Monday, December 14, 2009

The Eleventh Balloon (DARPA Network Challenge)


I was discussing the DARPA Network Challenge with another participant in the contest the other day and the conclusion of our conversation led me to ask why couldn't a similar system be used to locate Osama bin Laden?

For those of you unfamiliar with this challenge, DARPA offered $40,000 to the first individual or team to identify the location of 10 large red helium balloons that DARPA had moored at various locations around the US. The contest was to go on for a week but a team from MIT won the challenge in a little less than 9 hours (for the Lessons Learned from the Mercyhurst effort, see this post).

It occurs to me that the Osama bin Laden problem is a similar one and that social networks might well be as effective in identifying his location (particularly if he is not in the region where most people seem to think he is...).

Moreover, the conditions are much the same -- only better. The reward for bin Laden is 25 million! MIT's winning system seemed to work so well (to me) because everyone in the chain got a piece of the pie. With a much bigger pie, it should work that much better.

Furthermore, there appears to be no reason not to try. Secretary Gates recently stated that there has been no good intelligence on bin Laden "in years". In addition, GEN McCrystal recently testified that he didn't think we could finally defeat Al Qaeda without capturing or killing Bin Laden.

And its not like Pakistanis and Afghanis don't use social networks. They do. Facebook and Orkut are both popular in Pakistan. If the system were optimized for cell phones, it would likely be even more effective as 49 of every 100 Pakistanis (29 of every 100 Afghanis) has a cell phone (Note: These are 2008 numbers. They are likely higher today.) Even the FATA in Pakistan has cellular service. If bin Laden is not where we think he is, then the system could be even more effective.

Using social networks allows us to tap into those people in and around bin Laden who disagree with his message or tactics or both but see no way to get their info to those who can do something about it (and frankly, even if they could, they likely see no advantage in it). It allows us to negotiate with the innocents who are victims of radicalism rather than continue to do business with the guilty that traditionally operate within the elite levels of a state.

Crowdsourcing the bin Laden hunt would also be a real test for these new technologies. A successful effort here would likely have as much impact on the practice of intelligence as 9/11 had on our thinking about terrorism.

So what say you, DARPA? The "eleventh balloon" is out there! Get with the Rewards for Justice people and announce another challenge (this time for some real money).
Reblog this post [with Zemanta]

6 comments:

Anonymous said...

One of the tools used in the counter-insurgency campaign in Iraq was photographing and cataloging faces and identities. If, instead of orkut and Facebook, a tool like Flickr were widely used in Pakistan or Afghanistan a facial recognition crawler would seem like an easy tool to identify intelligence targets.

The MIT approach to the challenge was interesting in that it spread the wealth on the reward. It created incentives for sections of the network as opposed to an individual. I think that was a significant part of the innovation on their team. Brand-name and marketing helped build their team, they quickly got the word out, posting links on nearly every article on the contest I saw.

I am hopeful that the MIT team will produce an after-action on their experience, or include it in some sort of research. The DARPA response was, probably necessarily, simple and vague.

das said...

This wouldn't work as easily as with the DARPA Network Challenge, because the types of networking, and tools which enable such networking, aren't pervasive and ubiquitous in the area(s) where bin Laden is thought to be.

MIT built its own tool to create and track the network, and distribute the reward. There was no penalty for participating and publicizing it -- in fact, that was the desired mechanism.

What I find to be the most promising is the concept that people somewhat "removed" from the balloons could still benefit, and that's what made MIT's effort a success. No one who directly knows where bin Laden is cares about the $25M reward. But what about some people two, three, or four degrees removed? Could a structured network and incentive model reveal bin Laden's location?

Perhaps, but are people willing to possibly risk their lives in return? How can such a network be built without the tools to enable it?

Kristan J. Wheaton said...

I agree that creating a network people can trust would be the most difficult thing. There has been a good bit of research on this over the last several years, however. I think such an "add-on" is possible to the basic MIT system.

I also think it is worth opening up such a "contest" to other competitors. Rather than trying to figure it out in advance, I say let the crowd figure it out and the best system will win the 25 Million.

Intel and CT Weekly said...

Kris,

It's really interesting to read analysts' thoughts on intelligence collection.

US intelligence collectors have long had a very robust, "crowd-sourcing" method of collecting intel.

There is a link on the CIA website, and many other websites, where those with knowledge can provide tips: https://www.cia.gov/cgi-bin/comment_form.cgi

And the Rewards for Justice program offers the same.

All the MIT team did was to form a collection team--most likely by making use of its contacts with "elites" across the US. Very likely they used contacts with journalists, politicians and others in their own networks.

Unless your network includes terrorists, a "crowd-sourcing" search for terrorists isn't likely to get very far.

The US intelligence collection effort against terrorists was hobbled in the 1990s by the Clinton adminstration's making it illegal to use anyone suspected of "human rights abuses" to collect intelligence.

Losing intelligence collection assets who were "bad guys" pretty much made the US blind to terrorist plans.

The bulk of the US intel collection effort is focused on finding bin Laden. Maybe it would surprise analysts, but "crowd-sourcing" is nothing new in collection. However, the key is not in crowd-sourcing, but in finding sources who have real knowledge.

The past prosecutions, and the current investigations of intelligence collectors, weaken our efforts.

Maybe analysts are ready to get out in the field and get their hands dirty, and deal with human rights abusers. Should be an interesting operation....Good luck to them!

Kent Clizbe



Contact CIA
Online E-mail Form:

We read every e-mail we receive but with limited staff and resources, we simply cannot respond to all who write us.

We do not routinely respond to questions for which answers are found within this Web site.

If you have information which you believe might be of interest to the CIA in pursuit of the CIA's foreign intelligence mission, you may use the form below. We will carefully protect all information you provide, including your identity. The CIA, as a foreign intelligence agency, does not engage in U.S. domestic law enforcement.

If you have information relating to Iraq which you believe might be of interest to the U.S. Government, please contact us through our Iraqi Rewards Program .


--------------------------------------------------------------------------------

Please Note:

We use Secure Socket Layer (SSL) encryption to ensure the privacy of any information you submit to us.
If you do not see text editing areas below, your browser does not support forms; return to the Contact Us page for information about contacting us by mail, phone, or fax.
Our form does not accept keyboard characters commonly used to illicitly enter computer systems. We sincerely regret the inconvenience.
If you copy and paste text from other programs, the text often will contain characters we do not allow and which you may have difficulty finding and deleting. The characters most often at fault are quotation marks ( " ) and the "@" sign in e-mail addresses.
Non-English comment form: If you wish to submit comments in a language other than English, use our non-English character set comment form.

--------------------------------------------------------------------------------

Your e-mail address (required if you want a reply):


Your alternate e-mail address (optional):


Your name (optional):


Your phone number (optional):
Country Code: Phone Number: Extension:


Mobile/Cell Phone Number (optional):
Country Code: Mobile Number:

Your message:


Additional methods of communication (optional):

Kristan J. Wheaton said...

Kent,

I appreciate your comments but have to disagree with most of them.

As someone who has been on both sides of the issue, I believe that the distinction between an "analyst" and a "collector" is one that is not very helpful. Good collectors analyze and good analysts collect.

In addition, there is a significant difference between what the CIA and other HUMINT agencies traditionally do and the notion of crowdsourcing. Most HUMINT operations attempt to identify and approach those suspected of having specific knowledge. Crowdsourcing assumes that no one person has the bulk of the relevant knowledge but that all of us, collectively, do. It is "we is smarter than me". It is the antithesis of traditional HUMINT ops.

There is also a huge difference between what the CIA and Rewards For Justice are doing with their tips hotlines and what MIT did in the balloon challenge. Tips hotlines rarely gather much of interest because the signal to noise ratio is too high. It is not entirely clear yet how MIT dealt with this problem but there were two significant surface differences. First, the MIT effort rewarded everyone in the chain not just those in the know. It encouraged crowd participation rather than focusing on just the elites. Second, there was a very strict time limit on the exercise. This clearly had a psychological impact. Participants had to put up or shut up, as it were. No chance of a Curveball dragging things out to get more cash.

I also disagree that a social network would have to include terrorists to identify terrorists. There are clearly those that come into contact with terrorists and who may well recognize them as terrorists who are not terrorists themselves.

The key problems are hardening the system in order to protect those participating and motivating people to get involved. I think both of these are solvable issues. Given the stakes and the failure of traditional intel methods to identify where bin Laden is hiding, I certainly think it is worth a try.

Kris

Intel and CT Weekly said...

`Kris,

Analysts "collect" insofar as doing research is collecting. Collectors "analyze" insofar as all humans analyze. An untrained analyst is able to collect human intelligence just as well as an untrained case officer could write an article for the PDB. And that is...not very well.

That said, establishing reporting networks of HUMANS who provide information and INTELLIGENCE is better known as HUMINT Collection.

Giving it a trendy name just obfuscates what is really going on. "Crowdsourcing" is nothing but an attempt to rename the process of winnowing down a large group of people until you find the one with access to your target.

Furthermore, the notion that: "...no one person has the bulk of the relevant knowledge but that all of us, collectively, do. It is "we is smarter than me". It is the antithesis of traditional HUMINT ops." is just silly.

You could say that aggregating the "collective knowledge" of a crowd of 10,000 people passing through Grand Central Station in 1938 would have gotten you the formula for the atomic bomb.

But such an approach would be stupid. 9,999 of the crowd did not even know what the atomic bomb was. But Klaud Fuches, who was in the crowd, had the formula.

As a collector, why would I waste my time with "crowdsourcing?" Why not focus my efforts at identifying Mr Fuchs, and then recruiting him?

This is just elementary humint collection theory and practice.

Humint recruitment cycle:

Spot
Assess
Develop
Recruit
Run
Terminate

"Crowdsourcing" is nothing but casting a wide net in the Spotting process. It is absolutely no different than putting out the word for people to contact the American Embassy with details about the location of Osama bin Laden.

Once tips begin to flow, the collector must sift through them (this is part of the Assessment process) and follow up with personal contact of some sort. Then begins Development of the source, and further Assessment. Then the source is Recruited.

This is HUMINT COLLECTION.

Regardless of the media used to contact sources, whether it is Facebook, a chat room, a blog, a phone call, or a meeting in a hotel lobby, the whole recruitment cycle is used.

Even if the source is a clerk in a gas station that you call on the phone to ask him to look out the window to see if there's a red balloon in the sky, you still go through the Recruitment Cycle. It is abbreviated, but nonetheless it is there.

You must Develop the source before they will spill their guts to you about bin Laden's whereabouts, or the location of the balloon.

You must Assess the potential source's Access, Motivations and Suitability.

You can call the process of Spotting "prospecting," or you can call it "sourcing," or you can call it "crowdsourcing," but that does not change the fundamentals of the process, which are as old as humanity.

"Crowds" don't have intelligence. Only the specific people with access to information have intelligence.

Asking a random "crowd" where bin Laden is, is about as useful as asking a random crowd where a red balloon is. It's only worth asking those who have the requisite access to the information.

Targeting analysis is necessary to determine who or where to most profitably ask the questions. This IS an area where collection and analysis overlap, and where analytical training could be profitably applied.

The point of my comments is the fact that the DARPA Balloon Challenge required humint collection skills, and the humint recruitment cycle. Sending people (analysts, or computer programmers, or librarians, or anyone) to do humint collection without understanding what it is they are engaged in is a bit foolhardy.

Kent Clizbe
kent@cameronhalifax.com