With the UK's terrorism threat level being raised from substantial to SEVERE, following the 14 November incident in Liverpool, we are revisiting our podcast with Jerry Smith from CHC Global.

• https://www.gov.uk/government/news/uk-terrorism-threat-level-raised-to-severe
• ---

Ever wondered how we measure the amount of terrorism? We talk to Jerry Smith from CHC Global, who explains how the Global Terrorism Database has been keeping track of terrorist events all around the world for 50 years.

This podcast covers the painstaking process of collecting structured information about terrorism and considers how you standardise the measurement of something so chaotic and destructive. It explores how the Global Terrorism Database goes about collecting information, categorising events and maintaining data quality to create one of our favourite datasets.

A few things we mentioned in this podcast:

• The Global Terrorism Database www.start.umd.edu/data-tools/globa…ism-database-gtd
• CHC Global
• Analysis of terrorist event frequency

For more information on Aleph Insights visit our website https://alephinsights.com or to get in touch about our podcast email [email protected]

Hello and welcome to the Cognitive Engineering Podcast, produced by me, Fraser McGruer for Aleph Insights. In this series of podcasts, we take a look at interesting topics and discuss what we think they tell us about analysis and decision making. I'm here with Nick Hare and Peter Coghill of Aleph Insights and Gerry Smith of CHC Global. And this week we're discussing the Global Terrorism Database. And unusually, because of another lockdown, we are recording this remotely. Nick, perhaps you could introduce us to Gerry and tell us a bit about him and how you know one another.

I'm intrigued as to what the Honourable Artillery Company is. I think that's what you said. But anyway, we can come back to that later. Gerry, can you tell us a little bit more about yourself, give us a bit of your history and background?

Yeah, sure. So, yeah, as Nick mentioned, I met him at the HAC, the Honourable Artillery Company, which is a TA unit in the centre of London. And yeah, we were having dinner. Anyway, so my background is a little eclectic. I'm a lot older than you lot, I'm the grey haired one in this podcast. But I was originally in the British military in bomb disposal in the 90s. I then went into commercial landmine clearance. And then from that, bizarrely, it took me to DSTL Porton Downs, so the chemical, biological, radiological defence establishment, the UK one. It's where we keep the flying saucers as well. We don't have an Area 51 in the UK, so we keep the flying saucers there. And then from that, I ended up in the UN system working for the Organisation of Prohibition of Chemical Weapons. So you can see there's a bit of a thread there. And then four or five years ago, I came back from living overseas for a good chunk of time. And started doing consulting around risk management, primarily in the perils of chemical, biological, radiological, nuclear and explosive effects. Ended up working within the insurance sector. And a couple of years ago, became a partner in a small consulting company and broken company that is right in that sweet spot of what they call special risks. So that's terrorism, malicious actors, kidnap for ransom, that sort of thing.

So does that mean that, so is that through Lloyd's? Is that how that works or is it independent of that?

Well, Lloyd's is a market. So, you know, it's a collection of where people go to for looking at some sort of coverage. Lloyd's is sort of renowned as trying to be able to offer coverage for pretty much any peril that somebody comes with. So Lloyd's is one of the markets. But we work with other people as well, globally.

Great introduction. Thanks very much, Gerry. So over to you, Nick or Peter, what do we want to get from Gerry here? What do we want to try and achieve in this podcast? What do we want to know about the GTD?

Yeah. So what's the official sort of, what does the GTD consider to be a terrorist attack?

Yeah. So the GTD's got three criteria that must be kind of present. The incident must be intentional. So it must be malicious. There's got to be either violence or the threat of violence. And the incident has got to be perpetrated by a sub-national actor, i.e. not a government. And again, this might be something we want to talk about a bit later, because it is a bit weird when we look at definitions of terrorism now, whether a government can conduct a terrorist act at all. But the GTD, you know, they've stuck to their knitting. This is the definition that they pushed out. And they run with that. And they've also got.

Just to confirm that, so this would imply, for example, if that if ISIS, you know, were recognised as an official government for whatever reason in an unusual parallel universe, that they would no longer be capable of terrorism. It would just be normal state based warfare or something, according to the GTD.

Yeah. And that would potentially then fall outside their definition and therefore their any attacks that they conducted, if it was confirmed that it was them, then that potentially could then fall outside the GTD.

Yeah, that's interesting. So is I mean, I think the key thing there is that it really it isn't to do with the nature of the act, which is what I think a lot of people might instinctively think is terrorism is defined by a certain kind of activity and not necessarily by the actors. But actually, I think, you know, probably that's a much harder thing to define to separate it from warfare. But their definition, presumably, I mean, I mean, presumably that's pretty it's pretty clear. Is it is that because it's fairly easy to measure that whether or not these criteria are true?

Yeah, well, I think the issue is, is that, you know, they they they needed to have a definition. They need to have, you know, what's going to be in the box and what's going to be out. And particularly as an academic approach, we've quite rightly so is that, you know, you've got to have that those kind of criteria. Now, they might shift and you make a conscious decision to change aspects, but they felt that this is this provides consistency. I mean, they've got some criteria as well that they add into that. So as well as those sort of three key elements, they've got criteria, which the broad one is around aim about attaining political, economic, religious or social goal of some sort. One is around intention. So the intention to sort of coerce and intimidate. And the third one's got to be illegal as well. So and I don't mean in kind of domestic law, but in kind of international terms as well. So because, of course, bizarrely in the world, we have legal forms of warfare, legal ways of killing each other, which seems a bit odd. But that's the way international relations has evolved over the last century and a half. So there is a series of elements that must be included. And then criteria, two of those three criteria must be included as well. And then they have this really neat thing as well called the doubt terrorism proper. And what that means is that even though you might meet all those criteria, there is something about the attack that might also fall into some other definition. So what they call a definition overlap, whether that's insurgency or hate crime, organized crime. So you've got this kind of ability to even still filter out. So what they try and do is, you know, not have so much of the rubbish or they have all the rubbish on the cutting room floor, but the kind of credible events they want to try and keep in there. And what you can do, because it is a database, you can just turn off, you know, the doubt terrorism and proper switch if you want and reduce your kind of field to more narrow definition.

Yeah. And of course, it's a really tough, tough one, because, you know, the nature of terrorism has shifted. If you look at sort of, you know, 20, 30 years ago, again, in OECD nations, they were looking at big vehicle bombs blowing up, trying to hit economic targets. You look at it in the last sort of five years or so, and there's been the shift to sort of low technology, blades as weapons, you know, marauding knife attack or using a vehicle as a weapon, lower tech stuff and just sort of indiscriminately killing people. And so, you know, as those shifts occur, and particularly as you start to get, you know, the Internet, of course, starts to blur definitions as well of what's a state actor or a non-state actor. And then you've got, you know, a state supported actor or a state authorised actor. You know, the greying of those situations mean that, you know, straight definitions may not hold water. But I think what the GTD is trying to do is maintain that structure. So you do get that consistency. But it'll always throw up some controversies. Peter, what would you like to ask Gerry?

So, so, Gerry, so presumably, the people that run the GTD that you've mentioned already, but they go to great pains to make sure that there's a sort of standard approach and a standard methodology that goes way back to the 70s. Presumably, that's quite a sort of organisational procedural challenges. You know, people have their career, retire, new people come on board, the team changes, grows, whatever. Are there any kind of like key lessons that you've kind of spotted from running, from being involved with running the GTD about how they go about doing that?

Yeah, so I think, I mean, we're incredibly lucky to have hooked up with these guys. You know, this relatively small, limey company, the other side of the pond. But what we've certainly recognised with them is that the consistency and the approach that they've carried on. And I mean, the GTD costs a couple of million dollars a year to run. And it has historically been supported by the US government and their intent has been for academics then to use it for research. And what we've done over the last year or so is support them as the US funding has been, the US government funding has been slightly more variable. We've sort of come on board to sort of commercialise a little more. But the process they actually go through is quite interesting. So they've got, I wouldn't say really advanced machine learning and artificial intelligence, but essentially they're scraping the internet every day for around two million articles using kind of keywords. That search ends up looking at around about 200,000 articles a month, of which around 70% are potentially relevant to terrorist attacks. And then they will, we will then look at how they can populate their field. So they've got about 120 odd fields and, you know, they're from the very basic, the lat, long, you know, the location of the event itself. The type of attack, as Nick mentioned, you know, the weapon used, the suspected perpetrator, perhaps the number of fatalities, number of is injured. They might talk about damage. So there's a whole set of criteria. But as you start to get into the more detail and nuance, as you can imagine, when you're effectively using the internet as your source, when you get more into the detail and the nuance, that tends to fall away from things like popular news sites. And of course, into the mix, you've got to throw in, and this is probably one of the reasons why AI really isn't quite mature enough to handle this yet, things like social media. And it goes from sort of basic things. So, you know, you don't want to pick up a website that talks about shooting, because in fact, you're talking about soccer or basketball. You don't want to, you know, you don't want to talk about an explosive, because in fact, it's talking about a great new album that, I was going to say LP there, but you probably don't know what they are, do you? I don't know about Peter, but me and Nick do.

I studied them in a museum once.

I've, you know, there's the machine sort of works through filters, natural language processing, and to some sort of machine learning. But ultimately, then what's left is goes to what I broadly call some committees. And if the GT guys are listening now, they'll probably be, their toes will be curling. But yeah, broadly, they are sort of matrix committees. So they'll have people that are specializing in a region, and they'll look at the data for that region. You'll have people that are looking at, for example, weapon types. So they'll be focusing on certain weapon types. And they'll have a committee that's looking at perpetrators, for example. There are a few more as well. But essentially, you have this kind of matrix of looking at those reportings from a sort of a number of lenses. And ultimately, then you get to this kind of committee of committees where they'll sit round and discuss each event and almost by a sort of jury service, decide whether it's in or not. Now, that is, you know, in some ways, well, I was to say inefficient. That's the wrong word. What it is using is it is using the significant academic horsepower that the University of Maryland has to extract, you know, a lot of the or get rid of all the chaff and look at just the key elements and then make those decisions based on those criteria.

Yeah, so I mean, I think everything you said, I just speaks to the the kind of perennial problem with data, especially when you're collecting data about the real world and real world things. It's difficult. It's complicated, expensive and messy because the real world is all of those things. I mean, the fact that it's roughly $2 million per year to run the GTD for 200,000 entries, that's $10 per entry per year, essentially.

Well, it's not two million per year, though. It's sorry. 200,000 entries per year. That's all of it ever.

Yeah, but maintenance of those two. Yeah. So, yeah, it's that that's the $10 per sort of record. That's when you think about all the data that flies around on the Internet, much lower quality, much lower fidelity, much less, you know, potentially less a different kind of data, very different sort of flavor of data. But I mean, but it's interesting how the production of the GTD has changed with it started before the Internet was a thing, before AI was a thing. So there must have been a significant challenge in maintaining a sort of standard quality, a standard definition through it as the process has sort of changed. But I think that's probably handled by these this highly manual process at the quality and the quality of checking end to decide what actually is what qualifies, what doesn't qualify, and then how to measure it is all very conversational between experienced academics.

Yeah, absolutely. Absolutely. I mean, there's there's a sort of a dozen or so full time staff often on working on this on on this project. And then you've got kind of postgrads and interns and all sorts. And, of course, particularly the beauty of technology now, you know, you don't have to be those those virtual committees, you know, exist virtually so they can work on that. I think what's interesting with the GTD and I mean, there are other databases out there. There are some sort of commercial ones. But what the GTD, if it's coming for criticism, it's around its timeliness. So it still does does publish on a sort of a year hence. So it's always looking at the previous year's database. And that's really a function of the fact that because you've got this manual element to it, you know, the December, if we're going to talk sort of calendar months, the December data is still going to be processed. I mean, the January data, you know, that's been that will have been squared away. But rather than having this sort of incremental deployment of the database, you know, it's being looked at on an annual basis. And that's something that, you know, they're looking at into the future. But that balance between, you know, timeliness, thoroughness and volume is a bit of an interaction that is always a challenge. And, you know, that's where they're working. And, you know, let's not forget, most governments, if not all governments, use the GTD as being one of their primary open source databases. So, you know, it must be doing something right. Otherwise, governments wouldn't be wouldn't be using it and funding it.

Sounds like you only want to get on trains or planes that have got 99 passengers.

Well, that's it. So the yeah. So I mean, that that tells you something important, you know, which is that we're dealing with the GTD can only be as good as the information it's got. And, you know, there's going to be rounding, as you'd expect.

So I just jump in there. I mean, that's that is really interesting, I think. And you're right. It's an absolute function of the reporting. But it also brings again highlights the sort of nuances of this. So, you know, define a fatality in an event. Is that the number of people that die on the scene within the first 10 minutes? Or is it the number of people that die that recorded as a result of that death a month later who people have died from their wounds? Or, you know, so there's I mean, you're absolutely I'm not you're absolutely right. There is that rounding element, which I'm sure is a function of essentially news reporting. But there is that also that nuance of accuracy that may be around the timeliness of things.

Yeah, no, I think it's just worth sort of that people who use data just need to understand that at the other end of this data, there is this process. I think a lot of people and I come from a background where, you know, I've used data all my working life and including having done data entry, you know, as a top up my income at university. And, you know, at the end of it all, there's there's someone like me, you know, making a decision about what to put down, you know, sort of, OK, this report says 52. This one says, you know, 48. What should we put? Well, let's put 50, you know. So, yeah, I think it's also worth I haven't mentioned the rest of the work of START, but it might be worth I mean, I'd say they are, as far as I'm concerned, the really the most reliable source of analysis about about terrorism, because the GTD is only one of the databases that they they run. They have other very detailed databases about different types of extremism. And it's really notable when you see what they put out that everything is clearly stated hypotheses tested with data. You know, it isn't that narrative driven stuff, which you often get when you get academics talking about terrorism and it's all kind of narrative and case study. This is they're really, really good. And, Gerry, I just want to say, is it worth mentioning some of the other some of the sort of general work of the START consortium?

Yeah, well, the guys, again, the work that we are building with them and we've we've got to you know, we're looking at a long term relationship with these guys. But I mean, they have another number of databases. They've got one that looks at chemical, biological, radiological attacks in terrorism. As you can imagine, that database is somewhat smaller than two hundred thousand entries. And it is pretty specific. They've got one they're running at the moment as well, which is all around physical consequences of a cyber attack. And that's certainly something from sort of my professional background. We're looking at more and more interest now. So the consequence of, you know, these so-called SCADA systems or industrial control systems where, you know, if you were able to take control of a, you know, a valve in a in a factory that had hazardous material on, could you could you open it and let a, you know, an uncontrolled release? Ultimately, you know, we hear these stories of things like, you know, could you get on the Wi-Fi of an aircraft with your laptop and take control of the aircraft systems? I think we're, you know, potentially a bit further away from that. But certainly, you know, the start guys are looking at that. So this unconventional weapons approach is another thing. I mean, they also have contracts where they're looking purely for the U.S. government on U.S. government issues. And again, I don't want to, you know, labor the point with the U.S., but clearly they have some some unique situations around, you know, far right and far left extremism, as well as, you know, the issue of things like militias and that sort of thing. So there's a whole scope of things that they they also do other than the GTD, as well as, you know, other specific research. And I'm even down to things like, you know, doing doing crisis management scenario stuff. They've got a whole set there that run government departments through crises that can inject real world data into into a scenario.

Well, I was going to give some examples of interesting, interesting things in the GTD. So if that's that sort of moves it on, I think, towards the end. So if Peter's got something more sort of procedurally.

Yeah, yeah. Thanks, Nick. Another very procedurally question. Being the engineer, I'm very interested in how this whole data thing is produced and how the quality is maintained. But I mean, are there any historic examples, Gerry, of how perhaps the GTD has tried as sort of wanted to or had to adopt a new field or adjust its definition and then revisit the entire data set to to make sure that it maintains sort of current or relevant or useful? I'm thinking because if you when you if you set a definition of terrorism back in the 70s, what what you were talking about, what you said was recognized as terrorism that now is possibly have diverged from that definition somewhat. So when you're trying to market and sell a database to analysts, to decision makers saying we can tell you about terrorist attacks in historical terrorist attacks, they may have a different notion of what terrorism means to them. Are there any examples of where additional fields have been retrospectively added or changed?

Yeah, so absolutely. So every year, well, there's two elements to that, really. First thing is that once they push the data out, once they put it out on an annual basis, which essentially is just a giant Excel spreadsheet, they can have the position to sort of update it. But in that within the year, as it were, if new information comes into light, but within the annual updates, they also have this codebook, which they publish every year and any changes to fields and they might change it. They might and I can't I can't remember specifics now, but in the last couple of years, they have, you know, they've broken out definitions. So you've got a different field for a weapon type, for example, or they have they've narrowed that definition somewhat for whatever reason, you know, and sometimes it might just be for the fact that, you know, common terms are going out of use. So, you know, yeah, that's the sort of thing that they can that they are flexible in that. But I think more importantly, they're transparent about it as well.

Which as an analyst is incredibly important. You need to understand how your data was collected, what processes it went through before you actually start using it. Otherwise, it is much less useful. You can only you can you have to caveat your work much more heavily.

No, absolutely. Yeah, yeah.

I mean, the codebook they push out annually is 60 odd pages. And it's a good read, actually, in itself. It's quite interesting discussion of terrorism. So I've got a few interesting things which I've looked at. And it's interesting, Jerry mentioned that that sort of distinction, the US does have quite a unique profile. So we can I'm reluctant to make lice of terrorism, but we could do this as a sort of quiz. So if we think of the main categories of terrorist groups, right, Jerry, I feel like you might be cheating if you join in. But do feel free. UK, what's the most prolific type of terrorism in the UK? Now, this is a classification that I've. I think I know this. I think it's probably animal rights protests. Ah, interesting, because you've missed a really, really obvious thing that we all grew up with.

So you're talking about Northern Ireland still, no?

Would it be a sectarian party?

Is it anti or pro gun lobby people?

I don't know. Maybe I've not looked at the data. Presumably, you can see the improvement in fire safety standards in public buildings played out.

Okay, so I just want to draw things to a close. It's a shame because I feel that we barely even sort of touch the surface of this. And I just feel, you know, we could just keep on going with this all day. But nonetheless, we need we do need to wrap up. You're the one who's stopping us, Fraser. I know it is me, isn't it? Yeah. But there's a couple of questions, Jerry, that I want to ask you. Nothing to do with GTD. Nothing at all to do with that. But a couple of things that intrigue me. There's one question, which I always ask, I guess. But there's another one, which is I think you're the first. This is the first time I've met a bomb disposal expert, which is very exciting for me, since my most of my knowledge of bomb disposal comes from watching The Hurt Locker, which I didn't like very much that film, actually. I don't know. And I'm sure I'm sure bomb disposal experts around the world just tatting away at all sorts of stuff in it. But my question to you is this. This is my first question. Can you tell me in the world of bomb disposal or being a bomb disposal expert, what could you tell us about the job that would perhaps be most surprising to someone like myself, a layperson in this area who knows nothing about it? What's the most surprising thing about the role?

Well, I tell you what, I will use The Hurt Locker for this. By the way, I hate the term expert, but we'll run with it. Do you remember the scene right at the end when he stood in the supermarket staring at the choice he's got of food? I don't, actually. It's right at the end. It's a really benign scene. So you've seen all the running around and the explosions and the gunfights and the helicopters and the this, that and the other and the cutting the red and blue wires, which I thought was the question you're going to ask me. Which color do you cut? Of course, that's a secret. But no, he's standing there in the supermarket looking at the benign. And I think the times that I've been away and done some, you know, sometimes, let's be honest here, it's been quite scary. Actually, when you step out of that scary piece and then come back, it's that transition to normality. And yeah, sorry, this is getting very heavy and deep now. Slightly different from data. Yeah, I'd say that would be the most surprising thing is that, you know, I suppose the other thing you'd see on the Hurt Locker is that everything's done at quite a lot of speed. So if you can slow stuff down, that's always good. But yeah, I would say it's stepping away from, you can quite rapidly be transitioning from something that's really, really scary into something quite mundane. And that sort of adjustment that you make is quite, can be quite a transition.

Interesting. And I suspect, I wonder if it's the same. I'm sure it is. But again, you would know better than me for, you know, other parts of the military and that sort of thing. I suspect it might be similar.

Yeah, I guess so.

Maybe it's more acute. Maybe there are higher pressures. I don't know.

Well, I think, sorry, just very quickly, it depends what situation you're talking about. So the Hurt Locker was, so I didn't deploy doing EOD in what you'd call a kinetic environment. So when there's, you know, there's lots of shooting going on there. My EOD tended to be in kind of minefields or, you know, in the UK where you're in a slightly more stable situation. So you don't have that. You might have the imperative time, but you don't have that imperative people shooting at you. So I can't comment on that. There's far braver people than me that do that. Right.

And so my second question, I guess it sort of relates a little bit. Had you not had the career that you've had, which, as you said, is, you know, it's been very eclectic. And yeah, what would you have done? Had you not gone down that path? What's the alternative career for Jerry Smith? What do you think is in that other universe that Jerry Smith would be doing?

Goodness me. I have no idea. I have absolutely no idea. I mean, I trained as a mining engineer. I've never done a day's paid work down a mine or in a quarry.

I think you ended up looking at the wrong kind of mine, Jerry.

Yeah, I signed up. I saw that. I signed up and lo and behold. Yeah, but it's too late by then.

Yeah, it's too late to say anything at that point. Hold on, mine? What's this? Okay. Unfortunately, yeah, we do have to finish there. So that was absolutely fascinating. I really, really enjoyed that. So yeah, suffice to say, thank you, as always, for listening to the Cognitive Engineering podcast. We've been here with Nick Hare and Peter Coghill of Aleph Insights. But a special thank you to Jerry Smith of CHC Global for providing us with your time and your expertise, your insights. Thank you, as always, for listening to the Cognitive Engineering podcast. Until next time, goodbye.