 Good afternoon, everyone. I'm Hassan Daulju, professor of computer science from Arizona State University. Is it better now? Okay, so this is another 15 minutes of follow-up to what cleaned what's presented to you earlier, but maybe in an international context. So the title of my talk is how can we analyze big data to understand while in text-dremism? So we developed this tool at ASU during the last decade called Looking Glass, which is a culturally and linguistically sensitive social media monitor. It basically finds radical and counter-radical hotspots. It shows breaking news, viral messages, shifts in key sentiments, bots, groups, and their ideologies, as well as their geographic footprints. We have deployed this tool in three continents, Southeast Asia, Indonesia, Malaysia, West Africa, Nigeria, Nigeria, Senegal, Western Europe, as well as recently in Latvia and Ukraine. Because we have this diverse footprint, we were very fortunate to work with a multi-country, disciplinary team of researchers with us, and we were able to tap into multiple methods to develop our technologies. In a nutshell, the way Looking Glass works is it's basically a machine learning system, and it learns its cultural models by looking at the media outlets of the groups on the ground. Real groups, real political parties, real NGOs, journalists, people like that. And then once these cultural models are learned, then we deploy them on social media, whether it is Telegram, Twitter, Instagram, YouTube, Facebook, whatever it is, and whichever language it is. And once we deploy the model on the social media, basically, we get to see the leaders and followers and how they are trending online. So my first case is how Russia manipulates ethnic Russians in neighboring countries. So what you see there is a Looking Glass where the line chart is showing the volumes of the underlying tweets. The pie charts are trained to show the sentiments. And in this instance, we are looking at three key sentiments, pro or anti-NATO, pro or anti-EU, or pro or anti-Russian. And the court diagram, that colorful big circle, is basically monitoring the political ideologies. So the color circles are trained from the local political parties and their outlets. So if we click on, for example, anti-NATO and separatist bands on the pie chart and the court diagram, Looking Glass will show us the relevant actors and messages. It separates the bots from people. It shows us the provocateurs that are at central locations. It shows the social networks who are following them, as well as the sources of this information coming out of these individuals. So when we analyze the findings, we are able to see the Russian propaganda sources and when we look at the viral messages, we are able to see the teams that the Russian propaganda employs. So these teams are basically revert of fascism in the governments, bonds on the use of the Russian language, violations on the rights of the ethnic Russians, assault on Soviet history, as well as integration of Crimea into Russia. So when we tell Looking Glass to pay attention to these teams, in the news articles coming from those propaganda sources, it gives us a propaganda signal. So if you look at this signal, I'm not the one moving the cursor, if you look at the signal, you can see that it basically breaks out, begins breaking out at what we call phase one of the Ukraine crisis. So these are news articles that mention Ukraine coming out from those propaganda sources and we are monitoring the density of those teams. That's what this signal is. So that first spike breaks out with the success of the Euro Maidan and the Ukrainian Revolution with the collapse of the pro-Russian government. You can see it breaks out there. And then it spikes all the way until basically the annexation of Crimea. Then it falls down and there's a second spike, which is basically the Russian propaganda in the southeastern regions of Ukraine, which basically picks with the declaration of the Donetsk and Luhansk Pupils Republics. And then as the signal collapses and kisses the zero, basically you have the Russia's humanitarian convoy and the full-scale invasion in Ukraine. So basically spikes in propaganda are early warning alerts for malicious intent. That's what we can tell. Next case is Libya. Basically, this is Facebook. This time we collected data from hundreds of civil society and political Facebook groups. Internet penetration in Libya is 43 percent. Facebook popularity is 93 percent. There are groups with millions of followers on Facebook. What we observed every time while volume spikes on Facebook, it points to a key event. So we are going to use this scenario to show how looking glass can be used to assess local attitudes towards US policy. Basically, the US air strikes in Sirte. So this is a very complicated chart. I'm so sorry about it. It looks like a reverse that are crisscrossing all over the place. But I want to give you guys a I want to point to a point there, which is marked by the green triangle, which is basically when the US bombs Sirte. So this is measuring sentiment for the government of National Accord, which is an interim government for Libya that was formed under the terms of the Libyan political agreement, which was signed on 17 December 2015. It's the United Nations government that we are trying to establish in Libya. Now there are two other governments in Libya competing with us. One is on the east, the secular Haftar dignity elite government. And the other one is on the west, the Muslim brothers, the Islamist. So everybody is fighting with everybody else in Libya. But basically what this chart shows us when we bomb ISIS in Sirte on August 2nd. Basically you can see there are communities that are flashing pro-government of National Accord for the first time. The orange and the red ones are our opposition, the Muslim brothers and the Haftar and al-Qaeda. So our data shows a surge in support for the GNA post US bombing. This shows how Looking Glass helps assess local support and our criticism of US policies. And their impact. This one is ISIS. I just want to, I have to go through quickly over this one. So basically ISIS conquers Mosul, they behead James Folly, they burn the Jordanian pilot, and they claim responsibility for the Paris terror attacks. Now what was happening in the social media as these events unfold? The key message I want to give you guys here. One in six was pro-ISIS, producing one in four of all the messages. The beginning configuration is a split with majority pro-ISIS and minority anti-ISIS. Basically these are Arabic tweets that mention ISIS. But following the Jordanian pilots burning, you see that huge green community. Which is basically a large contingent of anti-ISIS people showing up. And what do they do? During the next four months, you look at the ending configuration, the communities are fragmented and penetrated and majority anti-ISIS. So basically those people showed up after the burning of the pilot and they decimated the propaganda online. So this shows how Looking Glass can identify shifts in global sentiments on key issues and actors and their impact. Key takeaways, we can successfully detect and measure foreign influence teams, targets and levels, such as the Russian playbook. And we can use ideologies and sentiments to monitor the effects of US and adversaries policy decisions and actions, such as US in Libya or ISIS in Syria. Now I'm going to leave this talk with an emerging provocative idea. Social media, only 2% of messages goes viral and that's the part that we tend to pay attention to. Our research suggests that social media messages that goes viral and make a specific claim about concrete acts in the future, such as an assassination or a military action, turn out to be true. That is, social media can be analyzed in a manner that actually predicts events. And I'm going to show you two examples and leave you with those examples. So this is next week following the US bombing. We see a community that is comprised of Muslim brothers and pro-GNA people. Basically they are saying on August 12, 2016, the war is now about who is going to govern the oil. And they make a prediction. Basically they say this is from the National Oil Corporation of Libya. They call rival groups to protect the oil ports of the Raslanov, Sidra and Zutina as clashes are expected between the National Army of Haftar, the National Army of Haftar, that's the Dignity Allied, the Red Circle and the pro-GNA affiliated petroleum facilities guard. This was one month before that Financial Times report that occurred on September 11, which basically says, Libyan rebels loyal to Khalifa Haftar says key oil facilities. So there you go. If you were paying attention, we know this was coming. Next one is a little bit more controversial. So this is Al-Qaeda group number five. It's that little number five community, which is Al-Qaeda at that point, one week later. So basically they are predicting political unrest in Turkey, assassinations among army and government officials, causing Turkey to spin into instability and unrest. So this was predicted on August 22. I shouldn't say predicted, I should say remored. It was remored. Three days later, we got that news. Turkish opposition leader escapes assassination attempt. So viral remorse, if we pay attention and if they are specific, can be predictive. So that can have applications in threat casting and other things like that. So that's all.