earlier.And we measure ideology at two pointsin time, so 2013 and 2014.Almost like two years apart.

  • 39:48

    PABLO BARBERA [continued]: And then we looked at the extent to whichindividuals were exposed to disagreement on Twitterand what happened.Did they become more moderate or did they become more extreme?And here on the y-axis, we have the predictionfrom the statistical model that webuilt in terms of whether people became more politicallymoderate or more extreme.

  • 40:10

    PABLO BARBERA [continued]: And for the average Twitter user,we predicted that that person becamemore politically moderate.So being exposed to new ideas on Twittermade people converge towards the center.So it's not the case for everybody.Twitter was a moderating force.For some people, a minority of people, but for some people,

  • 40:31

    PABLO BARBERA [continued]: it actually had a polarizing effect.But for most people, they were becoming more moderatebecause they were exposed to different new political ideason Twitter.So if we take this line and we put it on top of that histogramthat I showed you earlier, our prediction was for close to 90%of people on Twitter in our sample

  • 40:53

    PABLO BARBERA [continued]: that were becoming more politicallymoderate because they were exposed to cross-cutting ideason social media.Again, some people were becoming more extreme,but it was just a minority.This is consistent with some other researchthat is being done right now.So there's a paper that was publishedin the Proceedings of the National Academy of Scienceslast year by Boxell and Gentzkow and some others

  • 41:16

    PABLO BARBERA [continued]: that looked at trends in political polarizationacross different age groups.And what they found is that, sure, polarizationis increasing in the US right now,but it's increasing less among young people than older people.And young people, we know they'remuch more likely to be online and on social media than olderpeople like this when they looked at this.So it cannot be the case that social media or internet

  • 41:40

    PABLO BARBERA [continued]: in general is the main force driving up polarization.And they have a more complicated analysis that, again,was reinforcing this idea.So this is consistent with the research that we've done,but there's also another paper that just came out, again,also a few months ago that is showing the complete opposite.So Chris Bail from Duke University-- they

  • 42:00

    PABLO BARBERA [continued]: ran some experiments on Twitter as well with some bots.So they targeted people with messages that whereof the opposite.For some people that were on the left,they were targeting them with messagesthat had some conservative slant, and the other way aroundfor conservatives.And they found that actually peoplesaw evidence of this backlash effect of people becoming

  • 42:22

    PABLO BARBERA [continued]: more extreme as a consequence of beingexposed to these messages.One limitation of this study is that the messageswere messages from political elites,from politicians, and things like that, whichis different from what we did, whichwas looking at messages that are coming from your social ties.So it could be that depending on the source of those messages,some might be more polarizing than others.

  • 42:44

    PABLO BARBERA [continued]: Some might be depolarizing, and some might be more polarizing.Another way to think about this--and I'm almost at the end--is that maybe what social media is changingis not so much ideological polarization,which is what we measure the extremity of our views,but maybe it's like how we think about peoplethat think differently from us.And this is what some researchers

  • 43:06

    PABLO BARBERA [continued]: in American politics or Shanto Iyengar and coauthors, they'vecalled affective polarization.So the way they measure this is, this feeling thermometer of howyou feel about-- so if you're a Democrat,how do you feel about Republicans?And they also have these very interesting survey question,which is, how would you feel if your daughter or son wouldmarry someone of the opposite party?

  • 43:28

    PABLO BARBERA [continued]: And what they found is, right now in the US,there's more support for interracial marriagethan inter-partisan marriage which is mind blowing.And here you can see-- so this is all of the feelingthermometers for within party, and thisis from one party to the other, and over the last 20, 30 years,there's been this steady decline in this out party sentiment.

  • 43:51

    PABLO BARBERA [continued]: And 2016 is not here, but 2016 is actually the absolute lowestpoint in [inaudible] series.So if anything, maybe--I don't have evidence.We're working on testing some of these hypothesesof these real-time research being done right now.But to me, these might be where weneed to look at if we want to understandthe effect of social media might be having on polarization.

  • 44:14

    PABLO BARBERA [continued]: If you want to read more about someof the research that we've done and we're planning to doand the things that we know and what we don't, a coupleof months ago, I wrote with some of coauthors and collaboratorsa literature review and like an overviewof some of these things that was funded by the HewlettFoundation.And it's available online if you want to read our report.

  • 44:36

    PABLO BARBERA [continued]: And, yeah.That's all I have for now.Thank you for your attention.I look forward to your questions.[APPLAUSE]

Video Info

Publisher: SAGE Publications Ltd

Publication Year: 2018

Video Type:Tutorial

Methods: Computational social science, Social media research

Keywords: algorithms; blogs; data mining; data visualisation; digital technology; disagreement; Facebook; homophily; information dissemination; network visualization; polarization processes; political ideologies; Social media; Social network analysis; Social networks; Twitter ... Show More

Segment Info

Segment Num.: 1

Persons Discussed:

Events Discussed:



Pablo Barberá, PhD, Assistant Professor of Computational Social Science at the London School of Economics, discusses concepts of political polarization and digital technologies, including how the consumption of political information has changed; whether echo chambers, digital enclaves, and filter bubbles increase political polarization; whether social media increases exposure to diverse ideas and political polarization; how political ideology of social media users is measured; ways to visualize social networks; ways to measure inter-ideological dissemination of information, exposure to disagreement; ways social media increases and/or moderates polarization; and what affective polarization is and how it is measured.

Video Info

Publication Info

SAGE Publications Ltd
Publication Year:
SAGE Research Methods Video: Data Science, Big Data Analytics, and Digital Methods
Publication Place:
London, United Kingdom
Copyright Statement:
(c) SAGE Publications Ltd., 2018


Pablo Barberá

Segment Info


Segment Num: 1


Segment Start Time:

Segment End Time:


Things Discussed

Organizations Discussed:

Events Discussed:

Places Discussed:

Persons Discussed:

Methods Map

Computational social science

Computational social science refers to computational methods and approaches to study the social sciences, particularly incorporating big data, social network analysis, social media content and spatial data.
Computational social science
Political Polarization in the Digital Age with Pablo Barberá

Pablo Barberá, PhD, Assistant Professor of Computational Social Science at the London School of Economics, discusses concepts of political polarization and digital technologies, including how the consumption of political information has changed; whether echo chambers, digital enclaves, and filter bubbles increase political polarization; whether social media increases exposure to diverse ideas and political polarization; how political ideology of social media users is measured; ways to visualize social networks; ways to measure inter-ideological dissemination of information, exposure to disagreement; ways social media increases and/or moderates polarization; and what affective polarization is and how it is measured.

Copy and paste the following HTML into your website