 Hi, I'm Laura Dugan. I'm a professor of criminology and criminal justice from the University of Maryland. And I'm here to talk to you about methodologies and media encountering violent extremist research. So before you zone out, it is TED. So it's only 10 minutes, and I think it will be compelling for you to think about what are the things that go into the research that we do and what makes good research. I've been studying violence and public policy and terrorism for more than two decades now. And one of the key things that's important in really understanding this dynamic and what works and what doesn't work in reducing violence and possibly even stopping it is we need to have data that measures violence, like systematically and comprehensively as possible. We also need data that measures what we do in order to stop violence. And so how do we find these sorts of data? Now, coming from criminology or being a crime researcher, criminologists get their data on crime from three major sources, official data from police departments, also from self-reported data from perpetrators, and also surveys from victims. Now, none of these are going to provide us a comprehensive source of violent extremism incidents. They just aren't. Particularly from a global scale, because the scale is just too big and it crosses national boundaries and we don't have a central agency to collect official reports. Despite the problem being severe, the events are still rather rare, making it difficult to collect information from either perpetrators or from victims. And so this is when we turn to media. Here's the thing. Terrorists like attention. Criminals don't like attention. Terrorists like attention because they want to be part of the public narrative and they want to shape public opinion. The other thing is that media likes to report on bad things. And so what we have is a synergistic relationship between violent extremists and the media. And this presents an opportunity that scholars have taken advantage of in order to chronicle events of violent extremism. And so I'm going to talk a little bit about the Global Terrorism Database, which is the most comprehensive database that's available on terrorist incidents across the globe. And that particular database, I'm one of the principal investigators that collected those data. And it began a long time ago with the Pinkerton Global Intelligence Services had ex-military personnel scanning newspapers looking for incidents of terrorism. And when they found them, they wrote them down on index cards. In fact, these are two of the index cards that they used. These index cards were basically how they were recording terrorist incidents from the 1970s, the 1980s, and the 1990s. The boxes that are in the upper right-hand corner are what the Pinkerton gave to Gary Lafri and I back in 2002, 2003, and they capture all of the global incidents that occurred from 1970 to 1997, except for 1993, which was somehow not in the boxes. We don't know where those were. But the key here is that this was a really important opportunity, because at that time, there was no source of global terrorism that included both domestic and international attacks. Now, what you may not have picked up on is that Gary Lafri and I got these data in 2002, 2003. The last incident that occurred in those boxes was in 1997. And so we were already lagging behind for five or six years. And so if we had the incentive, which we did, to continue the collection up through the present, we were already behind, which has implications. And so throughout the years, the global terrorism database, which is what those boxes became, relied on whatever technology was available at the time in order to pull the information together. Now, this is what's done today. And I'll get back to what we did before. This is what's done today. And basically, we rely on news aggregators to pull in information, the technology used in machine learning tools, that is based on input from analysts that either accepted or rejected a news story as terrorists, so that more and more stories come together that are actually terrorist attacks. Now, this is the front page of the global terrorism database that's housed at the Start Center where Bill directs. And according to the definition of terrorism that is used by the global terrorism database and that meets the criteria of that database, this is what global terrorism looks like today. And so these are the trends that you've heard referenced to from the speakers up to today. Now, I'll come back to this graph, but I also want to talk about what governments do. Because part of the policy equation is what is it that can be done in order to reduce terrorism? And as the speakers have made hint to, just focusing on countering terrorism and reacting to terrorism isn't enough. And so in order to collect information on what governments have done to address terrorism, we need to look at a broader perspective. And so this chart shows you how a colleague of mine, Erica Chenoweth, and I have looked at government actions in a counter-terrorism environment or in a terrorism environment. And we are interested in looking at government actions that span from fully repressive to fully conciliatory. We're interested in looking at whether these actions affect only the perpetrators, are they discriminant, or do they involve bystanders or the constituencies for whom the terrorists fight. We are interested in looking at material actions, like raiding a town or releasing prisoners. And we're also interested in looking at non-material actions like praising a leader or making bold statements and things that are more rhetorical or threatening or even empowering. And so how do we get data like this? Once again, once again, there are no public sources of information on this. We can't look at those who are doing the actions. We can't take surveys on those who are targeted by the actions, and so therefore we turn to media. And this slide is not for you to look at the details of, but it does give you an idea of how we've gone about collecting data from news sources by downloading information, using machine technology to identify events, and then rely pretty heavily on human coding to get a sense of what is happening here. To give you an example of some of the information that's captured in what we call GATE, which is government actions and terror environments, we are able to capture the Israeli government allowing utilities to be installed in Palestinian refugee camps, which may not look like counter-terrorism, but it's relevant to the terrorist threat. We're able to capture things like leaders threatening, government leaders threatening somebody who's a leader within a constituency movement that also has a terrorist component to it. We're able to capture things that are going through the policy process that's going to affect constituencies in different ways. And so it's a move in the right direction. However, we're now working to update it, but it's a little slower. It's not quite as technically advanced as the efforts to collect the global terrorism database, although we're working with computational linguistic scholars to find ways to more efficiently collect these data and to preserve the accuracy of the data. Here's what the two data sets, GTD and GATE, look like when they are combined. The red bars, these are just crude measurements, but they show repressive actions by the governments that are listed there. The blue bars show conciliatory actions, and these are directed toward constituencies of terrorist organizations or their organizations themselves. And the line represents terrorist attacks, just number of attacks in those countries according to the global terrorism database. And what's important by this slide is for you to recognize that they track pretty well. Now, what's beneath these data, so I have a minute left, and this is what I'm here to talk about. But I can say it all pretty quickly. Because they rely on media sources, we suffer from the same biases that all media gives us. If it bleeds, it leads. We are able to capture the most compelling stories, but the less compelling stories may not be able to be captured in these data. As I mentioned before, the GTD was pretty much tied to the technology that was available at the time. And so we have a mixture of prospective and retrospective collecting. We have a mixture of human coding versus machine coding, and that all has implications to analysis. Do we really think that North Korea had only one terrorist attack in 48 years? So media collection is tied to freedom of press. My last slide is this, because it shows the combination of these two. We have problems with using just media, but does that mean that we don't do it? No, it doesn't. But it does mean that we need to understand the biases that are inherent in media. And when you're analyzing these data, you need to take into account those things. And I'm not going to get into modeling strategies, but what I urge you to do is read the documentation on how things are collected in order to properly analyze it and not just take the data as truth in order to be able to assess what works and what doesn't work. So thank you.