Skip to main content
SAGE
Search form
  • 00:00

    [AMBIENT MUSIC][UPBEAT ELECTRONIC MUSIC]The mission of Social Media Lab is to basically study us.Why do we go online?

  • 00:21

    What do we get out of it?How is it changing our society?How is it changing the social contract of howwe connect to each other?So it has a huge impact.Yes, so maybe let's put this user back--We started online communities.As more and more people are hanging out online,we as researchers have to go to where the people are.There is suddenly a toxic conversation

  • 00:42

    happening between this user and that user.One of the unique features of the Social Media Labis that we are very interdisciplinary by design,and the reason is that because when we study social media,you really need expertise from multiple backgrounds, includingsociology, psychology, management, informationscience, and of course, computer science.

  • 01:04

    So you do need computational techniquesthat will then help you to analyze the avalanche of datathat is out there.Today, let's talk about this dataset that Twitter just released, and itincludes bots that were created by Internet ResearchAgency in Russia, as you know.And they were trying to influence the election

  • 01:25

    in the US and other countries.And so this data set is interesting because it covers--and this is a quick histogram of number of poststhat this data set includes--it covers data from 2010 all the way to 2018.And certainly, there are some timeswhen those bots were more active and some not as active.

  • 01:46

    And it would be interesting, I think,to understand this engagement level with other Twitter usersthat these bots have trying to understand whatthey tried to do on Twitter.The Lab is a very interdisciplinary space.We attract people and students from a wide varietyof disciplines and backgrounds.

  • 02:07

    We employ everybody from undergraduate studentsall the way up to postdocs.So today, let's just brainstorm types of research questionswe can ask, and just generally thinkabout why this data set might be useful for research.Since you showed that chart, and they started in 2010,some of the bots started becoming active then.

  • 02:29

    I'm interested in seeing how long it took them to, I guess,work their way into the network because as wecan see from the network graph, some of themare central nodes to those clusters.I'm a data analyst and web developer.The reason I switched from economics to computer sciencewas because I got a basic understanding of, I guess,all the macro and microeconomic principles,

  • 02:49

    and what I saw was that there's a lot of dataat every level of each of those steps.And computer science was always a passion of mine,so I decided to switch and then see if Icould try to combine the two.We are embedded at a university that values entrepreneurialism.So this Lab, even though it's a research lab,

  • 03:11

    we behave and act as if it's a small startup company.So one of the ways that we bring new blood and new ideasinto the lab is via our visiting scholar program.Every year, we put out a call for researcherswho are doing similar stuff as us, but who might need

  • 03:31

    and can benefit from our expertise.And usually, we get 50, 60 applications.We pick four or five projects thatcan add to or enhance our projectand also enhance the other researcher's project,and then we invite them to come and hang out with us.So it's been a wonderful way to meet colleaguesfrom areas and universities that we might not normally

  • 03:54

    meet or run into.So it makes for a very dynamic and interesting workspacebecause literally, we have newnessbuilt into our infrastructure because every few months,we have somebody else coming from another part of world,from another discipline.I actually do want to see the visualizations with that

  • 04:14

    set over time because you see how at the bottom right,there are some notes that are just kind of outside.My primary role at the lab is to do data analysis.So first, that contains the full breadth of it.So I will collect data.I'll clean data that needs to be cleaned.I'll do a bunch of different analysis, anythingfrom visualization to getting raw values of distribution

  • 04:37

    charts.So I'll do that full range of data analysis,but then I'll also do a full stack web development.So that means managing the servers,creating the front end, creating the back end, and then 2A dataanalysis.I'll also integrate tools like APIsand create proprietary code to facilitate data analysis

  • 04:59

    on the websites themselves.Do you produce graphic mail file?Yeah.The skills that I need to do this workwould be primarily computer science skills, so programmingskills, knowing how to solve various problems involve APIsand data collection, and also having

  • 05:21

    an idea of the structure, so data structures.So what format is this in?What format do I want it in?How can I speed up data collection?How can I speed up data visualization?How can I speed up data processing?If I'm pulling this from this website, that's in this format.How do I transform it to this format?

  • 05:41

    And along the way, not only do youneed to know the actual structureand understand the data structure,but also understand how to code thatand then how to actually use it and make it useful to people.Because of the size of this data setwith over 9 million accounts in there--I mean, tweets in there-- we probably will have to look

  • 06:03

    at distributed computing solution to figure out howto visualize this because otherwise,we're going to be sitting there for a very,very long time trying to wait for--A mistake.--a computer, a desktop computer to render this image.I guess it depends what kind of databaseyou really want to use to visualize it.Would you like to use Neo4j?Would you want to create the network graph

  • 06:24

    and then visualize in something like Gephi?So I guess that's one of the first questions we want to ask.In order to understand what that data is trying to tell us,we have to have computer science people on boardwho know how to get that data, clean it up, and visualize it.We have to have social scientists whounderstand and are trained in social theories

  • 06:47

    so that we can have a good framework to understandwhat it is that we're seeing.Is this an anomaly?Is this something that is part of a bigger pattern?Otherwise, you're just reporting numbers,and numbers are meaningless without the theoryto back it up.I'm a post-doctoral researcher here.My background is in political sociology theory,

  • 07:08

    and I analyzed Twitter networks of the Egyptian revolutionover different time periods.And that's what introduced me to social network analysisbecause this type of research, you cannot study it withthe regular methods of sociology.Based on your research that you'vedone before on social media and activism,

  • 07:28

    how do you think dangerous to have bots like this so activefor the society?Do you think it's just an online-only phenomenon?So the problem is usually, in most of the Twitter networksthat were used to study activism,you can find a link between online and on the ground.So my research showed that during the Egyptian revolution,there were some kinds of links between what was happening

  • 07:51

    in terms of communication online and whatis happening on the ground.So though maybe in this one, you won'tbe seeing the bots on the ground,but their effect that happened onlinecould also affect things that will happenon the ground, which probably we have seen before during the USelections and what happened after.

  • 08:12

    My experience with the Lab was through their conferences.So I used to attend their conference.They have a conference every other year here in Toronto.So I used to attend these conferences.I found that the subjects that they doand the different studies that they makeis pretty much close to my research interests,so I found also a very nice way of collaboration.

  • 08:35

    The teamwork at the Lab was very good.So I said that would be my post-doc destination.The lab also has a visiting scholar programwhere different scholars from different disciplinesand different countries come through the laband collaborate on different research,and that's, again, helped me start thinking about new ways

  • 08:57

    to do stuff, different cultures, different questions they mightwant to ask.We set up the Lab in a way so that it canchange very quickly on the fly.Just like the equipment in this lab, everything is on wheels.Our people are very flexible in their skill set,and we train them so that they canjuggle many, many different balls at the same time

  • 09:18

    and many different projects.And we start that process as soon as we hire them.We have an onboarding process because, again,some of the students are here for four months.Some are here for two years.And depending on how long they are with us,we need to have a way to basically get them up to speed.So yeah, let's take a look at one

  • 09:39

    of the subreddits that tend to be more hostile in nature.So it's Donald Trump supporters.So we have observed a lot of conversationsare happening-- that reddit is very hostile.So let's identify whether this hostility is based on oneindividual or it's throughout.

  • 09:60

    So let's visualize this network.The Reddit collector tool would be a toolto help you do your own analysis.You can look at a community.You can start to ask what are they arrested in,how are discussions on this subreddit,and then you can collect data.

  • 10:21

    Clearly, we see some nodes that tendto be more active than others, and that'sone of the advantages of visualizing as a graph,and you immediately see that some participants aremore active than others.It was developed firstly as a web-based tool,so that means we needed a web server.It was coded in Python using Django,and that allowed us to integrate a lot into it,

  • 10:42

    so we could develop our own Python code to do stufflike draw the network graph.We can also integrate APIs in our workflow,like perspective to-do sentiment analysis.Tell me a little bit about what the scores are hereand which one you would recommend to use.So I'd recommend Toxicity firstlybecause it's the most fleshed out in the API.

  • 11:03

    Also, it gives the best scores basedon the actual content in it.So are they using profane language?Are they specifically attacking?So I guess it's the most general.The Perspective API is an API developedby Google in collaboration with places like the New YorkTimes, which have large comment sections,

  • 11:24

    and they want to foster positive discussion.So they trained a machine-learning algorithmto identify what they classify as toxicity.So what is a toxic comment?And how can we stop them before they happenor before they're commented and degrade the discussion?

  • 11:45

    So we use that, which automatically scores commentsbased on their context and wordingto give them a toxicity score.Say, this is a toxic comment.This is not a toxic comment on a scale from 0 to 1.And then you can use that to score your own comments.So we run that through, we get scores,and then we give them to you.

  • 12:06

    This comment is this toxic.This comet is not toxic, and we allowyou to do what you want with those scores.Big data sets essentially allow us to do as data scientistessentially to study society at scale.It used to be that you can only study a very small sampleof your population that you are interested in studying

  • 12:26

    through more traditional ways of collecting datalike surveys and interview.So with the emergence of social media data,and especially of publicly available social media data,it allows us to study the issues that society is concernedabout at the large scale.And it also allows us to see how society and the opinions

  • 12:49

    change over time.Just being in this lab with peopleof different disciplines, you startto need to ask questions and need to interact and startto gain new thought processes of,I thought about it from this angle,and I'm a computer scientist, but they thought about this,and they're pulling from this angle,and they're political science.

  • 13:09

    Being an undergraduate myself, I learned a lotfrom people who are doing Mastersand PhDs and people who are post-their PhDor post-doctorates.And then the kind of questions they come up with their stuffthat I haven't even thought of yetor begun to think on that level.And that's really helped me elevate my own thinking

  • 13:31

    of how can I solve problems.How can I think about problems?How can I think about the data I'm collecting?And what kind of questions do I want to ask,which has been the biggest ones because a lot of the questionsthey ask me have completely set me on a new path of thinkingand a new way of thinking about the work I'm doing.

  • 13:52

    And since social media and specifically technology movesat such a fast rate, we're alwayslooking to do new things like integrate the PerspectiveAPI into the Reddit collector, liketake a look at different social networks,like take a look at ongoing discussioninside of subreddits.And then, from there, you start to need to ask new questions.So what does it mean to be toxic?

  • 14:16

    Why are discussions devolving into toxic discussions?What's happening on social media?How are people connecting?What are they talking about when they connect?Why do they even want to connect?The great thing about education in North America and Europe,basically, is that we teach people to zoom in and go

  • 14:40

    in, dive deep into something.But one of the things we try to do hereis, let's zoom out a little bit.It's important that you have depth,but it's equally important that you have breadth.So what we try to do here is we say,OK, you need to be an expert in whatever it is that youhave chosen to be an expert in.

  • 15:01

    But always ask yourself, how is your expertisecontextualized in the larger world that we exist in?I think the Lab has changed my vision of myself,as I've started to ask more questions about what canmy work enable, rather than beingthe end point of I've created this thing, and now it's done.

  • 15:24

    What kind of work can this platform I've createdenable other people to do?Social media is growing.It's everywhere, and it's very importantright now because literally a country willhave a Twitter handle, a president willhave a Twitter handle, a prime minister willhave a Twitter handle.

  • 15:44

    But then there's also a lot going on that can be good.So how do you create movements like March for Our Lives,but then also, how do you swing an electionwith misinformation?And both those things are happeningon the same platforms.So I think it's really important to understand, to knowhow to analyze, to know how to look at,

  • 16:05

    and then to know how prevent negative thingsand try to foster positive things.

Abstract

Social Media Lab's Philip Mai, MA, JD, Director of Business and Communications, Anatoliy Gruzd, PhD, Director of Research, Tomi Adegbite, Data Analyst and Web Developer, and Deena Abul-Fottouh, PhD, post-doctoral researcher, discuss their experience at the Social Media Lab, including the type of researchers the lab attracts, what makes it unique, important skill sets to have, its multidisciplinary team approach, social media research tools used, and the benefits of being part of the Social Media Lab.

Looks like you do not have access to this content.

Creating an Interdisciplinary Research Laboratory: Social Media Lab

Social Media Lab's Philip Mai, MA, JD, Director of Business and Communications, Anatoliy Gruzd, PhD, Director of Research, Tomi Adegbite, Data Analyst and Web Developer, and Deena Abul-Fottouh, PhD, post-doctoral researcher, discuss their experience at the Social Media Lab, including the type of researchers the lab attracts, what makes it unique, important skill sets to have, its multidisciplinary team approach, social media research tools used, and the benefits of being part of the Social Media Lab.

Copy and paste the following HTML into your website