 So turning to Takas and Andres. So Andres Monroy Hernandez is a postdoc researcher at Microsoft Research and Takas Metoxys is a professor of computer science and founder of the media arts and science program at Wellesley College. And we have the privilege of having them speak to us about a paper that they have co-authored that they have been presenting on Narco Tweets and the use of social media in the Mexican drug war. And so we're excited to welcome them both. I believe that they're open to questions throughout the presentation. So please feel free to interrupt. We want to make this a really discussion-oriented conversation as they share our work with us and lead a discussion. And the hashtag, so if you want to tweet about it, is hashtag Narco Tweets. Welcome Takas and Andres. Thank you. Thanks. Well thanks everyone. We don't have to clap yet. So the topic of today is going to be this. We're analyzing how people are using social media in moments of crisis, in particular during war time. And we focus on the case of Mexico. So I just wanted to say that this is both work in progress and published work. So Takas and I are collaborators at Wellesley and Microsoft Research. We recently published this conference called ICWSM. But it's also part of this work is coming out in future venues as well. Yeah, two papers. So first of all, I just wanted to kind of set the tone of this conversation. And you know, there's been a lot of debate about the role of social media in promoting or supporting civic engagement. So you might remember when the Arab Spring was in full speed, you know, Clay Cherokee and Malcolm Gladwell got into this debate about, you know, what was the role of social media and so on. So our hope is that this could add to this discussion. So I'm just going to give an outline of our presentation. So first we're going to provide context of what is the phenomenon that we are observing. And then we're going to dive deeper into the different analyses that we did of the Twitter sphere in Mexico. So first, the Mexican drug war. So first of all, I just wanted to say, you know, that the drug war in Mexico is a slightly different concept than what we think of the drug war here. It's not just a policy. It's also an actual war. It's an armed conflict. So this conflict started around 2006. And some people, you know, some researchers have said that the drug war should not be considered a domestic insurgency that should be instead classified as a non-international armed conflict. So again, trying to set the context of this, which is actually, we're talking actually about an armed conflict happening in primarily urban areas in Mexico. So again, to give some context, the numbers of people who have died in the drug war are contested, as you can imagine. But these are some numbers that were released at the end of last year. So there are about 60,000 people that have been killed. There are about 230,000 people that have been displaced because of the violence. And you can see here in this chart, you know, how the violence started to pick up around 2006, which is when President Calderon took office. And, you know, he changed the policies, and then the way he was going to fight the drug war. As you can see, this is the rate of homicides per month, actually. So you can see how it kind of continues to go up and worrisome rate. Now, just to give you an idea of what this looks like, you know, crisis in Mexico, in many cities of Mexico, are part of everyday life. There are shootings, there are grenade attacks, there are car bombs, all sorts of things that are, you know, happening in people's neighborhoods, in the streets, et cetera. It's not just in some areas of the city. It's generalized in many different areas of the cities that have this problem. So again, crisis are part of everyday life. And now, you know, the war is not just a war of, you know, bullets. It's also a war of information. And, you know, there is a lot of tension around, you know, what kind of information can and should be released. Typically, when there are emergencies in a country like the U.S., you have situations like this, you know, if there is a shooting or if there is a natural disaster, there is a government there. And this is a role typically known as the public information officer. And these are people who are, you know, part of the government and part of the kind of emergency response team that tries to talk to the media. So we have these two kind of elements of the information flow during crisis in places that are not at work. However, in Mexico, as we see in this report on the Washington Post, journalists, the media is not able to do the job. And here I'm going to quote, it says, feeding for their lives and safety of their families, journalists are adhering to a near complete news blackout on the restrict orders of drug smuggling organizations and their enforces who dictate via daily telephone calls, emails and news releases what can and cannot be printed or air. So again, you know, the challenge here is that the journalists cannot do their job because they're being threatened by the cartels and sometimes even by the government. Similarly, the government itself also is a victim of this news blackout. And here I'm going to quote again, it says, the news blackout extends to government officials. And then they mentioned the case of Novolaredo. In the city of Novolaredo, the major mysteriously disappears for days and refuses to discuss drug violence. The military general who presides over the soldiers patrolling the city does not hold news conferences, issue statements or answer questions from the media. So this is kind of the environment in which this is happening. So if you think of the image that I showed before where you have the government and the media, you can see that both of these entities, both of these institutions are weakened by the war. Now, social media has played an important role in this information war, as we called it. And I'm going to describe why. First of all, internet penetration in Mexico has increased dramatically in the past few years. So now you can find internet and tacos at the same time. The number of users in Mexico grew from 17% of the population to 34% of the population in 2010. And now you can imagine it's a lot more. Now, in terms of social media use, 61% of those internet users are users of social media sites from Facebook, Twitter, to YouTube, etc. Of those people who use social media, about 20% are active participants on Twitter. And in fact, I just saw a report from the Oxford Internet Institute from last week that showed that Mexico is the fifth largest country in terms of Twitter production of knowledge. So that's the environment. And so when you have this intersection of weakened institutions, the media and the government, you have increased violence, and you have an increased adoption of social media, what you have is the following. We said to see tweets like this one. I'm going to translate this to Spanish, to English. So the tweet says caution on Lázaro Cardenas Avenue around El Paseo. People report a recent risky situation, which is a euphemism for a shooting or a grenade attack. And then they add these hashtags like empty why follow, empty why is the code for the airport of the city of Monterey, and then follow kind of an invitation to follow this hashtag. And then also risk Monterey, and then a timestamp that some people add to their tweets. So what we are seeing is that people are using Twitter as a way of alerting other people of which areas of the city to avoid what is happening in the city is because the media either is not reporting this, or when they do report it, they report it at such a slow pace and they wait for many hours or even days that it's no longer useful information. What a lot of people do, a lot of the people that I know in some of the cities in Mexico, for example, before they leave home or before they leave their office, they check on Twitter on which areas of the city to avoid. Kind of like a traffic report. So that's what we are seeing. And I just want to make sure that we understand that Twitter is just part of an ecology of different information systems. It's not the only one, it's the one we studied in this case, but there is YouTube, there is email, there is Facebook, and there is also traditional medium like the cell phone, even face-to-face communication, as well as the media, despite the fact that it's being threatened. There's still a lot of journalists who are doing their job or at least try to do their job. So just want to make sure that that's what we're studying, but it's also part of an ecology. Now what we did is we looked at four different cities in Mexico where this intersection of high violence, high internet penetration, and weak institutions happened to converge. So the first one is Monterey, the city of the size of the greater metropolitan area of Boston is four million people. Then there's the city of Sreinoza, which is a border city. Then the city of Veracruz, a city in the Gulf of Mexico, and the city of Saltillo, which is very close to Monterey. So just to give you an idea, the city of Boston, just the city of Boston is about 600,000 people. The greater Boston area is 4.5 million people. So just to set the dimensions in your head. This is where the cities are located. Number one is Monterey, two Saltillo, three Reinoza, and four Veracruz. So what we did is we looked at, first of all, we wanted to understand how many tweets are there. We've been hitting and seeing on Twitter that there are people doing this, but we wanted to get a sense of how many of these tweets there were. So we looked at all the tweets that were posted from August 2010 until November 2011, and there were about 600,000 tweets posted with particular hashtags. So each city has organically developed hashtags that are used to report the violence. So we look only at one of those hashtags per city, and this is the numbers that we got. There are more hashtags that emerge, and people compete for attention on different hashtags, but just to make the analysis simpler, we looked at one hashtag per city. So again, Monterey, Reinoza, Saltillo, and Veracruz. These are the hashtags, and the number of tweets you can see kind of corresponds to the size of the city. You can see also the number of unique users who are contributing to these communities of practice around the hashtag. Now we also wanted to know what kind of tweets we have. So if you use Twitter, there are different kinds of interactions. There is something called retweet, which is like spreading a tweet to the people that follow you. There is also this thing called mentions, which is kind of mentioning somebody else in a tweet, and there are like the novel or original tweets. So we wanted to see what is the rate of dissemination via retweets and interaction via the mentions. And one thing we saw is that for most of the cities, the retweets or the dissemination are the primary modes of participation on these hashtags, which if you look back at the tweet that I showed you, that tweet will be posted by someone who perhaps is in the scene or who heard from someone that something happened, and everybody else kind of retweets that, and they retweet a lot. Just as a point of comparison, although this is kind of almost comparing apples and oranges, we looked at tweets from the city of Seattle, for example, and the numbers reversed. So we didn't look at a hashtag. We looked at all the tweets posted by people that said that they're from Seattle, and there were about 50 percent of them were mentions, and 15 percent were retweets, so kind of the opposite type of interaction. Again, we need to do a little more analysis into looking at comparing something similar, which will be perhaps comparing the number of tweets in the cities against the number of tweets with the hashtag, but just to give you an idea, that's what we're seeing. Now, how often do people tweet? So what we did is that over the course of these 15 months that we looked at, we kind of tried to quantify the number of tweets per day, and this is what we found. So the top is Rainosa, then Monterrey, Saltillo, and Veracruz, and you can see that there are spikes, and the spikes are, they match with what we are seeing are major events. So for example, one of the biggest spikes in Rainosa corresponds to some clashes that killed 47 people in one single day. In Monterrey, the biggest spike corresponds to a massacre that happened in a casino where 52 people were killed in a fire that some drug cartels started. So you can see that this matches somewhat well to some of the big events. One of the issues with studying this phenomena is that it is really hard to find the ground truth. As I was saying, if you look at the media, you don't see the totality of what's happening. If you look at Twitter, well, you might also see some tweets that might not be true, so it's really hard to know what is true, but if you look at each one of these spikes and you look manually at the tweets that people are reporting, primarily they're reporting the same event, and then sometimes that matches to what you see on the media. One other thing that we see here is this kind of stagger adoption, and you can see that the volume of tweets starts shifting or starts moving in this direction, and this kind of corresponds to the way the violence is spread geographically in this city. So we actually saw that violence in Ragnosa, if you look at the official numbers, started before Monterey and so on. So you can see that this practice of using Twitter and using it in this way through hashtags and so on spreads as the violence spreads. Now, we wanted to see also what are people tweeting. So we took a snapshot of all the tweets from one of the cities, and we looked at just the typical world clouds where we wanted to see what are the most common words mentioned. So you see the biggest word there is the word balacera, which means shooting in Spanish. And then there are words like reportant, which is report, precaution, the tonacionas, which is like blasts, and then also names of parts of the city, like Garza, which is a common last name in the city, which also corresponds to a big street in the city of Monterey, and Guadalupe, which is one of the municipalities in the city. So you can see that there is a lot of words that refer to events as well as location of those events. So, again, kind of putting this into a different format, we kind of looked at the most common words, and we kind of looked at the frequency of those words. And you can see the order of these words, primarily what we see is descriptions of places, then references to violence shootings in particular, then references to some version of the world report in Spanish, and then references to people that we also see in this corpus of data. Now, one of the things that we also wanted to see is who is tweeting all this? Is it 10 people, or who are these people? So what we did, we looked at each one of these cities, and we mapped the number of tweets and the number of followers that each one of the people tweeting kind of reports. So each one of the dots is a person. So for example, if we look at this person, you will see that they have like 10 to the three, which is like a thousand tweets, and then 10 to the four, about like 10,000 followers. So you can see that this is basically what we try to identify is different groups of people. So these are people, for example, that tweet very little, but have a lot of followers. And this is important because it helps us identify different kind of groups of people. So again, this is the group of people who tweet very little, but have a lot of followers, and this is the group of people who kind of tweet an average number of tweets and don't have that many followers, the average citizen. And these are the people who are tweeting a lot and also have a lot of followers. So when we looked at trying to identify some of these people, for example, one of the accounts there is CNN in Spanish. So when they were reporting the casino massacre, they used the hashtag MTYFollow, and that kind of set was represented here. We grabbed the number of tweets at the end of our kind of computation, and also the number of followers at the end of the kind of the sample that we did. So it's possible that some of these people started with zero followers, and then at the end they had many followers. Then these are the average citizens and then what we call the curators. And basically what the curators are doing, they are spending a lot of time on Twitter. They're browsing through all the tweets that they read with the hashtag of their city. And when they recognize that multiple people are reporting the same thing, they broadcast this event to their followers. So these people gain a lot of reputation for doing this. And as you can see, they have a lot of followers. Some of them have like more than 100,000 followers and so on. So they're very influential or very powerful people in the tourist sphere in their city. They have a wide reach and also people kind of trust them through their their followers. So we try to identify some of these people, and it's really hard to talk to them. As you can imagine, the situation in Mexico is quite complicated and there's been some attacks on people who are on social media. So they're very careful of who they talk to and what they talk about. So we reached a couple of them and not many replied, but three of them replied and they accepted to be interviewed. I interviewed them over Skype. One of them over Skype, the other one over kind of a chat, live chat, and the other one through email. So I'm going to focus on two interviews that I did. One with Angela, this is a fake name, obviously, and the other one with Claudia, also a fake name, a fake photograph. And so basically, Angela, in this case, is a woman who, at least self-reports, is a woman who has 25,000 followers. She has stood it 35,000 times. She spends, according to her, about 15 hours per day on Twitter, which is quite a lot. She's in her early 20s and then Claudia, she has about 30,000 followers. She has stood it about 60,000 times, a lot more. And she, when I asked her, how much time do you spend on Twitter, she said many hours. She didn't say exactly much and she didn't want to say what her age was. One of the interesting things is that a lot of the people that are in this kind of curator category are women. I didn't kind of examine that in a lot of detail, but I thought it was interesting that a lot of them are at least self-reports as women. All the interviews were conducted in Spanish and I'm going to present to you the translations of this, so if there are any mistakes, it's my own mistake. So first of all, I asked them, how did you get started with Twitter? So Angela, for example, she said, it was through a friend and then she said, her friend said, you have to go to Twitter, it's so cool. She joined around 2009. For Claudia, I was a little different. She said, I joined by chance. I heard on the radio about how celebrities will interact with their fans. She also joined around 2009. You can see the power of Justin Bieber perhaps being placed here. Then I asked her, how would you describe your role? And Angela said, I'm a journalist. And then she says, it is as if I was a war correspondent on social networks of the war we are living in Mexico. I thought it was a kind of nice quote that she calls herself a war correspondent. For Claudia, she kind of insists a lot on her role is that of another citizen. She kind of tries to not be, or at least during the interview, tries to not be a protagonist. She says, I'm just another citizen. But also she says that people tell her that she's like their angel for looking after them. So a lot of her followers, they ask her, I'm living work. Is this street safe or not? Or they have this established relationship over the course of several weeks or months that they kind of interact. I also wanted to learn about their motivations. So the motivations kind of came out of a particular question, but out of the whole discussion that we had. So for Angela, for example, she says, I consider this as a community service, even though people might laugh about it. And she kept insisting on this, it's something really serious. But at the same time, maybe my friends are making fun of this, or even though you might not think it's serious enough, it is actually serious. And I think of it as a altruistic kind of participation. And for Claudia, she actually referred to the word altruism. She said, tweeting is an altruistic community service. So you can see the similarities in the responses. And then I also asked her, what are their sources? I really wanted to know where they get their information. So for Angela, she says, not all the information comes from Twitter. There is a lot of people who know what I do. They have my number and they call me, they're 100% citizens. And then she explained how people in the city know her, some people, and that she has contacts with people in the police department or the fire department, or friends who live across the city. So when she hears a report either on Twitter or that somebody calls her, she often tries to do some kind of triangulation. So she will call a friend of a friend whose grandmother lives by that neighborhood and says, are you hearing any blasts or anything? So that's kind of the way she tries to confirm that the information is true or not. For Claudia, she says, most of the information comes from Los Tuiteros, which is a common term people use in Mexico to refer to the Twitter sphere. And so she says, my followers are the ones who send me the information. And then interestingly enough, she says, in other cases, it's the reporters on TV or the local news. So you can see that despite the fact that yes, institutions are weak, the media is weak, they still play a role in this kind of information ecosystem. And a lot of the tweets and retweets that people send are links to articles or links to news reports. One of the things that actually I think this comes in later, one of the things that I asked them was also, what are your opinions on mainstream media? So Angela said, for example, for Claudia, she said, when I started, the news media, the journalists and the government were non-existent. They did not report, they did not inform what was happening on the streets. So that was kind of a complaint that I also heard from Angela. She says, the media, they forgot, they have an obligation with the people. They started hiding information. Then society started demanding information. And this is when social networks took cover. So we started seeing this similar articulation of why they are doing what they do and the way they see the mainstream media. Now I asked them also, what are your opinions on social media? So Angela says, social media is more pure. So for example, she says, when a piece of news gets to TV or newspaper, it gets mishandled. She often talked about how she sees events and people report things. And then by the time these events are reported in the media, they get changed so much that they're not truthful to what actually happened. So social media, she claims, is pure. Similarly, Claudia says, social media is very important because of the speed at which one can reach people and the veracity, and because it's altruistic. So again, kind of the purity of social media in a way, kind of expressing these quotes. So one of the things to note here is also that they do acknowledge that the media reports things as I was saying. But there are two challenges with the mainstream media. One is the speed and the other one is how truthful it is. Last, one of the things that I asked them, what will you change of this system? What will you change of Twitter or social media in general? What are the challenges that you see? So as a designer of technology, I was really interested in this question to see what we could do. So one of the things that Angela said is that she's worried about people stealing her tweets. So she says, it sounds silly, but it's just that I am against the stealing of tweets. As I thought it was interesting, because I've been doing a lot of work on intellectual property in a completely different setting. And I thought it was interesting to see this also happening in this space. For Claudia also, she says, I do not like that people often do not give credit in my retweets of my tweets. So from what they told me is that oftentimes somebody reads their tweets and then they kind of copy paste them and then put them with a different timestamp on the tweet, not on the kind of official timestamp. So oftentimes what people do is that they report some event and they add a timestamp on what time that event happened. So it's not necessarily when you're tweeting that, but when that event happened. And then apparently people sometimes change that to claim that they knew about this event earlier. So you can see that there is a lot of tension around the ownership of this information. And despite the fact that these people are altruistic and doing this for the greater good, they still want to get credit for their work. They still want to be recognized. There were two other issues around ethics and morality that they mentioned. The first one that Angela mentioned is that Twitter verifies accounts that harm society. So I asked her more about this and what she says is that she sees that a lot of the Twitter accounts that are verified by Twitter are from mainstream media, a lot of media that she recognizes as harmful for society and she gets upset that Twitter is endorsing this. Now if you ask Twitter, they will say that verification is not endorsement, it's just a way of avoiding duplicity or people claiming to be somebody else who they are not. But from the eyes of these people, at least this person, the verification is actually a seal of endorsement. Similarly for Claudia, she complains about the quality of the hashtags of the trending topics and she says that the ones that are promoted are often insulting. She says how some of them are demeaning to people or racists or whatever. And again, if you ask Twitter, they will say, you know, trending topics are completely an automated process. It's an algorithm that does this. However, these people are saying this is kind of some sort of endorsement. Now one of the interesting things about the curators is that, you know, these are just two of them, is that these phenomena started to be recognized in the community outside Twitter. So there is a group of people in one of the cities that we looked at that started a non-profit organization called the Center of Citizen Integration or Aggregation. And basically what they did is they tried to replicate what the curators are doing in a more official way. So they have offices, they have phone numbers, they are public about who they are and they basically are trying to aggregate all the information from Twitter, from even short messages to phone calls, et cetera, and put them on a map using the system called Ushahidi that was developed in Kenya. So this is a really interesting development where now kind of people are taking this kind of user innovation that they've seen on Twitter and kind of implementing it into a more formal process. And then we start seeing this spreading to other places as well. So now I'm going to let Takis explore and explain to you some of the analysis of the retweet networks of some of these curators. Thank you. Good afternoon, everybody. So I am a computer scientist and I knew about the drug war in Mexico, but I did not have a particular interest in diving into it and trying to find out what was going on. However, I have an interest in another issue that has to do with trustworthiness of information we receive. How do you know what you know? How do you verify the information you're getting? And I've done some work looking at the way that the discussion about politics in the United States is being shaped. So when Andres last year told me about the collection of tweets that he was getting from there, I jumped into it. I said, I would like to see it. But he says, well, they are anonymous, which made me even more curious. Can anonymous accounts actually get any trustworthiness, any traction? And why on earth would you ever trust somebody anonymously, in particular, when it comes to have your life depending on the information you receive? So that was kind of mind boggling. So we started analyzing. And so I had two problems. First, that was the collection that we got was huge, was in the hundreds of thousands of tweets. I don't want to read that many tweets. Number two, I don't speak Spanish. That was a serious problem, of course. So I thought, well, as a computer scientist, maybe we have some tools that we'd like to see. So I started looking at them as a computer scientist. What you see here is a graph. And this graph represents the mutual friendships between members of a particular group in a particular city that we will call Greenville. It was a, thankfully, very early decision that we made to use anonymity as well in everything we were doing that the question was that, yes, these accounts are anonymous that we're dealing with. But if at some point somebody finds out about some of those, they can be exposed, they can be in danger. So let's go with anonymous. So we did that. Now, this graph, interestingly enough, I did not do it in purpose to look like a human brain. It does. With five regions colored, I will explain there are three different parameters drawn here. The first parameter is the proximity of the nodes. How this algorithm works? Well, it is looking, as I said, that mutual friendships. On Twitter, how many of you are familiar with Twitter? So not everybody. In Twitter, it's not exactly like Facebook in which the moment you become a friend of somebody else, they become friend of you. But it is you can follow somebody and only if they follow you, you have kind of closer relationship. All right? So mutual following means a greater degree of friendship. So we looked at that particular city's Twitterers and see how much friendship mutual following that is was among them. Whenever you see two nodes close by means that they are connected by many other friends. And when you see them far apart, they are not connected by that many friends. So that was the proximity is one of the parameters. The second is the color. And the color is given by a different algorithm. This is an algorithm that tries to find sub regions which happen to have more often connections between them than outside of the sub region. For example, for all of us here, if we were to draw a similar graph, those of you who happen to work on the same project and you get to see everybody every day, you will eventually get the same color versus others who are working on different projects. So in this one, it seems like we have like five different at least sub groups connecting. Actually, we do not draw in this graph every friendship, but only those that they have at least 75 mutual friends. And the third parameter is the size of the node. A node has a larger size if it has more friends with others. So you can see here that in this green sub area, these nodes are having more friends than others. So again, you get this and you do not know anything yet about what they're talking, but you say, well, I can see communities in there. I can see that the communities are actually cooperating. They have this kind of friendships that you can do. To compare this against, let's say, the friendship you could find in the US of people tweeting about political issues. You will not see anything like that. You will actually see at least two different groups of people kind of tweeting. As a matter of fact, we have drawn that in another place and actually you can see, you know, the Democrats here, the Republicans here, all the liberals and conservatives. And if you were to color them within the conservatives, for example, you would see a subgroup which would be, say, the Tea Party. And all of this you could essentially extrapolate just by looking at the way they behave. So that was one interesting thing, made us look even further into that. The second thing that we looked at is the pattern of their communication, in particular their retweeting. For those of you not familiar with Twitter, retweeting is when you are transmitting the very same information without changing an iota. You just get a message and you send it to your followers, those who are listening to you, and this is not yours. This is an act of trust in the information you're receiving in some sense. Somehow you want to inform others that you are believing that this has some value. It's not happening 100% of the times, but most of the time that's what happens. Again, if you were to see the political system, you would see something similar. So here's what we see. There are some clusters far away from the others. These are clusters we found out when we looked closer to that, that they're referring to events like musicals or some kind of cultural event that people were talking about or, you know, famous personalities on TV and stuff like that. However, by and large, there is a very strong cluster in the middle and the size again means who is being retweeted. So think of that as the PageRank algorithm that works on Google. Whenever you do a search on Google, the first few things that will come up on your web page are likely the ones that Google believes are the most important. We are drawing the very same thing here and one of these nodes becomes particularly large, means that this node is being retweeted like crazy. And we have some smaller ones, 13 that they're kind of larger than the group, but some of them they're quite larger. This is the picture throughout the period we are observing. All right? A project that we're currently working is how the shape of this communication is shaped over time. We're always people retweeting these nodes all the time or not. So that's the kind of thing that we got. And again, we have not read anything of the tweets. However, these big nodes attracted our attention. You know, why on earth people tweet these nodes so much? All right? So we decided let's take a closer look at this node. Oh yes, I supposed to show you this when I was talking. So here is what I was talking about. All right? We started looking at this particular Twitter. We call this person Alex. We have no idea if it's a male or a female. We do not have an idea who's the real name of this person. We call this Alex. And so we show here in red the tweets of this very person and in blue the tweets of the community that they mentioned this person. So it's not surprising that the blue lines are much larger because, you know, the community essentially is referring to Alex much more often than Alex can, you know, do it by himself. Makes sense? And so a little investigation here is important. You know, you see a graph like that. What does it tell me? Well, first you see the spikes. You wonder why you have the spikes. That's a curiosity you need to dig deeper. However, you also find that there is a correspondence between the spikes of Alex and the spikes of the followers. So somehow the community is following or is in sync with Alex, whatever it's doing. However, there is another pattern that we notice that we find particularly interesting. If you look overall, how much Alex is talking about in this first period versus later on, you will see that Alex is becoming kind of more quiet down the road. He is talking less, which is weird. The sampling we have gotten is always with the same means. So it does not explain why at some point Alex starts talking less. His followers also start talking about Alex less as time goes. And interestingly enough, the separation between the two regions comes with a huge spike for Alex. So we look at this and we say, wait a minute, something, we're told something here. Something is going on because Alex changes attitude and the community changes attitude. And that happens after the community has a huge interest in Alex. All right? The deficit of not speaking Spanish, it gets more and more severe there. But we thought maybe it has to do with particular events that are happening. So in particular, as Andres said, balacera. This is about shooting. Let's see how what people are talking about corresponds to shooting events. Maybe that explains everything. And here is the interesting thing. We have divided the people who talk on Twitter in three groups. Those who will talk only rarely. They have all together less than five tweets. This is the first picture you see here in blue. And next to it, we have the tweets referred to balacera in purple. And an interesting observation right there between these two is that blue and purple are quite correlated. You know, when one goes up, the other goes up, when one goes down, the other goes down. So these people who rarely talk, by and large, they might be talking about shootings. That's interesting. Now, we go into the next group. This is the green group, which is the group that's talking a bit. These are people that they had five to a hundred tweets. And indeed we see that in the beginning, this group, the green group, follows what balacera is doing. But from a point later on, they're talking more and does not always follows the spikes in balacera. So this group actually is interested in something else after that. And to make things even more, you know, suspenseful, those who are talking a lot, they have a hundred or more, which is the blue, which you see them very low in the beginning, here are talking a lot. And the other observation is that the change, by the way, this deep is just meaning really that we don't have data for that particular date. It's not, if you were to just follow the lines, you would, you know, see that the transition is smooth. It's not like things are going suddenly dead and then going up. So the dates that we see here correspond to the dates that our friend Alex was talking before and after. So the whole thing becomes really darn interesting. And we realize that something is happening in the community around that time. Those talkative people talk even more. There is an interest in Alex. And after that, you know, the interest in Alex persists for a while, the interest of the community persists for a while, but Alex is not participating in that. This is where we give in and we start hiring students who speak Spanish and we start saying, okay, here is this time, you know, this 10, 12 tweets, tell us what they say and so on and so forth. So instead of trying to read everything, we just focus on those that they have something interesting to say. And here is what we find. Here is what we find. Here's the picture I showed you before. It turns out that in the beginning of the data collection we have, we find out that our curator's account is being hacked. It's one of the three accounts actually our curator is using. So this is what's happening right there. Why somebody would want to actually hack the account of that? It could happen accidentally, but from the tweets we realized it did not happen accidentally. We actually, beyond this data, we started going back and we find several repositories of tweets and we start sampling there. We find out that the tag that actually created all of this data was created about eight months before these events. And it happened from a random person, somebody who cared about the community. But Alex started getting, using a lot of these hashtags in a very effective way, that is people appreciated the way he was using it, and he becomes prominent. Suddenly he gets into the tens of thousands of followers. So people suddenly start following Alex. So we're thinking maybe there is some kind of animosity or jealousy towards Alex. We'll see what's going on. It turns out also that Alex is trying actually, is a rather idealistic person, because he's trying actually to create a larger event behind all of these things. He's trying actually to say, us Mexicans, as a big community, we have to do something about the narco war. Later on we find that actually the attacks of Alex do not end. There is a fake Alex that starts ridiculing this guy. And it's rather ridiculing. We find, however, that within a few weeks period, this account is being sat down by Twitter because the community starts reporting this fake account as spam to Twitter. And Twitter, you know, has this kind of automatic algorithms that if a lot of people complain about a particular behavior, then they will sat them down. So here you have a community actually rallying around Alex to try to shut down this fake account. Pretty interesting. Well, what happened in between? We were looking essentially at the eight top picks here to see what people were talking about instead of trying. You know, just going with a surgical accuracy and you say what they were talking that day, what they were talking about that other day. It turns out that halfway in between some major events were happening in Greenfield. A lot of slaughtering is happening and people are really getting really upset in between. So it turns out that Alex actually tries to do something about that and it's doing something really major. He's trying to organize in a very formal way a group of informants. We found actually a large number, you know, into the tens, the dozens of accounts that they were created on the call apparently of Alex trying to give information. So all of these explains the, you know, the role that Alex is playing, explains what he is doing, how he is doing it. There is however a major event that happens just before this spike. Alex is being accused as actually working for one of the cartels and you can just imagine the response of the society. In the beginning they would not want to think of that but after a while Alex stays quiet for reasons we do not know and eventually those who are retracting credibility for Alex are winning out. Yes? No. It was one of three accounts that was hacked and that has been resolved. I should have mentioned that this issue was resolved here. This account was abandoned at that time. So Alex had three accounts. Again I did not describe it. No, it is one account but we know that it is referred, Alex said, you know, that's my account has been hacked. So what it shows is an attempt by somebody or actually a collection of different people. We believe actually there are multiple players, they are trying to harm Alex because of the way they behave, because of the way they are giving the messages. So we are continuing as Andres said this is, you know, part of what we are doing but we already have developed a clear, a clear rare story about what we observe right here. At that time I think we should draw the summary and maybe we could get together and start talking about the different things that we learned from this experience. Yeah, let's try to play it. First of all, you know, just to sum up, what we see is an increase of violence combined with a weakened institutions combined with higher adoption of social media led to this kind of form of civic engagement on social media. And one thing that we also talked about was, you know, how citizens form these kind of alert networks. Almost like, reminds me of like birds, you know, like some birds are chirping and then the rest kind of follows. And then we also presented how these practices spread geographically as the violence starts spreading. And then, you know, how this represents a new element in this kind of information ecosystem. And then we also presented this phenomenon of what we call the curators, citizen curators who are aggregating information kind of like operators receiving lots of information and then broadcasting these alarms to everybody. And these curators are anonymous. They manage to get prominence and they manage also to get harmed. It's a very interesting play here that we are observing how, you know, an anonymous can come up to some prominence and be harmed because anonymity is not always helping you. Although I must say that for the interview, the people that I interviewed, one of them, it's actually not anonymous. She actually puts her name out and her photo. And when I asked her if she's worried or not, she said that she's not worried because what she's reporting is kind of like factual information. You know, things that if there is a shooting outside the street here, you know, lots of people will know it. So reporting that she considers is not a problem. While the other people are very anonymous and they try to be, you know, they try to hide their real information. But I bet it's not Alex. And then, you know, one of the other things that we wanted to put forward, the idea that these people, at least one of them, kind of phrased it this way, that they consider themselves as new work correspondence. They spend countless hours on Twitter working altruistically. But also, they want recognition for their work. They also challenge the assumption that technology is neutral, as we saw with the mentioning of the trending topics and Twitter verification accounts. And also, these carriers, you know, we can identify them just like the people living in the cities through the volume of contributions, as well as through the reputation signals that they accumulate via retweets or follow numbers. But this reputation, as Takis was saying, is fragile and, you know, can be challenged by others. So with that, that's it. I just want to thank some of our sponsors. Microsoft also has been a lot of our collaborators at Microsoft and Wellesley. And we have gotten a major fund from NSF. And I should mention also the great help of my colleague and Mustafa Ray at Wellesley who has been instrumental in doing a lot of this work. Thank you. So we're open for questions if we have time. To what extent do the cartels and their members participate in these social networks either openly or not openly? That's a great question. So they are very open on one particular website, which is YouTube. And I mean, with anything that I presented here that we talked about, it's really hard to know who is behind anything that you see online. What we see a lot, independent of the research, is that a lot of the cartels, for example, they kidnap one of their opponents and they interrogate them in front of the camera and then they kill them. But the interrogation is filmed and this video is posted on YouTube. Oftentimes, the video includes the beheading or the killing of the person. And obviously, YouTube terms of use don't allow that. So the video gets taken down. So oftentimes, these videos get reposted on this website called VK, it's like a Russian Facebook where there is apparently less moderation. So you can still see some of these. So I will say that the cartels are probably primarily trying to use social media as a way of broadcasting their message. It used to be early on in the drug war that they will leak some of these videos to the major news organizations in Mexico and they will play it. That no longer happens. So now they're using YouTube for this purpose. It's unclear on how much they are on things like Twitter. So there's a lot of debate and a lot of speculation, but that's been a lot harder to identify. There has been, however, some major accusation that they try very actively to infiltrate this kind of citizen networks. It's one of the accusations that happens from time to time in the tweets we read in which somebody has been accused that they work for this cartel and they work for the other. Within the community, I think also some cartels, within different communities, some cartels have kind of a better reputation than others. There is one particular cartel which is kind of the champion of hatred, I guess. They are being hated by everybody more. So it is, it's interesting how even the government of Calderon, in some reporting that was, I read a couple of weeks ago in New York, New Yorker, I think, was accused in quotes, accused that he was trying to attack other cartels, leaving one in charts so that eventually will be an easier target and others say no, they are actually collaborating and so on and so forth. These news actually also appear in the tweets. Yes, Mary. Thanks for the presentation. The question I had, it's incredibly provocative and persuasive for me to think of a community of actors who are really looking out for each other and then their curmudgeon in me wants to be a bit more cautious and tentative about calling them a community and particularly in terms of interpreting retweets as a reflection of community and notoriety within that community or recognition within that community. So could you just talk a little bit more about how you came to think of it so coherently or cohesively as community? Yeah, I can take that. So it is, the term community actually is a technical term when we are recognizing this kind of graphs and it comes from recognizing a behavior in a sub group which is more consistent within the group than outside. Given that all of these anonymous people, it's very unlikely that they know the vast majority of them in person plus we're talking about networks that they have 50, 60, 100,000 followers. It's unlikely, humanly unlikely that you know that many people right or even remember their faces. So it has to do more, we recognize communities more on the way they behave, that is when willingly they're using the same hashtag to report, if you want to report something you don't necessarily have to use a hashtag. You use a hashtag when you want to say, everybody who's following this hashtag, I want you to know what I'm saying and that's how it's through their behavior. It's not a community the same way that you're getting outside in Cambridge and you're thinking. In addition to that, I feel like at least to me one signal of identifying a community is when members of a group identify themselves with a certain name that everybody agrees on. So I feel like this term of Los Tuiteros to me in some way signals this sense of collective group that they try to or they feel like they're part of. I mean that's kind of not universally used but I feel like it kind of signals this sense of like we are part of something as a group outside the rest of the people who are not here. Last summer there was from my understanding anti-violence movement throughout Mexico that was triggered by social media and I'm wondering if you studied any of that in relationship to the Twitter sphere. I've been aware of that movement of course. I have not studied quantitatively or qualitatively as much as this one. That movement was primarily driven by this poet whose son was killed and was kidnapped and killed and so there was a big effort around this kind of call like basically people who had who had been victims of the war they came together and they formed this no more violence kind of movement. They went around the country you know talking to victims and so on. So I think social media played an important role in making this movement more visible but I think the kind of face-to-face interactions and the kind of you know the fact that they are part of the same victims of the same crimes they brought them together much more than social media will do but I think social media play a role in kind of making this a lot more visible in you know outside Twitter. So I don't know how much you can claim that social media caused that movement as much as you can say that the movement was helped by social media. Well so actually yeah so for example Alex the hitter that the tag is looked at this person started a website that is all about you know bringing Mexicans together against violence and so on. So there is a lot of rhetoric in all the tweets that we've seen of this idea of like not only reporting things but also complaining about violence oftentimes the complaints go you know people who are complaining about their governor or the president or sometimes people are saying things like you know we should stop waiting for the government to do something and we should organize ourselves and do something. It's really hard to you know to achieve much just on social media and actually one of the interesting things is that I think part of what we're seeing in the past few weeks in Mexico during the election this movement organized primarily by students on social media trying to fight what they perceive as an electoral fraud I think part of that kind of drives on the similar practices that we start seeing that we saw in these different cities around violence so they see kind of the usefulness of social media to bring their message to the forefront. So the the work's really extensive and there's a lot of good like mixed analysis and the qualitative interviews is great but there's one thing that I think is missing that would add a lot of validity to the research which is looking at geography and looking at it in sort of like maybe doing geocoding maybe looking at geocoded users to really get an idea of the extent to which these people are actually physically located in Mexico the extent to which they're located in these metro regions and and sort of you know assert that this is actually the case because you see Arab Spring stuff and you know it's mostly Western audiences looking at looking to them rather than like these people on the ground and same thing happened in Iran so. I actually started from there because I could tell and I could write a program that would determine the names of the streets in you know that particular city and then connecting with Google Maps I could check to see for example this street and districts actually meet you know is it is it a fictitious or it is the all of these accounts the by and large do not report location a few that reports give you location the middle of the Atlantic Ocean or Norway or you know the North Pole or things like that they do not you know have their geolocation on however I stopped doing that because I never found in the beginning any fictitious location in the cities we were seeing so it seems like they were reporting about the cities then very often they would also post a picture and then I could take a look at the picture and you know see who else was reporting when you have many people retweeting the same information about a location and about a city it's very unlikely that this is not there I will add to that there are a couple of signals that you can identify as kind of being that they represent that these people are actually in these places one against the language the locations as they mentioned like unless you live in the cities which I have a lot of familiarity with the cities I know that these locations actually exist but I agree that there's there's a need for a more systematic analysis of how much of these tweets how many of these tweets actually come from Mexico but from just kind of anecdotal evidence that we've collected through reading tons and tons of tweets it does seem like the majority are actually from Mexico and one of the challenges that is mentioned is that it's really hard to find location on tweets so first of all the number of geo located tweets overall on Twitter is like less than 0.1% it's like really really small the other thing is that the again the location that people set on their profile oftentimes is not the real location or sometimes they use a different version of that location so for example Justin Bieber's heart is one of the most popular locations on Twitter so it's really hard to identify that but you know but not for our group not for our group is not reporting we were looking we were looking at the case of T1 at one point and getting uh upwards of that but we can talk about that later we I mean if you have some suggestions would love to talk about it yeah I have a question regarding anonymous sources and I just said anonymous sources in this situation became very influential influential and had a really high impact but we are aware of many concerns regarding anonymous sources so there could be many problems with with them so do you have any conclusion regarding the regulation of anonymous sources in general in this situation and how do we keep the quality of anonymous sources and can you see any perspective of views of anonymous sources in future can we believe that we will have anonymous opinion makers in future based on this example yeah I can take that so um for the anonymous sources it's different when you're looking at your own Twitter account and you receive an email a tweet from some anonymous person and different when hundreds of thousands of people or maybe you know thousands of people actually look at the same thing and they decide to rebroadcast this information the chances are that the more you throw this kind of crowdsourcing principle on the problem the more human eyes look at this information and decide to broadcast it or not the more likely you are that there is some validity with information you are getting the person who's giving the information the anonymous source itself slowly gains reputation it's not that if you you know start an account right now and you start tweeting about say narco wars in you know pretending you are in mexico you will have an audience right away it's very unlikely that you will have very quickly somehow the community and i'm using it you know in the previous term has to accept that you have something valuable to say and report your information so one of the things that we found is that we did not we we find things about confirmations and non confirmations you know somebody says yeah i confirmed that there is this thing happening right there and i do not confirm we do not find too many of the non confirmed in there so for cases like that you will never be a hundred percent sure in in a few cases as i said we try to verify as much as possible with the surrounding information with the surrounding you know location or description or the picture stuff like that but it seems that when thousands of people think that this is worthy broadcasting and they're not robots it's not like somebody wrote a program to just broadcast like that i have many reasons to believe that then you have some validity that's my short answer but i i'm going back to the question of regulation this has been a very debated topic in mexico about this so one of the states the state of veracruz which is one of the cities that we looked at um there was a case where people on social media were claiming that kids were being kidnapped from schools at the end it turned out to be not true or at least seems like most most things indicated was not true but the news spread on social media really quickly and so what happened is that the parents when they read this they left work early and then they went to pick up their kids and so on so the government was really upset and primarily the governor actually has a twitter account and despite the fact that he was tweeting saying this is not true people didn't believe him partly because he hadn't built a reputation through a long period of time where he engaged in this kind of practice of tweeting and being part of this kind of community of practice um part of that made it kind of not trust a not trusted source despite the fact that he's the governor he was very upset about it as you can imagine and and what they did is that they went after some of these accounts so they identified like 17 accounts that apparently were retweeting or tweeting about this uh that were the source and they send them to jail on the charges of terrorism and not only that but they kind of changed the law so that you know it will be uh illegal to spread misinformation on social media and again in this environment where you don't know what's misinformation and what's not misinformation it's really complicated so uh in general I say I will say that you know the approach of looking at regulation or using the law as a way to to kind of uh control the flow of information here doesn't seem like the best approach and this is compared to other cities where the government has taken a more active approach and they have a long-standing you know engagement with this uh twitter sphere uh where their tweets and retweets are considered a lot more trusted um so I guess just in terms of regulation I will I will kind of uh raise the concern that that doesn't necessarily it will work there is another thing that we did not talk about at all though it it's because you are still looking at it and it's it's also important for your question it has to do with the fact that let's imagine that you are actually working for a cartel and you are um you know having your own tweet account and you can follow by seeing what people are reporting you could get information about where it is safe for you also and your cartel members to move and where not to move so you could have this kind of dual purpose because citizens are using it to save themselves and cartels are using it to actually uh find out about the movements of their opponents or the police this is one you know heavily debated argument in this tweets environment and what we find at the end you remember the the brain-like structure I showed you one of the reasons that you have this kind of connections is that these groups don't exactly uh agree or trust completely the other friends in the other group you know you may have some common connections but they don't and from time to time they start spreading rumors about the other group we do not know at the end who is telling the truth all we know is that they're not on the same board and this is you know the thing that if we could figure out if you could start you know binding you know finding the ground truth for some of these groups then you might be able to find for the rest. Hi so last year uh it was it became public that there were some threats to the people doing narco blogs in Mexico and they became public I guess mostly when the few narco blogs still covering these things or the few mainstream media covering they started posting the pictures of these couple of bloggers that were killed and put on a bridge because of their engagement on blogs uh but looking at it I mean just from the inside on Twitter I haven't never never found like threats posted at least publicly to people like these curators or people who are influential in on Twitter and I was wondering through your analysis or even through the through the interviews done with with these curators like did any information on threats they have received appear? So I think there are two issues here one is that the kind of things that we are looking at are reports of events that are public the the claims that people have been killed for tweeting or for using social media have actually been related to the use of other websites not not Twitter in particular to pinpoint the hidden locations of the cartels so I just I'm just trying to make sure that we separate those two practices and the one that we're looking at is the practice of reporting events and violence and so on and that's a lot less uh in some ways dangerous than actually reporting where the cartels are hiding um and that actually is not as active on Twitter so the fact that you haven't seen it on Twitter as much is because that doesn't happen as much on Twitter as it happens on other sites so that particular case in Valera where they killed two bloggers was because they actually created a separate website where they kind of had a Google mash-up of the Google Maps API and they kind of helped people pinpoint the location of the cartels and the claim is that some of these people were killed and but even that is being contested you know there were two bodies found next to the bodies there was a sign that said you know these people are killed for what they're doing online uh some people say that perhaps they they were not actually the people involved in the blog that they were just random people that they killed and they were trying to kind of demonstrate their power so it's very murky and it's really hard to know what's true or what's not true um but in general I will kind of separate these two phenomena the one reporting you know violence that is public and the other one reporting the hidden kind of information and that's something that I'm personally not as interested in looking at just because it's a lot more dangerous and more uh you know it's a lot harder to to analyze because it's just a lot more hidden yeah I agree thank you for your really interesting presentation on your really interesting research and I know you said that your research doesn't cover this but I was wondering if you might consider talking a little bit about what you saw as the people who are curators presenting themselves as female I don't know how news operates in in Mexico but I know in the US it's pretty uncommon to have investigative journalists be female particularly war correspondents be female but females have historically been people concerned with kinship and and civic and welfare of community so I'm wondering if you could kind of talk to have social media might be involved yeah actually when I asked one of the curators why do you think that a lot of these people are women and she said exactly what you were saying that you know women are more uh civic or more you know pro-social in some ways that's what she said um and when I asked this same question to other people so they said well they claim to be women just to gain status and trust but they might not be women so I think that is this play with gender there that is really interesting and that I think will be really interesting to dive more deeper into that yeah we're thinking exactly the same thing because the overwhelming kind of fake names that you see there are women's names and you wonder are they really women but I guess one reason is that women naturally might care more about that the other is that if you're a man you want to muddle the waters you say you're a woman well thank you very much thanks