Skip to main content
Search form
  • 00:01


  • 00:09

    DR. STEVEN LLOYD WILSON: Hi, I'm Dr. Steven Lloyd Wilson,and I'll be talking about social media and its usagein social science.I was an engineering and history undergraduate major,and then worked as a programmer out in the Silicon Valleyfor about seven years before getting into political science

  • 00:32

    DR. STEVEN LLOYD WILSON [continued]: and social science research.And I've always been interested in Russia,and mass organization, and protest.And back in December, 2011, therewere huge protests surrounding the Russian electionsthat were happening then.And basically in the course of an afternoon,I packed together a very, very crude Twitter downloader.

  • 00:53

    DR. STEVEN LLOYD WILSON [continued]: So I was downloading all of the tweets coming outof central Moscow during the protests.And from that, it sort of snowballedinto a big infrastructure that I'vebuilt that collects all geocoded tweets from the developingworld and I use that for a number of social sciencequestions, analyzing, basically, what is itthat normal people talk about.

  • 01:15

    DR. STEVEN LLOYD WILSON [continued]: As someone with a history background,historians get very, very excitedwhen we have a diary or just a scrap of lettersfrom a normal person from, like, an entire century in the pastbecause it's so rare for normal people to have a voice.And in the modern world, social mediahas billions of normal people not only having a voice,

  • 01:37

    DR. STEVEN LLOYD WILSON [continued]: but we're actually recording it in real time.So we can actually, for the first time in history,see what it is that normal people talk about.I kind of see that there's two different directions.One of which is the first generationand the second of which is, to me, the more exciting

  • 01:59

    DR. STEVEN LLOYD WILSON [continued]: generation of questions.So the first generation of questionswe were trying to answer--what effect does social media haveon various forms of politics?And we had a number of theoretical beliefsthat social media, by making it easier for peopleto communicate, would make it more likely for thereto be revolution and democratization,as normal people could organize much better than they

  • 02:22

    DR. STEVEN LLOYD WILSON [continued]: were able to in the past.That's interesting, but as social media has becomefar more of just a universal quotidian, everyone has it,we care less about what is the effect of social mediabecause whatever the effect was is a question for history now.So our social science questions have moved on to what can we

  • 02:42

    DR. STEVEN LLOYD WILSON [continued]: actually use this data for?And so a lot of my research now focuseson how can we use social media to actually produce real data?Tell us things that we didn't know about the world?And one of the things that it's great foris just normal people having conversations,there's a lot of stuff in there that

  • 03:03

    DR. STEVEN LLOYD WILSON [continued]: is all of the background noise of normal peoplethat used to be lost to history.So one good example of that is that if youlook at any research on mass protest,it almost inevitably is about big, giant protestsin the capital cities of countriesbecause that's where the reporters are atand they can actually see it happening.We know next to nothing about protests happening

  • 03:24

    DR. STEVEN LLOYD WILSON [continued]: in regional capitals or small areasbecause there's never a foreign journalist there taking noteson what's happening.With social media, we can tell that thereis a protest of 500 people in rural Bolivia, for instance,and we can tell what they're protesting about and why.And this is something that's never been a data source for usbefore.

  • 03:44

    DR. STEVEN LLOYD WILSON [continued]: So social media allows us to multiply what we can actuallytell about the world by seeing into the lives and the worldof people that didn't have an ability to communicate whatwas going on before.There's literally billions of tweets, trillions of posts.

  • 04:08

    DR. STEVEN LLOYD WILSON [continued]: One of the figures that I really enjoy reading aboutis that every single day on social media, thereis the text equivalent of 50,000 copies of War and Peacepublished every day.Now, none of it is up to the snuff of War and Peace,but that's because we've never had it before.Or on Instagram every single day, more photographs

  • 04:28

    DR. STEVEN LLOYD WILSON [continued]: are posted than exist in all of human historycombined before 1990.So the pure amount of data is something staggering.And one of the things to realize with big datais it's not just that we have more datathan we ever had before, but it's a different kind of data.And one of the best ways that we'reworking on being able to actually

  • 04:50

    DR. STEVEN LLOYD WILSON [continued]: make this pile of noise into actual datais to use a combination of both computers and human beings.So computers are very good at doing very, verylong, tedious tasks, but they do exactly whatyou tell them to do.They're not very smart.People are very smart, but you can'tmake them read a billion tweets, no matter how much funding you

  • 05:10

    DR. STEVEN LLOYD WILSON [continued]: have for graduate students.So a lot of us, what we do, we do these hybrid methodswhere we use advanced statistical techniques to tellus what is the text saying, but then we have humans review whatthat's saying.So for instance, I have projects thatare trying to measure nationalist sentimentthroughout Russian regions right now.

  • 05:32

    DR. STEVEN LLOYD WILSON [continued]: And instead of doing something like just,let's keyword search for the word nationalismon Russian social media.Instead, we do that as the first step and thenwe have human coders who are fluent in Russian,both linguistically and culturally,going through and hand coding 5,000 random tweets.Saying, yes, this is pro-Putin, this is anti-Putin,

  • 05:55

    DR. STEVEN LLOYD WILSON [continued]: and 95% of the time, this is irrelevant.And then we use a computer programand statistical software to be trainedby what those human coders did and saidto generate a second batch that says, OK, the computer ispretty sure this pile is really relevant.And then we have humans go through those.And each iterative step, we get better and better.

  • 06:16

    DR. STEVEN LLOYD WILSON [continued]: So we can train the algorithms suchthat we're really confident they're working wellbecause humans have sort of trained it graduallyhow to work, and suddenly we can measureRussian regional nationalism running on literallytens of millions of tweets.And running in a few hours and make a nice pretty graphand a nice pretty map, and mapping software

  • 06:37

    DR. STEVEN LLOYD WILSON [continued]: that we can put on our PowerPoints.One of the fascinating things that I learned on one projectwas dealing with the Ukrainian protestsback in 2014, the Euromaidan protests.And I was collecting geocoded Twitter data in real time.

  • 06:59

    DR. STEVEN LLOYD WILSON [continued]: And this was not keyword based, itwasn't me searching for the Ukrainian word for protest.It was, give me everything in a box that contains Ukraine.And so all this is is it's all of the tweetsthat people are posting from their GPS enabled smartphonesand such.So we know their position within a meter or two of accuracy.

  • 07:22

    DR. STEVEN LLOYD WILSON [continued]: So we can put up maps of central Kiev with all the buildings,and we can see this person posted right hereat this moment in time.Which, from a historian's perspective,is we can actually generate sort of real time oral historiesof people that we know for a fact,they were on the ground at a placeand said that this is happening.

  • 07:45

    DR. STEVEN LLOYD WILSON [continued]: And then on a much larger level, oneof the technically fascinating things we found outwas that nationwide, when a mass protest is happeningor a mass event of any kind, the tweets basicallygo up overall in the country because everyone in the countryis talking about the big event happening.But in absolute terms, the numberof tweets in the center of the capital, in Kiev,

  • 08:08

    DR. STEVEN LLOYD WILSON [continued]: where the protests were happening,they fell through the floor because--and you could tell what was going on when you zoomedin on that building level map.Where all the violence was happening,there wasn't a single tweet coming out.Because very rationally speaking,if you are fighting at barricades,you aren't on your smartphone tweeting, no matter

  • 08:29

    DR. STEVEN LLOYD WILSON [continued]: how obsessively you tweet.But only two blocks away, there was a huge numberof tweets where people were organizingthe action that was happening.And we've never been able to see sort of that dynamic otherthan just narratively people telling us thisis happening when they're on the ground in a particular place.And one other interesting thing that came out of itwas that when we weighted the tweets by number of followers

  • 08:54

    DR. STEVEN LLOYD WILSON [continued]: that a person had--which most people had a handful of followers,a few people had millions of followers,it was very, very skewed is how we described it statistically.What you noticed was that the pattern flipped nationwide.So what a mass event is, if you want to be almost pedanticallytechnical about it is it's when the network central actors

  • 09:18

    DR. STEVEN LLOYD WILSON [continued]: geographically converge on the same location.So all of the Ukrainians, or a great bulk of the Ukrainiansthat had tens of thousands or hundreds of thousandsof social media followers, they were all in a two block radiusduring the Euromaidan protests.And what we know from classic political scienceis that what drives mass events ironically has nothing

  • 09:41

    DR. STEVEN LLOYD WILSON [continued]: to do with how strongly you feel,how much you hate the government, whatyour political standings are.Any of these things that would be the obvious definitionsare terrible predictors of whether a single person willgo to a protest.The single thing that is the biggest driver is simply,do you already know someone who's going?

  • 10:02

    DR. STEVEN LLOYD WILSON [continued]: And social media has demonstrated that, well, it'sway easier to know somebody.50 years ago, to know that someone was going to a protest,a friend had to call you on the phone,you had to run into a friend.Now by the multiplication effect of just social media,you always are connected at all times,for good or for bad, if you have anyone

  • 10:23

    DR. STEVEN LLOYD WILSON [continued]: in your extended kind of tie network of peopleyou know going to a protest, you will know.Which increases the odds that you will go to that protest.There's definitely a distinct urban/rural divide,

  • 10:44

    DR. STEVEN LLOYD WILSON [continued]: and this goes trans-nationally, wheremost tweets and most social media activity that'sgeocoded, at least, is coming out of the capitals.So for instance, if you just put a map up and show numberof tweets per province or what have you,it will look exactly the same as a population density map.So we have to do a lot of tricks in terms

  • 11:05

    DR. STEVEN LLOYD WILSON [continued]: of-- or not tricks, but a lot of techniques, whichis what tricks are, basically.It's a fancy word for saying it.Is to figure out-- to weight things properly.So that way we're tracking thingsby per capita in the area and such.In terms of cross country, like what countries are different,what trends are there in countries.

  • 11:25

    DR. STEVEN LLOYD WILSON [continued]: There's been some really interesting findings there,where it's not like you might havethe presupposition or the assumptionthat this is basically wealth driven,that places that are either wealthier or more educatedare going to have far higher social media usage.And that doesn't actually necessarily end upbeing the effect.It's definitely the trend because the richer places,

  • 11:48

    DR. STEVEN LLOYD WILSON [continued]: more people have smartphones, more people have free timeto do things like tweet.But we're definitely getting a huge amountof data out of places you wouldn't necessarily expect.So for instance, a full 20% of the tweetsI have in my enormous worldwide dataset are all from Indonesia.Now I personally, as a political scientist,know nothing about Indonesia.

  • 12:09

    DR. STEVEN LLOYD WILSON [continued]: So this is fascinating to me, just as a research question.Latin America uses social media on a staggering level,far higher than their per capita incomewould suggest if you just thought it was a GDP element.

  • 12:31

    DR. STEVEN LLOYD WILSON [continued]: One of the advantages of social media datais that it's very low threshold for entry.There's basically zero cost if you have a computer.There's startup cost in terms of havingto have some knowledge on how to do it,but you can start very, very basic data collectionliterally from your web browser by just searching

  • 12:51

    DR. STEVEN LLOYD WILSON [continued]: on Twitter for a keyword and starting to downloadthose tweets and read them.That can, with just hours of effort,give you thousands of things thatmight be relevant to your research question.You can obviously scale up to very, very sophisticated thingsvery quickly.But there are a number of software packagesthat have nice graphical interfaces that

  • 13:12

    DR. STEVEN LLOYD WILSON [continued]: allow you to search for and download tweets and thendump them to, for instance, Excel spreadsheets.So that way you can move on to doing more advanced things.And then, of course, you can alwaysstart to talk to professors, like me,who specialize in it, who, generally speaking,can start downloading whatever you want to in general.So for instance, in the classes that I

  • 13:34

    DR. STEVEN LLOYD WILSON [continued]: teach using these methods, I frequently,for undergrads when they have their own researchquestions and their own research projects,simply tell them to give me a list of keywords or geographicareas that the Twitter--the tweets from those areas or on those topicswould help them answer their research question.And I already have the infrastructure set upto just start downloading that.

  • 13:54

    DR. STEVEN LLOYD WILSON [continued]: SoI've had dozens of undergrad projectsthat have had original data collection.Which, 50 years ago when we really relied on survey dataor on the ground data, the idea that undergradswould have in their grasp the abilityto get the opinions of normal people in different countriesas original research data, this is

  • 14:16

    DR. STEVEN LLOYD WILSON [continued]: a huge leap forward in our abilityto know what's going on.

Video Info

Publisher: SAGE Publications Ltd

Publication Year: 2019

Video Type:Interview

Methods: Social media research

Keywords: data analysis; data processing; geocoding; oral history; politics; politics and society; real-time communications; Social media; Social network analysis; Social networks; Social science research; Software; Twitter ... Show More

Segment Info

Segment Num.: 1

Persons Discussed:

Events Discussed:



Dr. Steven Lloyd Wilson, PhD, Assistant Professor of Political Science at the University of Nevada, discusses the use of social media data in social science research, including the kinds of research questions that can be answered, the types of analysis to use, what has been learned through research projects, how social media is used differently from country to country, and finally, how to get started using social media data for research.

Looks like you do not have access to this content.

Steven Lloyd Wilson Discusses Using Social Media Data for Social Science Research

Dr. Steven Lloyd Wilson, PhD, Assistant Professor of Political Science at the University of Nevada, discusses the use of social media data in social science research, including the kinds of research questions that can be answered, the types of analysis to use, what has been learned through research projects, how social media is used differently from country to country, and finally, how to get started using social media data for research.

Copy and paste the following HTML into your website