Skip to main content
Search form
  • 00:00

    [MUSIC PLAYING][Using Social Network Analysis to Understand Disinformationon Social Media]

  • 00:09

    YEVGENIY GOLOVCHENKO: Well, my name is Yevgeniy.[Yevgeniy Golovchenko, PhD Fellow,The University of Copenhagen]And I'm a PhD student at University of Copenhagenwhere I'm part of a large interdisciplinary projectcalled digital disinformation.And we are a team consisting of sociologists,political scientists, computer scientists trying to understanddisinformation in social media.And this is the main topic of my dissertation, that

  • 00:32

    YEVGENIY GOLOVCHENKO [continued]: is, disinformation of social media in the contextof the Ukrainian crisis.[How did you become interested in studying disinformation?]My interest in disinformation startedin my interest in politics on social media in general.I was observing many different social movements

  • 00:54

    YEVGENIY GOLOVCHENKO [continued]: and many different activists trying to change the world, soto speak.And then the war in Ukraine happened.And people started talking about disinformationabout military operations in Eastern Europeand use of information as part of warfare.And that caught my interest and made me dig deeper

  • 01:18

    YEVGENIY GOLOVCHENKO [continued]: into this topic using both theory from social sciencesand computational methods.Disinformation, in general, refersto intentionally misleading content.And we're looking at a very particular kindof disinformation, that is, disinformationthat occurs in the context of international conflicts.So if, for example, a war event is happening, for example

  • 01:42

    YEVGENIY GOLOVCHENKO [continued]: in Ukraine, and the big question is, who isresponsible for this actions?Because, essentially, disinformation,just like any information, can beused to mobilize populations, to mobilize war support,to mobilize dissent.[What computational methods did you use in your research?]

  • 02:04

    YEVGENIY GOLOVCHENKO [continued]: We collect large data sets about what do people actually doand how do they interact on social media.Particularly, I've been focusing on Twitter.But I'm also studying other social media siteslike Russian social media VKontakteand various other sites.And then we use, in particular, I

  • 02:27

    YEVGENIY GOLOVCHENKO [continued]: use a social network analysis to tryto understand what is going on, so to speak, online.Social network analysis, it's not just a methodology,but it's also a broader analytical frameworkwhere the point is to study relations between entities.So, for example, when you study social science,

  • 02:50

    YEVGENIY GOLOVCHENKO [continued]: typically, in your undergrad, youlearn how to analyze different attributes relatedto an individual, for example, gender and income.So you study, for example, what is the correlationbetween the two.In social network analysis, the unit of analysisis the actual relation between individuals.Their relation is at the core of the method.

  • 03:11

    YEVGENIY GOLOVCHENKO [continued]: When social network analysis was first, so to speak,conceived, among other times, in the '30s,it was used on very small scale.For example, among prisoners, maybe 30, 40, 50,100 prisoners.Now, we can perform social network analysison hundreds of thousands of individuals.

  • 03:33

    YEVGENIY GOLOVCHENKO [continued]: [Are there any tools or techniques that you use to dosocial network analysis?]I personally do social network analysisin R, which is a programming language.And to any of those who would like to learn social networkanalysis, I would recommend starting masteringthe package called igraph.

  • 03:54

    YEVGENIY GOLOVCHENKO [continued]: Both R and the package igraph are free.And that, I think, is a good start.[How does traditional social science theory influenceyour social network analysis work?]In my particular study, I examinethe role of citizens in international conflicts.Now, in international relations, normally what we analyze

  • 04:17

    YEVGENIY GOLOVCHENKO [continued]: are states.What does Kremlin or what does US do?Here, we try to show, using empirical data,that states are not the only actors that action matters.[How do you access the data you use in your research?]Usually, social media sites have something

  • 04:39

    YEVGENIY GOLOVCHENKO [continued]: called an API, which to put very simplyis sort of a gateway where the social mediacompany, for example, Twitter, can control what kind of mediayou can get and how fast you can get it.And so I recommend anyone who wantsto collect relational data to start using an API.

  • 05:00

    YEVGENIY GOLOVCHENKO [continued]: Alternative is to scrape the data.So imagine that we log in with a browser on a site,and then we download everything we see in our browser,and we do this a million times.This is what many people do.But you should always do this with cautionbecause, well, first of all, you may overwhelm the server.

  • 05:21

    YEVGENIY GOLOVCHENKO [continued]: And second of all, it may be a breach of the termsof service of that website.[What kind of research questions can you answer using socialnetwork analysis?]So how, or for what purpose, can weactually use social network analysiswhen studying disinformation?And there are many, many different kinds of questionsthat we can answer.

  • 05:41

    YEVGENIY GOLOVCHENKO [continued]: For example, you can try to understandwho plays a central role in spreading disinformation.You can try to understand whether disinformationspreads in one community more than the other.You can try to understand whether disinformationis isolated in different communities.And you can also ask yourself, how fast does disinformation

  • 06:06

    YEVGENIY GOLOVCHENKO [continued]: diffuse and how broadly?That's just an example of a broad varietyof ways to use this method to understand disinformationonline.[What are the ethical considerations when workingwith social media data?]There are a lot of ethical considerations.Unfortunately, it's one of the topics thathas been kind of neglected in many computational social

  • 06:27

    YEVGENIY GOLOVCHENKO [continued]: studies.And I think one should start with ethical questionbefore the actual data collection,or maybe even before the actual research question.And I think one of the main questionsthat we ask ourselves, and anyone should ask themselves,

  • 06:47

    YEVGENIY GOLOVCHENKO [continued]: is whether this can be used to harm the individual.That's one thing.However, that's not enough.It's not enough just to think about arethese people, the ones that we study, whether they're harmedor not.It's also a matter of respecting their autonomy as individuals.So, for example, it's possible for me, maybe

  • 07:10

    YEVGENIY GOLOVCHENKO [continued]: in some occasions, to download some of your private postsand I'm not causing any harm.You won't even know that I'm downloadingall this data about you.But in the end, I'm transgressing a walled gardenthat you have built yourself, so to speak.[What steps do you take once you have collected your social

  • 07:31

    YEVGENIY GOLOVCHENKO [continued]: network data?]So in the context of social analysis,social network analysis, the first thing we dois to try to think of data in relational terms.For example, when you download tweets,you can count how frequent are the tweets.You can look at different characteristics of the ideologyof the one who is tweeting.So the big task in network analysis

  • 07:52

    YEVGENIY GOLOVCHENKO [continued]: is try to structure the data in a way that is relational.So the most common and simple technique for beginnersout there would be to start thinking of datain terms of edge lists where we have at least two columns.Column one, let's say, we're having a conversation.

  • 08:15

    YEVGENIY GOLOVCHENKO [continued]: I would be in column one and you would be in column two.And each row would reflect some sort of interaction between us.A good way of starting analyzing the datawould, of course, be to try to ask yourself, whatare the relevant questions?And this may sound trivial, but actually,when many, especially people in computer science,

  • 08:39

    YEVGENIY GOLOVCHENKO [continued]: approach the data, they often start by just simply lookingat the data and asking, like, what kind of techniqueswill be cool?But you can use all the cool techniques in the worldand still not get a relevant answerthat's relevant to an actual research environment.So this is the first step.And once you have actually asked yourself, hopefully,

  • 09:02

    YEVGENIY GOLOVCHENKO [continued]: at least from social science perspective, a theoreticallymotivated question, you think what measurements canhelp us answer this question.For example, if you want to understandwhat are the key players in spreading disinformationonline, then you can think what kind of centrality can

  • 09:23

    YEVGENIY GOLOVCHENKO [continued]: we use to measure popularity, or importance, and so on.[What challenges have you faced during research,and how have you overcome them?]Well, what surprised me the most is perhapshow difficult it is to actually say what is disinformationand what is not.Typically, when researchers do that,

  • 09:43

    YEVGENIY GOLOVCHENKO [continued]: they often rely on fact checkers and someone elseto do this difficult work for us.But obviously, different fact checkersuse different criteria as to say what is disinformation,when is something disinformation.So what we have done, instead of focusingon many different cases of disinformation,

  • 10:06

    YEVGENIY GOLOVCHENKO [continued]: relying on a huge database of facts, so to speak,or non-facts, we focus on one casethat we try to understand in detail,taking one case at a time.So in a sense, in my work, I also try not onlyto focus on something from a quantitative perspective,but also having a qualitative, so to speak,

  • 10:29

    YEVGENIY GOLOVCHENKO [continued]: understanding of the context.You don't have to do this, but the pointis you can also combine network analysiswith a more qualitative understanding of whatis actually going on.[What are your recommendations for students pursuing this kindof research?]My top three advice is for people startingin computational social science with a social science

  • 10:51

    YEVGENIY GOLOVCHENKO [continued]: background would be one, start to code and be good at it.Two, don't sweat about not knowing everythingabout sophisticated methods, such as machine learning,or natural language processing, or even social networkanalysis.

  • 11:11

    YEVGENIY GOLOVCHENKO [continued]: Choose people from computer science who are good at it.And lastly, in continuation try to use your backgroundto bring to, hopefully, to a collaborationwith a computer scientist, or physicist,or from another discipline.For example, often, when I meet people

  • 11:35

    YEVGENIY GOLOVCHENKO [continued]: who have backgrounds in physics or computer science,I see that they are often very, very, very, verygood at methods in ways that you don't usuallysee if you study sociology or political science.But on the other hand, in my experience,seems like many of the sociologists, political science

  • 11:57

    YEVGENIY GOLOVCHENKO [continued]: I know, or psychologists have a very strong backgroundin methodology.Now, there's a big difference between methodsand methodology.Methodology is a more broader studyof which tools to use, whereas a method, a question of method,is like, how does this tool work?Those are two completely different things.

  • 12:19

    YEVGENIY GOLOVCHENKO [continued]: And I'd recommend social scientiststo remember their strength.And that is, I would say, definitely methodology.But with that being said, understand the methods, too,as well as you can.I think one of the things I'd recommendis to think about your audience very carefully.

  • 12:40

    YEVGENIY GOLOVCHENKO [continued]: Once again, it may sound trivial when you say to someone,think carefully about your audience.But actually, I notice many people start without--example, when undergrads start doing projects,they sometimes forget what is the audience.

  • 13:02

    YEVGENIY GOLOVCHENKO [continued]: And depending on your audience, or ratherwhich audience you communicate towill play an enormous role in what kind of questionsyou should ask.So, for example, if your audienceis a more network science approach where the people are

  • 13:25

    YEVGENIY GOLOVCHENKO [continued]: interested in, for example, in universal structuresof networks, maybe you would wantto understand some kind of universal patternin this network.If your audience is an international relations crowd,they don't care about the universal structuresof networks.They care about asking a specific questionabout specific political event, or maybe say

  • 13:47

    YEVGENIY GOLOVCHENKO [continued]: something about a more general theory about politics.I'd like to say that social network analysis isa very, very broad discipline.It's a very old discipline.And there's space for all sorts of peoplewith all sorts of backgrounds.You can be a humanist.You can study art history, and youcan do social network analysis.You can be a physicist.You can use that knowledge, that unique background,

  • 14:10

    YEVGENIY GOLOVCHENKO [continued]: to do social network analysis, or you can usepolitical science or sociology.So I would ask myself, once again,what kind of social network analysisI would be interested in, what would be my discipline?And I think I would begin by focusing

  • 14:33

    YEVGENIY GOLOVCHENKO [continued]: on one single tool at a time.For example, if you decide to use R, be good at R. Yeah.Don't just focus on R, Python, Java, and everything at once.Choose a package, understand its functions.

  • 14:56



Yevgeniy Golovchenko, PhD Fellow at the University of Copenhagen, discusses his research using social network analysis to understand disinformation on social media, including his interest in this field; computational method, tools, and techniques used; the influence of traditional social science theory; access to data; research questions that can be answered using social media research; ethical considerations working with social media data; challenges faced and overcome; and recommendations for students pursuing this type of research.

Looks like you do not have access to this content.

Using Social Network Analysis to Understand Disinformation on Social Media

Yevgeniy Golovchenko, PhD Fellow at the University of Copenhagen, discusses his research using social network analysis to understand disinformation on social media, including his interest in this field; computational method, tools, and techniques used; the influence of traditional social science theory; access to data; research questions that can be answered using social media research; ethical considerations working with social media data; challenges faced and overcome; and recommendations for students pursuing this type of research.

Copy and paste the following HTML into your website