Digital Humanities Lab: Social Platforms Small & Large Scale Research

[00:00:05]Good morning everyone. Thank you very much for  joining us today for the University Library's   Digital Humanities live webinar. This is  the first in our five-part series for 2022.   I'm very pleased to see a number of you attending.  My name is Alexis Tindall, I'm the manager of   Digital Innovation here at the Library and first  up I just wanted to mention I live and work on   the lands of the Kaurna people and before I  start I would very much like to acknowledge   and pay my respects to them as the traditional  custodians whose ancestral lands we gather on,   even virtually. I acknowledge the deep feelings  of attachment and relationship with the Kaurna   people to country and respect and value their  past present and ongoing connection to the land   and cultural beliefs. Given that this is an  online session I understand there's people   probably joining us from other parts of Australia  so I extend that respect around the communities   and especially to any Aboriginal and Torres Strait  Islander people who might be joining us today.   I'm really pleased to welcome you to this  session, very excited to kick off for 2022   and I encourage you to visit the University  Library website to look into the program ahead.   The University Library's Digital Humanities  Lab webinar series is designed to inspire and   inform Humanities researchers and their potential  collaborators who are interested in using digital   approaches in their work, and by potential  collaborators I'm including other researchers,   other disciplines, students, representatives of  the galleries, libraries, archives and museum   sector and industry anyone who wants to attend.  In these webinars. We showcase real projects and   researchers that are using data digital tools or  other digital research methods to drive, enrich,   complement or communicate their research and our  intention is that these sessions are approachable,   enjoyable and thought-provoking. The Library  started this series in 2020 as a response   to our researchers' interest in showcasing their  digitally innovative work and in raising awareness   of new tools and support that is out there,  perhaps stimulating a bit of peer learning   and in the best cases facilitating networking  could lead to interdisciplinary collaboration.   As you have registered for this webinar I'm  going to add you to my circulation list for   future sessions and please feel free to opt out  of any time by replying to me. And just a bit of   a housekeeping about how today will work we have  the lovely Matt from our Academic Liaison team   on board, he will be monitoring our chat window  for questions or anything that you want to mention   there. Please feel free to introduce yourself  in the chat as well. Today's session will be   recorded and it'll be shared after the event on  the University of Adelaide Library YouTube page   and you'll be sent a link to that recording after  the session. Feel free to share it with anyone.   If you have any questions please look to the  bottom of your Zoom window and you can find a chat   icon and you can ask questions in the chat  window; they can be sent through at any time   during our presentations today and i will post  those questions to our two speakers at the end of   their presentations. So today we're going to hear  from two researchers in really different research   disciplines who are both researching social media.  As our social, working and creative lives depend   ever more on online platforms these environments  are of enormous research to researchers from a   number of different angles. They're exploring  how we engage and interact in these environments   and all the positive and negative impacts those  interactions have on our lives in our community.   Effort is being put towards not only saying,  not only understanding and advancing technical   environment, the security and integrity concerns,  how information spreads and how users create their   online identity, their intentions and agency in  this space. So with that i would like to perhaps   introduce our first speaker Francisco Zanartu  who is a researcher from our School of Computer   Science who's working on a Facebook or Meta funded  project to consider integrity and privacy issues   through social testing. Francisco is a former  Chilean broadcast television executive, he was   involved in the development of major productions  as well as an intensive digital transformation   process that positioned the media holding as  the country's undisputed ratings leader. He's   joined us and completed a master's degree in  Data Science where he studied state-of-the-art   methods for automatically classifying GitHub  repositories and with that Francisco I'd invite   you to share your screen and I look forward to  hearing what you've got to say today, thank you.   Okay thank you for your introduction  and I'll share my screen one sec please.

[00:04:20]Okay can you all see my presentation?

[00:04:26]Yes that looks great thank you. Cool thank you.   Well my name is Francisco and I will be presenting  about socials a Facebook found research project   that I'm working on along with professor Christoph  Treude from the School of Computer Science   of the University of Melbourne and Marcus Wagner  from the University of Adelaide. Well, as you may know Facebook is a social network for you to  connect with friends and family but to give a bit   of context to this project the first idea that  i want to install is that Facebook is massive.   So it's almost 3 billion users worldwide  on average each user spends half an hour   on the on the platform and that usage  generates 4 million gigabytes per day.   So that is Facebook and Meta a Facebook owner  has deployed a Facebook scale simulator to detect   harmful behavior in his network and provide  better security a standard for his users.   So this this simulator is called a Web Enabled  Simulation or WES and it has three main   attributes: it uses machine learning to train bots  and WES is able to automate the interactions of   these bots and it's also able to deploy this bot  and test them on the actual, on the real Facebook   the the one that we all use. But there are many  challenges in building a platform like WES so in   order to speed the progress Meta is seeking input  from others in the research community inviting   academic researchers and scientists to contribute  with new ideas and that's where we came in.

[00:06:29]So WES can simulate user behavior but  what behavior should it simulate? This is   a research question that motivates our project.

[00:06:42]Social times a diversity to find the  user behaviors that will allow us to find   new problems so the idea is to diversify the  user behavior to help WES finding this problem.   I will explain this in more detail later but what  challenge here is that we don't have access to WES   in in our project so we will build our  project with data from GitHub. We have it's an   online platform as well and here we come with  the expertise of professor Christoph Treude,   he's a recognized world leading researcher for  mining interaction from repositories and also   with the expertise of professor Marcus Wagner on  diversity optimisation. So the first thing is like   we will capture the data from users on GitHub  we will characterise this data and we will test   this useful that we found in in GitHub  with some abnormal or random behavior.

[00:07:53]But you might be wondering what is GitHub  right, so GitHub is a big community   of 72 million users and around 233 million  repositories. It has the largest host of open   source code on the internet, is a cloud platform  where software developers build and maintain   the project and it have integrated kits, a person  control tool to keep track to change to the code.

[00:08:25]You can think about these repositories as folders  in your computer that contains code or text files   basically and the users here contribute to  each other improving those text files with   new ideas or better implementation of the  codes. And GitHub also allows you to create   bots that automate tasks in the platform so  let's say that for it change in the code, you   have to perform some tests you can create a board  that make that test for you and GitHub recently   released copilot that's an artificial intelligence  that autocompletes code helping develop solutions   faster right. And to be more specific a typical  workflow might be as follows: so let's say here let's say you're using a GitHub code and you  found a problem with that code, so you log   in into GitHub, you go to the repository that  contain that code and submit an issue. So this   would be an event inside the GitHhub network  and the next thing that will happen is that   the owner of the code or some someone that is  contributed to that repository will fork and   clone that repository in his own computer and he  will modify the code and submit a full request.   And from here let's say another person in this  project, in this folder, will review and test the   code the new code and will merge  it with the original code right.   But how can we analog this to Facebook? Well we  can think on this event as Facebook events right   so for instance a fork event will be similar to  giving a like to a public profile page on Facebook   and we can build a graph data structure  that in our case is a set of nodes, these   big blue dots here that represent users on  repositories on GitHub and edges that are   the actual events that occurs in the network and  that will allow us to explain how users interact.

[00:10:56]And so how we are capturing this behavior,  we're taking the GitHub database that is   publicly available on Google Cloud and building  our own custom graph data structure with the list   of events that are registered there and among  other things we are calculating the centrality   metrics that indicate the relative importance of  a node in a graph so let's say for instance this   graph here with the highest degree of centrality  is the user that has participating more events   inside the network and also we're calculating  some other thing like profile similarity metrics.   That would be for instance how many  attributes you share with some other   profile in the network if you have  the same name, the same picture,   the same location on all that sort of information  that we fill when we sign up for for a network of   these characteristics or some anomaly scores  based on how abnormal an event might be   for a user depending on the  time that such event happens.

[00:12:21]Okay so here i will introduce the  concept of evolutionary algorithms   and that will allow us to understand how we are  optimizing the device diversity of GitHub users.   So these algorithms are inspired by the biological  evolution and use operation operators such as   selection mutation and reproduction to find  candidate solutions for optimization problem.   So let's say let's have a practical  example of this, let's say you want to climb a hill right now draw it here.

[00:13:07]And you start, you you don't know the  landscape and you start at a random point   so this will be x and y coordinates  right. You just start here and   keep random steps at an unknown  direction and on each step   you elevate your altitude right. So if you work  here they say okay I'm in a higher attitude I'm   here, here, so the higher up attitude means that  you're getting closer to the optimal point right.

[00:13:50]So this is an example of how diversity  algorithm helped us finding a diverse   distribution of users in the network. We are  using the so-called star discrepancy score as our evaluation metric this star discrepancy  measures how well the data points are distributed   in our space and here we we're calculating the  star discrepancy based on centrality metric and and on the degree of each  node just an example okay. And so for instance this is the an initial set  of actual users that we grabbed from GitHub   they seem to be clustered in these two  groups here and after our evolving process   they now look very well  distributed in the plane right. So is how they look now is the same amount of  user and this is how they look now they are and they are distributed with  our evolutionary algorithm.

[00:15:12]So here is a practical example of  this, let's say we have a popular user   this red node in the graph that has reached  the maximum number of friends in the network   so for instance for the case of  Facebook that would be 5,000 users   and let's say also that to avoid the spamming this  network has set a limit of the maximum number of   messages that you can send into a period of time  so that in Facebook will be like 150 messages in this period of time. But let's say also that  this user has been victim of some sort of hacking   thing, his account has been phished so you can  clone his profile after you can, you get access to   this victim account and access to  his entire network and spam his   friends but you still have the this  limit that the network is imposing.

[00:16:32]So a diverse set of clones could be created  to bypass the network security mechanisms,   and violate these spam rules and so this is our work  in progress and we will continue from here   developing new experiments and installing our  local version of GitHub or a GitHub server to   study what would happen from the server side  of the network with this diverse set of users.

[00:17:01]So in conclusion we are aiming at diversity to  find new behaviors that could potentially lead to   new issues and therefore simulation should focus  on this behavior in order to fix these issues.   So what Facebook would expect from our  work is we plan to develop a framework for   testing user interactions and potentially detect  these bugs or this harmful behavior that we took.

[00:17:33]So that is for me, I'll leave you with Alexis.  Thank you very much Francisco that's great thank   you. So what we'll do is if you have any questions  please feel free to post them in the window   and we will put them to Francisco and Kim  at the end of both of the presentations.   And with that I would like to introduce Dr  Kim Barbour who is a qualitative new media   scholar and Senior Lecturer in the School of  Humanities department of media. Kim's research   looks at online personae, the strategic production  of identity for digital media, particularly the   domestic use of social media. Kim teaches in our  Bachelor of Media and supervises post-graduate   research. She's published articles in Celebrity  Studies, oh hang on I've got acronyms here   and she's published articles widely and has  chapters in a number of edited collections.   She's also the managing a founding and  managing editor of Persona Studies journal   and has co-authored a book, Persona Studies,  and introduction which was released in 2019.   And with that I'll invite Kim to take the stage  thank you. Great thanks so much Alexis and thanks   to Francisco as well for a really really  interesting presentation. Algorithm stuff   always feels a little bit like magic to me and I  think that demonstrated clearly that that's still   the case but to begin. So as Alexis mentioned I'm  a qualitative new media researcher and I study   personas which are the strategic performance of a  version of self for an audience. Often that self   is constructed on or performed through social  media that's most frequently where my interest   lies, so my focus is on media types personas  produced and enacted through media technologies.   I usually avoid researching high status folks  such as celebrities as I'm more interested in   everyday and banal uses of these spaces. I have  a media communications background which means   that there is an influence in cultural  studies and considering the role of power,   personal, political, financial, platform,  audience power, that sort of thing,   along with sociological and underpinnings and the  interest in how socio-cultural norms and kind of   historical constructions of identity  influence persona development now.   So that's a whole lot of words to say that i  research what people do on social media, how this   contributes to their sense of self and how others  see them. So my research methods utilize online   listening to build a deep familiarity with the  habits and voice of a small number of participants   and often include one-on-one interviews where we  talk through the way that they use social media   and what they intend to achieve through that use.  So social media platforms are my field sites.   Using small data allows me to get deep into  experiences that may not be generalizable   but do add depth and richness to our understanding  of the ways that people use social media   and the range of ways that this new ubiquitous  software has produced ways of negotiating the   self. While small data research is limited and  its capacity to tell us trends across populations,   what it can tell us about is people and  a little bit about why things happen or,   just as importantly, don't happen. So in this  presentation I'm going to talk through some of   the challenges and implications of small data  research and where opportunities for insights   lie. So this will be framed by three key  constraints that guide human research   and all social media research is human research  so ethics law and institutional policies.   Okay so the default position in academic research  since the introduction of ethics frameworks has   been that protecting participants from harm means  obscuring their identities. Large portions of   ethics applications are devoted to privacy  security and anonymity; this is undoubtedly   important but also has set up expectations that  anonymity or pseudonymity is best practice. So   this is an area where there is actually now  alignment in terms of the increasing nuances   that need to be taken into account when thinking  about doing social media research and small data   research when you're looking at individual  people and where there's alignment between   legal frameworks around privacy and copyright,  around I guess institutional expectations of   researcher and behaviour and around ethics  as well. So firstly when we are considering   social media sites there are significant  differences in the use of their identifiers. So   some like 4chan are anonymous sites, your  identity is not noted there and I put   "anonymous" in closed quotes because there's  always ways and means if you're clever enough.   Reddit has persistent handles which are not a real  name but can be identifiers across sub-Reddits   or within a thread or can also be discarded just  as easily. Others like Facebook profess to using   real names but there are plenty of workarounds  for that policy including just lying. I think that   we often discount this idea that people just lie,  but people just lie sometimes. So all of this has   to be considered when we're thinking about how  we identify participants in these spaces and   how we think about these ideas of anonymity and  pseudonymity and obscuring identities by default   or not doing that, because participants may well  be comfortable with their existing methods of   identification depending on how close or far those  are from their identity and in human and in other   spaces. So the second component here is that  participants might not want to be anonymized so   this default to anonymity that stems from medical  research and some really really important things   needs to be considered as well. So for some  people they could gain status or credibility by   being included in academic work, by being featured  and named as contributing and having their work   I guess connected to them and being able to point  to that. And other people may want credit for the   labor that they put into their posts and that's  fair and it ties as well into the third point here   which is that the alignment of the  law ethics and policy moment is that   under Australian law social media posts are  automatically copyrighted as creative work.   So this means that as researchers we  need to respect people's moral rights   to their creative labor including when  that creative labor is posting online.   And I think that's something that often goes  under the radar with social media researchers   and researchers more generally as well. Okay  so a couple of case studies to put this in   the context of a few pieces of work that I've  produced, I realize they're very difficult to   see here but they are linked down the bottom as  well. So there's two case studies. The first case   study here is where i was using Twitter to explore  public audiencing; so that was looking at tweets   that were sent during the live broadcast of Love  Child which was a show on Channel Nine. The tweets   used the official hashtags, tagged the official  accounts associated with the show and the channel,   and also actors as well who were performing  within the shows, and generally position   themselves as part of a public discussion on  this site. But they weren't media commentators,   they were small accounts and so in terms of the  people whose tweets I was capturing through this   process. So to avoid surfacing those accounts I  used the tweets verbatim but I excluded the handle   and the publication. What that means is that the  user is able to claim their tweet if they wish,   they can point back to it, but if they  delete it then there's no real easy way   to connect that tweet to the person or the account  that tweeted it because it's not listed, it's   not named in the publication. So that was the the  kind of choice there. So the second study that I'm   addressing today is my doctoral research which was  a longitudinal study that also used interviews,   so I spent between 18 months and two years  following eight participants who were, to explore   the way that they performed artistness on social  media so to look at the artist persona. I was   checking their accounts at least daily, they lived  rent-free in my head and some still do, I still   think about these people on a regular basis quite  a few years on now. So the participants themselves   had kind of varying profiles in terms of their  level of public visibility but they all perform   all provided informed consent and they elected  how they would be identifiable and they all   elected to be identifiable in a way that connected  them to the accounts that I was studying them   through. So what this allowed is it allowed me to  use screenshots although I did need to be careful   to make sure that I obscured any identifying  information from other people who engaged with   the post. So that's this idea of where there  is alignment between law, policy and ethics.   The next one is an area that's kind of in  negotiation, there are areas of these three   kind of spaces which depending on their  interpretation would come into conflict. So this   is around ideas about profile and size of profile  and also around when we do and don't name people   and when we do and don't need informed consent. So  there's a tendency to collapse all types of social   media content in quite unhelpful ways, either  it's on social media and so therefore it's private   because it's where people connect with family  and friends or it's on social media and therefore   it's public and I can use it because I can access  it. So there are you know difficulties in making   practical and ethical decisions around  who's contributing to public discourse,   who's a public figure and how much of  what they say is reasonable to research.   So making those choices is complicated or making  those determinations is complicated. Follower   numbers by themselves are not necessarily a  useful metric particularly when the same somewhat   arbitrary figure, so numerical figure like a  thousand or ten thousand or whatever it might be,   when those numbers are used across platforms.  Because ten thousand subscribers on YouTube may   mean something very different to  ten thousand followers on Instagram.   So this is an area which is currently under  negotiation within institutions including this one   as Legal and Risk and human research ethics  committees and researchers work to find best   practice approaches particularly around if and  when it's appropriate to consider issues around   waiving consent. So here's a hypothetical case  study for you: a researcher wants to study the way   that different types of masculinity is performed  on Instagram to explore the way that high profile   men role model different ways of being men in  contemporary society. So in order to do so the   researcher wants to study those who have  influence and power in particular realms   and select a number of highly visible politicians  and high-profile successful footy players.   So having reviewed the methodology and the  methods and the data collection plan and so on   HREC determined that there's a good case to  waive consent pending Legal and Risk approvals.   Okay so Legal and Risk, and I'm actually kind of  paraphrasing here from advice received although I   am not doing this project, state that informed  consent is required for research that includes   data from social media sites on the basis that the  project will collect both personal and sensitive   information. So names and other identifying or  potentially identifying information is considered   personal sensitive information as a subset  of that and includes but is not limited to   information about a person's political opinions,  philosophical beliefs, sexual orientation or   practices and it's determined that studying  masculinity and looking at masculinity and   naming the people who you're analyzing  falls into both of those categories. So   taking this very strict line which as I  say is kind of under negotiation at present   this project they say you you must get consent so  you have to contact all of the footy players and   the politicians and say I want to study your, the  way that you perform masculinities on social media   and please give me permission to do that.  Right so sorry there is more there. Okay so similarly taking a kind of very strict  interpretation of this would mean that researchers   who study for example terrorism or gangs or white  supremacist organizations would also need informed   consent to look at how these folk use social  media. And I probably don't have to unpack why   those kind of examples are problematic and you  know the the profile and the status is not given   as a reason for waiving consent and so this  kind of very strict interpretation becomes   quite problematic quite quickly but as I said it's  kind of under negotiation at the moment. All right   so the final topic I want to kind of introduce  here today is where i think there is a direct   and to date unresolved conflict around social  media research that is worth quickly unpacking.

[00:31:40]So this is the the role of platform terms  and conditions and the way that we plan and   execute research into social media platforms. So  as users of a service we are bound by platform   terms and conditions; we see the implications of  breaching these when people are de-platformed for   repeatedly breaching terms and conditions, when  people get what's called zooped from Facebook   and have to serve 30-day or 60-day bans for  mentioning names of medication or being reported   for obscenity for posting breastfeeding pictures  or whatever the latest thing is I'm not getting   into those discussions today. But as researchers  we are also bound by University policy and by law   to abide by the terms of conditions of these  sites. I argue that allowing platforms to   determine how they are researched creates an  ethical and moral conflict. These are immensely   powerful organizations whose influence on  everyday society shouldn't be ignored and yet they   determine the conditions under what research can  like can be done and how that research is done. So   again I think there's there's some quite obvious  conflict here in terms of thinking about research.   Putting aside big issues here's a more  approachable one: Twitter is the easiest platform   to scrape data from, it's the most accessible  for research, it's the most researcher friendly   and therefore it's massively over represented  in internet studies research. This leaves other   sites critically understudied and it puts too  much emphasis on the impact that a site has   on a comparably small site, it's got a  comparably small user base for internet,   the social media platforms, and they are  quite a specific group of people as well,   the Twitter sphere. For fairly obvious reasons I'm  not going to connect you to a piece of work that   demonstrates that breaching terms and conditions  results in ethical and moral good because Legal   and Risk have my phone number but I'm just kind  of thinking this through as we go and thinking   about how we can address this conflict I think  is really important. Okay so that's the kind of   main part of my presentation I hope that these  provocations have sparked some thoughts in terms   of what it's possible to do both in social  media research without that kind of in-depth   computational and big data research methods but  instead with data that you can hold in your head   or in your hands. Both approaches and all of those  in between add to our understanding of the world   and the impact of these new new technologies on  the people who use them. If you're interested in   thinking about some of the ethical issues around  social media research or if you are a social media   researcher yourself I really encourage you to  start with the Association of Internet Researchers   Ethics Guidelines and they're available on the  AOIR website. It's a really great community of   scholars who if you're working with research and  online spaces you should definitely connect with.   The most recent version of the ethics guidelines  from 2019 includes specific guidance depending on   your research approach and method so it's  got sections on machine learning and AI,   partnering with industry, feminist research  ethics and an impact model. And from the 2012   version i really like the chart which introduces  a range of prompting questions to guide your   research design, so these are not rules, they're  guidelines so they're ways of thinking about   conducting research ethically online. So thank  you very much everybody and that's my bit.

[00:35:32]That's great thank you Kim. I was just looking  up the Association of Internet Researchers isn't   that right but yeah I'm just popping that  in the window for anyone who's interested,   it's worth having a look at there. Thank you  for both of your presentations, I found that   very interesting I mean at the most  basic level I find it interesting that   you're sort of studying the same thing  from completely different directions.

[00:35:58]But I think that's actually really  actually part of the point of bringing   the two different disciplines  here together because obviously   you know this is a, this is a social  sphere an interactive a creative sphere   and it's you know there are as many different  angles to take on it as there is on angles on the   real world and you know where does the distinction  lie these days anyway. It is the real world   we are still doing things we're still making  stuff yeah still people behind all the data. Yeah   one thing that I found interesting I did  like your comment early in the piece Kim   that algorithmic work feels like magic and I  just reflected on the fact that Francisco and I   caught up the other day and I found the  evolutionary algorithm section quite interesting   and I said can I can you tell me more about that  and he said it gets very complicated very quickly.   Which I think is the point but I do  think that actually was the point of   using the algorithm to blur the  people's interactions in that way   I have a thought sort of going around my head  as I listen to you both but these platforms,   people use each of these platforms for a purpose,  so the celebrities that you're talking about,   the politicians, the footy players Kim are  using it for the creation of a kind of profile   which would probably be using it for a different  purpose than, you know my auntie who wants to   share photographs of Christmas on Facebook and  uses it for a different use of social platforms   for a different reason from the GitHub companies  that you're talking about Francisco, who are there   sort of collaborating on code and sharing code.  And I'm wondering whether you have any comments   about how the objective of what the person is  trying to do would change your methodological   approach in how you research them. It was that  it, it feels to me, I'm going to answer my own   question, it feels to me like that's a bit of  a slippery point when you're trying to research   these platforms because it's difficult to  ascribe that objective as a researcher.

[00:38:03]Yeah well I think I'm sort of using the the data  that is publicly available on GitHub in a very   abstract way. I'm not revealing any of the details  of any developer on GitHub I'm not telling you   which are the events that they were in  building I just, we're just building this   graph data structure that is a very abstract  thing right and from there it's very hard to to read some kind of private information of any  use there so I'm very glad as well of been using   GitHub data instead of Twitter and not being  over representing the more studies there right.   But yeah I think I never heard of GitHub before  completing my Masters and when your coming into   this world you it's a really fascinating platform  to contribute to other projects so I would   encourage everybody here in this session to have  a look at GitHub and understand how it works and   how you can interact with other users there and  how they stated that there's publicly available   and it's hosted on Google right and how you can  wrap this data and build beautiful things for   researchers and this kind of studies. Yeah  it's interesting isn't it because intent,   so my work's at the opposite end, so Francisco  takes the data and abstracts it to the point that   individual people are no longer actually at the  center of what's happening it's looking at the   connections between them yeah. Whereas I'm  looking at a person and then the connections   come from that person and so intent is really  important and that's one of the reasons why   my research often includes those interviews  as well because I can't judge intent   without talking to a person. I can only judge  what I'm what I'm seeing, I can only analyze   and interpret what I'm seeing and so you do have  to get that kind of actually talking to people   and even online listening you can get a sense of  passions and rhythms of behavior and interactions   and those kinds of things. But I still can't see  if that's counted as successful for that person   unless they write a post saying I tried to do  this thing and it didn't work and you know very   few people do that when they're talking about what  they're doing online. But also GitHub as a site   as well as the level of kind of data analysis and  abstraction that Francisco and his team are doing   there is also other work that's being done on  that space which was really interesting so from   from my perspective there's some great work  that's been done on levels of participation   and engagement with women and the way  that women software developers and coders   interact on GitHub, the types of questions that  they ask, the number of responses that they get   and those kinds of things which are  quite, which are I guess visibilizing   the way that this field of software engineers  internationally interacts at a level of kind of   a gender where it's determined where  it's identifiable and the way that the   different types of coders operate and how that  also feeds into different types of projects.   And that can help address some of the larger  structural issues around women's participation in   computer tech. So as a space it's really really  intriguing there's some great work being done   there yeah so intent's what I'm all about and  it's really hard to determine if you don't ask.

[00:42:11]I'd love to, I feel like even the graphical  representation that Francisco shared of   users I almost feel like I wonder whether  there's a, I'm someone's surely done it,   what the graphical representation,  how different it might look for these   different types of users and whether you could  almost identify you know whether someone is a   just random punter versus corporation versus  celebrity, whether they're operating their   own account or not, whether you could tell  that from the data I don't know, I'm just,   maybe someone's looking at that. Francisco  do you have any thoughts on that. I mean yeah it's very hard to to answer that from  from the sort of data that that we are grabbing   here but if you look at this sort of graph  data structure you might realize that   normally bots are the actual user that  tends to have the more interaction with   with other users and that's sort of natural right  because every time that something happens that   triggers an action for the bot and the bots don't  sleep right they they're awake all day and they're   keep working, and I don't know developing  this action, these events inside the networks.   And I don't know and how they work on for instance  on Twitter or on Facebook but I think they're,   they act in a similar way right. Every time some  somebody is talking about something that this bot   is particularly interested in he will be there  answering and publishing comments on that, on the   topic and that of course will increase the the  activity inside the network. That would increase   their centrality? Well although different metrics  that they all sort of refer to the same thing,   it depends what are you looking at but all  those metrics look how how important and all is where it's a relatively  important you know in the network.

[00:44:48]If you are looking about what might  have the the highest degree because   some other common metrics, I don't want to dig  into the details here but some other some other   common metric for social networks is the let me back to the, between the centrality, so  that is how many times a node or a user acts as a   bridge connecting two groups of people inside the  network and those users are very important right   because the messages tend to travel through them.  Yeah so that's that's really interesting as well   that that's one of the things that is looked at  because what's often looked at from my perspective   is, yes those connections and and how those  connections work but and and how strong they are,   so looking at kind of the strength  of Dyer's theories from the '70s and   stuff and going are these people that are, that  you interact with a lot who you're close to,   do you get more value or or a connection out  of the I guess those people that you connect to   online who are more distributed in your network  because the people that you're close to are,   you engage with them all the time so you're not  necessarily learning anything new from them.   And so that's that's really it's interesting  to see it mapped at scale as well as looking   at it kind of the individual level and this  is why I think it's so important that all of   these types of research continue on and that we  find new directions to think about this stuff   because you know the maps tell us some things  about patterns but like you say Alexis that   intent question can sometimes get lost in that  process. But I think that's one of the strengths   of Francisco's teams' project is that actually in  that in the objective which is testing integrity   and privacy you're actually removing with that  evolutionary algorithm stage of obscuring what the   interaction of the event is, you're removing that  intent. And because yeah if you the researchers   interpretation will always go well if I'm trying  to try and find people spreading this information   or I'm trying to find people using the platform  for commercial purposes that it is allowed   then you'll sort of skew the research in some  way whereas taking the base level of users and making them weird through  that algorithmic process.

[00:47:33]Finding those weaknesses and sore spots and  the exploitable points and networks to allow   us to be able to continue to use them  with a degree of safety whatever that's   called and looks like and i  think really really disputed,   how safe these places are anyway and that I'm  as somebody who you know kind of depends on   them for my livelihood you know these are these  are questions that we need to to dig into and I   think this level of abstraction really helps  with that I'm just going to mute all this.   You're being, you're being suppressed Kim. I'd  just like to remind the audience feel free to   ask a question in the chat window if you're,  if you're interested in doing so. Francisco,   the people who are using GitHub, it seems like  it was fairly easy for you to get that data set,   so the data for GitHub users and interactions  is sort of hosted on google did you say?   Yeah you still need some technical skills because  it's, it's a really huge database I mean it's   for years you can have like four terabytes of  data so that it's hard to download everything   and work with that on your computer. So  we're building with the small pieces of data   that will be scalable in the time so we're  sort of developing a prototype for a project   with the expectation to find something  and from there develop something bigger.   But if somebody knows how to deal with  really big data that is Facebook right,   if they found some interesting  thing in our project   they will certainly know how to deal with this  amount of data. I mean what it takes GitHub a year   in terms of the weight of the data for Facebook  it's just one day. Yeah do you think that,   the I'm wondering whether the hosting of that data  and making that discoverable is part of the terms   and conditions of the GitHub site like that the  users know that, do you think the users know that   their interactions are available for downloading  or do you think they don't care? Presumably   is it safe to assume that GitHub users will be  slightly more data and technically confident   or literate in terms of those terms and conditions  environment or is that just is that suggesting too   much for them? Well I guess that is one of the  most magical things of GitHub that people know   that this is an open source platform and by the  open source that means that you can actually   publish your code and you know that somebody could  download your code and use it in his own benefit   without your consent. I mean you are implicitly  giving your consent to the other person to use   your code. What developers do is they share a  license for the code so sometimes you have to,   the most common thing is like if you use  open source code to to develop your solution   your solution also has to be open source.  That's the most common license in GitHub.   So yeah Kim do you think it would satisfy your  privacy concerns or your sensitivity consent?

[00:51:20]Look I think the level of abstraction  that we're dealing with here is   I guess to a point relieves those concerns anyway  it's certainly there's no kind of informed consent   thing going to happen in this space when you've  got the number of users and the level of data   that you're scraping and like that's a functional  boundary right for this kind of research if you   insist and this is why as well I think it's really  important to think about when the impacts are   on smaller cohorts or on smaller pieces of  research to think about how that could be scaled   up if people decided to be super strict in terms  of their interpretations because I think what's   most actually most core here is that people, as  researchers we're thinking about the implications   of the work that we do. What happens as a result  of that work both to individual people whose data   that we're using and for example that might  be that there is no impact on them other than   potentially that they could gain if the results  of the research lead to changes that are positive   but also thinking about flow-on effects and I  think this is so core in the tech industry as well   that so many things including social media itself  have had really significant impacts which were   not anticipated because people didn't look far  enough ahead right. And yes we can't predict the   future and and all of that kind of thing but there  are times when it's, when things like real name   policies that were introduced and then nobody went  oh wait there's a whole lot of people who obscure   their identities for really really important  reasons and maybe we shouldn't insist on seeing   a driver's license in order to verify somebody's  name like these are these are questions which,   that's like quite a small one, but we can think  about you know platform-wide changes even changes   of hands in terms of ownership and those kind  of structural issues as well to do least harm   and and not just break things right just  because it's possible but go okay well what   happens if this breaks and does that mean that our  society becomes more polarized, does it mean that   our relationships become more challenging, does  it mean that you know our education system shifts   in ways that we're not necessarily going  to see a benefit from and I think that   when we're dealing with these larger structural  issues that's when we need to think about those   flow on effects and I think that starts with  ethics at the beginning of the research for us   and I think it starts with ethics  with the companies as well. Rant over!

[00:54:15]Thank you Kim.

[00:54:17]We're getting close to the end of the hour  and so I think I might wrap up discussion   there unless either you have any final  comments that you would like to make.

[00:54:28]Thanks very much for the opportunity to  take part. Thank you for joining us today,   I think that was a really interesting discussion  I think you actually really both did a very very   good job of bringing together quite different  projects and seeing finding some commonalities   there and I appreciate that. Thank you very much  to all of our speakers, thanks to Matt for holding   down the fort behind the scenes and being there  ready to be poised in case it all falls over. Just   in winding up I'd just like to point out that our  next installment in our webinar series is already   promoted on our University Library website and  I will send you a link when I send you the video   afterwards. We're going to be looking at a few  different research projects that are using digital   technologies to explore musical instruments  including 3D scanning and digitization of historic   musical instruments both for preservation and  understanding the relationship between space and   sound and also looking at how we can use  3D printing to adapt instruments or build   completely new instruments that work with the  needs of professional musicians with disabilities   Really looking forward to that one it's going  to be Wednesday August the 24th at 10 a.m,   as I said I'll email all attendees details  on how to register. So thank you very much   to our audience your time today, to our  speakers for some really thought-provoking   presentations. Thank you Alexis. Yes thank you  I look forward to seeing you next time. Bye bye.