(Shane Harris is a writer at Washingtonian magazine and author of “The Watchers).
ROSENZWEIG: Tell me what you mean about data mining, when we talk about data mining, and if you think it’s any different from link analysis, or do you distinguish between the two? Give me some flavor of what you think the topic area is and what differences or similarities or definitional problems you might have in it.
HARRIS: I think a lot of people use the term “data mining” as sort of a catch-all for any number of what I would refer to as computer-assisted analytic activities. So, when I think of data mining, what I think of is using computers and algorithms to find statistically relevant or even otherwise relevant, depending on your purposes, connections between disparate pieces of information in a very large set of data.
And to my mind, because I’m interested in the counter-terrorism aspect, I break it down into two groups. If you’re talking link analysis, I do consider that data mining if you’re doing it on a large scale, obviously. I think data mining has a lot to do with scale. And so, if it’s link analysis, I think of that as sort of an investigative tool. So, if you’re an FBI agent who is looking into connections between – or a particular suspect or a known terrorist or someone in custody and is trying to find the connections to other people, places, and events, pretty straightforward stuff that’s largely an investigative tool.
The other category I think about, which is where in the kind of total information awareness style of, quote-unquote, “data mining” falls in is when you’re talking about looking for connections, not obvious relationships, about people who aren’t necessarily on your radar yet. So, looking at a sea of ostensibly innocent people who’ve done nothing to alert anyone’s suspicion but looking for the things that might alert you to their suspicious behavior.
And I break it out for those two reasons because I think that the first example of looking for the links among different actors, to my mind, is a lot easier than the second category. You can have a bit more, I don’t want to say certainty because that’s the wrong term, but you get a little more assurance that if you’re looking, you know, if you’re finding records that tie people together and looking for those relationship in a link analysis kind of way as opposed to the likelihood that somebody might be someone who commits fraud or commits a crime, it seems to me that the first category is a bit more sort of dependable in a way. And it seems to me, too, that the first category is the one where there’s been more, I guess you’d say advancement in the technology. The second category of almost trying to predict things that are going to happen is still, I think, not as solid a science, maybe, as the first one, if I can call it that. Just for my purposes, I kind of break them into those two categories, because it’s also easier for me as a reporter to think about sort of the work that people in the government do to investigate acts of terrorism versus to pre-empt acts of terrorism. Those are two discrete sets of activities. So, that’s another reason why I like to sort of break it out into those two categories.
ROSENZWEIG: How do you think that reporting in this area on data mining is different from writing about other subjects?
HARRIS: It’s a lot more technically, technologically demanding. You’re writing about things that computers do. You’re writing about technology. You’re writing about things that are opaque to most people because, A, they involve fairly complex processes and algorithms, but, B, you don’t really see it.
What is the activity of data mining that one can imagine seeing the way you could imagine seeing cops break down a door or some kind of other physical activity. It happens inside machines, and it happens at the level of a network. I think for that reason, it’s largely opaque to people. So, my challenge in reporting on it has always been to look for the thing that animates it, so to look for analogies or to look for metaphors or to write about the people who are building these machines as opposed to what the machines themselves are all about. And I think that’s the best way that I’ve found to make this a subject that doesn’t just make people’s eyes glaze over.
ROSENZWEIG: Tell me some of your experiences. What do you think were some of the interesting experiences you had in writing about data mining.
HARRIS: Well, one of the things I really enjoyed, you know, when I was writing, right around 2004, 2005 and ’6, that kind of time frame, was really kind of digging into the various data mining – again, broadly speaking, data mining programs that were at play in the intelligence community and finding connections among them that had not been previously reported. And one of the really neat things about technology in the government is that you find a lot of the same ideas getting recycled but with different names.
So, for instance, I wrote a story that was the first one to report that John Poindexter’s Total Information Awareness program had not actually been completely shut down, but that it had been shut down in name only, broken into smaller subcomponent programs that were given new code names and then switched over to the intelligence budget for classified funding. And that was kind of a neat, like, sort of treasure hunt in a way to kind of do investigative reporting on these programs themselves.
And then when I was reporting for the book, “The Watchers,” what I found was that a lot of the ideas that Poindexter had under the auspices of Total Information Awareness were getting picked up and started as kind of new programs to different agencies, like in the Homeland Security Department, for instance. And it was really fascinating to sort of, you know, as a reporter to be kind of combing through lots of documents, like technical documents, saying, you know, here are the umpteen different programs that the Homeland Security Department is working on for data analysis.
So, that’s kind of fun. That’s the point where I think you realize that you’ve gotten so close to this that you start to understand the language that technicians speak, and it’s a little scary. But it’s helpful because I think that there’s a tendency to kind of sort of paint all the government’s various activities with information and information technology with this kind of broad brush. And that’s useful to a point, but when you really want to start discerning the different kinds of activities for different missions and purposes that the government is using technology and data mining and things like that, then you really have to start to understand the subtle differences in each of these ideas and these programs.
ROSENZWEIG: What would you say are the good sources for information? And along the way, if you could speak to whether or not – how you deal with this in the context that some of these programs are obviously highly classified.
HARRIS: Contractors tend to be, I think, just better sources in general because they’re not governed by the same – well, they technically, if they’ve signed a secrecy oath, they’re governed by the same constraints that a government official is. But I think that practically speaking, people who work in industry are a bit more forthcoming for a couple of reasons. One, they’re not working within the government system, so I think they feel a little bit less constrained by it. And second, many of these companies are interested in promoting their product or promoting their idea or concept, so what you’ll find is that, I think, in my experience anyway, there’s almost an attitude among people in the private sector that any publicity is good publicity.
If I have a story that sort of comes across as very negative or alarming about a particular technology, let’s say, and the company’s name is associated with it, I’ve always presumed that at some level, that company looks at that and says, well, you know, at least now everyone in the government knows that’s what we do and we’re the best at it. Yeah, I think there’s a little bit of a different incentive in the corporate sector to get your information about yourself out there.
The other thing is just really that I’ve found was one of the most helpful things in writing about these programs is our documents. In a variety of places you find references to programs, and they’re really often buried deeply within things like congressional reports, inspectors general reports. One place that’s actually a great place to look is the government’s main contracting website, FedBizOpps. Oftentimes, you’ll find, like, calls for – requests for proposals or ideas or information sent out broadly to industry through that FedBizOpps website. In fact, one of the programs that I wrote about that was a successor to the Total Information Awareness program was advertised in this system, and it was because people in the government were looking for people in the private sector to send them ideas about it.
ROSENZWEIG: What do you think are some of the underreported areas or issues in this field of data minding and link analysis? And are there any programs or areas where you think people should look for good opportunities for stories? That you’re not using yourself, of course.
HARRIS: In terms of the underreported aspects, I think that generally, a lot of the kinds of research that the government is doing in the areas that I write about, like pre-empting a terrorist strike by looking at large amounts of information and this kind of thing, I think that’s been sort of underreported. But I think a lot of that is because it’s really hard to get. You kind of have to dedicate yourself to writing a book about it and taking time off to do it.
I think also one thing that’s not terribly well reported now, and it’s something I’m interested in writing about, it doesn’t really have to do with data mining, per se, but a lot of the technology that’s being used now to assist in drone strikes involves a lot of different kinds of signals intelligence and gathering huge amounts of data for the purposes of targeting, for identifying targets on the ground. And I think that’s something that’s just fascinating that I haven’t seen a lot about.
I can’t emphasize enough how important just document research is for these kinds of stories. It still amazes me how every now and then, you will see a reference to something, whether it’s explicit or just sort of implied in the document, and, you know, if you just kind of read between the lines and really start asking good questions.
ROSENZWEIG: Any other advice you can give someone? Any other pieces of advice you want to throw on the table for the aspiring young reporter?
HARRIS: Don’t be daunted by the technological aspects of it. You know, find a set of sources who are very good at explaining things in lay terms and really know what’s going on. For me writing my book, for writing “The Watchers,” John Poindexter was an outstanding resource because he was so technologically skilled, and he could really explain in fairly straightforward terms what a particular technology was doing and what it wasn’t doing. And then don’t be deterred by the fact that the government is not going to help you out and help you do the story. You have to be very persistent, and you have to be very patient. Intelligence is one of those beats that, stories don’t develop every day, and stories don’t just fall out of a tree. And when they do, you should probably be pretty suspicious of it. It’s a tough beat, but it can ultimately be a very satisfying one. It’s a place where you just have to persevere as a journalist and not get either overwhelmed or deterred or burned out. Just, you know, it’s kind of a slow and steady wins the race kind of thing.
ROSENZWEIG: Have you made any mistakes that you regret in this area, things that people shouldn’t do?
HARRIS: Sure. I think early in my career, I probably was sometimes too trusting an official explanation for something. We should be professional skeptics. And so, I think there were some times in the beginning where I was perhaps a little bit too trusting or a little bit too willing to believe, well, you know, what bad could come of this? Now, obviously I changed my tune as I got older and I started writing about this with a different angle. But I think had I been a little bit early on a little bit more skeptical, that probably would have been a good thing. Hopefully I didn’t get turned into too much of a cynic.
