« Lotta Lott | Main | Grand Rounds »

December 27, 2005

NSA Questions

Matt's post on the potential real world results of relatively small margins of error in a hypothetical NSA data mining algorithm is very much worth your time. As he shows, tiny inaccuracies will ensure the government's net traps large majorities of innocents. The question, then, is what next. And that's why so many of us liberal moonbats are in such high dudgeon. Let's go through a bit of what we don't know here:

The Algorithm: Is it useful? How is it targeted? What percentage of its results are false positives? How effectively are we able to process the true positives and put them on the correct desks?

Now, granted, we can't simply release the details on Slashdot and let the distributed intelligence of the net do it's work, but neither is it true there's no way to create some oversight here. Convene a panel of eminent computer scientists and mathematicians from MIT and CalTech, fully brief them on the program, and have them report back to the Senate Intelligence Committee, the House Intelligence Committee, and the congressional leadership. Or, better yet, convene a bipartisan committee led by, say, Sam Nunn, and have the scientists report to them. Let the panel speak to experts, evaluate the program, and release one classified report to the Senate and a more generalized report publicly.

The Suspects: What happens to them? As Matt wonders, are we then tossing them in for "coercive interrogations," creating endless false confessions and a grand unified theory that encompasses both torture and surveillance? How do we catch mistakes? Who evaluates them? Does each positive result get turned over to FISA? Why don't we create a panel specifically meant to separate the wheat from the chafe? Why isn't there some oversight body keeping track of the whole process, from suspicion to wrongful incarceration?

Why? I'm not necessarily against a program of this sort if properly executed, but why is it such high priority? Bush keeps mentioning how we needed to "connect the dots." Only problem is, he doesn't seem to understand the phrase. Pre 9/11, the NSA, the FBI, the CIA, the FAA, and a variety of other groups had collected isolated bits of information and surveillance that, if laid out on the same desk, would've laid out the 9/11 plot in considerable detail. We had FBI officials noticing the Al-Qaeda members in flight school, NSA intercepts calling 9/11 "zero day," agents theorizing that hijacked planes would be turned to missiles, and so forth. But none of that mattered. Because while we had the dots, we lacked the ability to connect them.

At best, this NSA program collects tons more dots, but are we connecting them? Do we have the ability to process this much information? Or are we dangerously decreasing the signal-to-noise ratio? In short, why is this program necessary? We had the intelligence under the old protocols, we just didn't process it. Why is the answer more data and why should we be confident that the government has the resources to sift through it?

December 27, 2005 | Permalink

TrackBack

TrackBack URL for this entry:
https://www.typepad.com/services/trackback/6a00d8341c572d53ef00d83424796953ef

Listed below are links to weblogs that reference NSA Questions:

» Ask the question, get an answer from Common Sense
Reading Ezra today, I see a question that doesn't seem to have been asked yet: Why? I'm not necessarily against a program of this sort if properly executed, but why is it such high priority? Bush keeps mentioning how we [Read More]

Tracked on Dec 26, 2005 8:18:57 PM

» More data mining skepticism from BookBlog
Matt Yglesias, John Cole, and Ezra Klein have picked up the question about data mining math. If you're looking for... [Read More]

Tracked on Dec 26, 2005 10:58:29 PM

» Ezra has some questions... from In Search Of Utopia
Ezra Klein has some questions about the latest revelations on NSA Spying. My top question remains, "Why no court orders."... [Read More]

Tracked on Dec 27, 2005 11:37:01 AM

Comments

mathematicians from MIT and CalTech

That's me! (It's Caltech btw)

Why is the answer more data

The more you have, the easier it is to cherry-pick from, no?
Perhaps the whole thing is an end-run around Congress, to reward Poindexter for his good works in Iran-Contra. Pesky Congress, defunding TIA just because Poindexter was convicted of lying to them.

Tie-in: Poindexter has two Caltech degrees. (In physics, though.)

Posted by: Allen K. | Dec 26, 2005 8:09:39 PM

Yeah, but with posts like these, I doubt they'll let you on the committee. And don't think they're not watching...

Posted by: Ezra | Dec 26, 2005 8:17:50 PM

great post

Posted by: Kate | Dec 26, 2005 10:06:11 PM

As Theorajones commented, the problem Matthew is discovering is an old one in public health and science more generally. Screening tests need to be applied to an appropriate population in order to be useful. Testing a nursing home full of 80 year-old ladies for HIV is dumb, testing a prison population makes more sense.

None of this is really for or against the NSA's algorithm. It is a tool and can be used wisely or foolishly, like others. In the case of the current administration, the smart money is on misuse.

Posted by: QuietStorm | Dec 26, 2005 11:20:12 PM

Data mining for political advantage : Likely the one place the admin would try for competence . Analysis? Hahahahaha....

Posted by: opit | Dec 27, 2005 12:55:11 AM

I'd feel safer and more open to ideas on how to use technology for intelligence if it was nearly universally believed that each tool/technique was evaluated by independent observers for potential for abuse (as well as efficacy, as Ezra and Matt seem to suggest). By abuse, I mean political retaliation or dirty tricks, or just trying to find out what domestic political opponents are thinking.

If the abuse potential was high, then close oversight independent of the executive branch should be provided. Lower abuse potential should have lower review safeguards. In no case should just one branch of government review their own actions.

But since near universal agreement is impossible to obtain, then lower safeguards should not be acceptable. Even the FISA court seems to need more oversight, given their record of approving almost everything the executive branch requests.

Posted by: JimPortandOR | Dec 27, 2005 2:31:39 AM

The simple and correct answers to a couple of your very astute questions about the whole process are: no, there are no "false positives", and thus no need to worry about separating "wheat from chaff". ALL the identified "bad actors" are terrorists or tied to terrorists. There can be no innocents caught up in the NSA dragnet because the definition of "bad actor" is "someone that the NSA snooping identified as a threat". See? It's EXACTLY like the death penalty in Texas (according to Bush): they have never executed an innocent person in Texas, thus they have/had no need to use new-fangled DNA testing to test the guilt-innocence of those on death row. They wouldn't BE on death row if they weren't guilty...see? Ipso facto, there's no need for oversight. There's nothing wrong here, only the "bad guys" are getting caught (and renditioned). Don't worry your perty little heads on this whole thing.

Posted by: Praedor Atrebates | Dec 27, 2005 9:57:49 AM

I don't think we have the slightest evidence that any American citizen has been dragged of the street for coercive interogation and forced to confess to terrorism.

At the most, I would imagine that a false positive would generate an unneeded search warrent for further monitoring and they would almost certainly never be prosecuted.

If people are being dragged of the streets and confessions extracted from them, then whether they were initially identified by a warrentless search or through some other means is pretty irrelevant. Certainly the majority of people that warrents are issued for are probably innocent. Part of the purpose of the warrent is to answer that very question. Claiming that the NSA monitoring program is a great wrong because of some imaginary wrong of hugely greater magnitude is poor logic indeed.

It is paranoid thinking, not good analysis.

As for eminent computer scientists and mathematicians, my understanding is that the NSA has some of the best and the brightest already.

Lastly, the connect the dots you describe is somewhat misleading. Even if all of the information was at a single desk it probably would not have been actionable. Without knowing who and where there was nothing that could have been done. Imagining otherwise is simply wishful thinking.

The more 'dots' you have, the easier it is to connect them.

Posted by: Dave Justus | Dec 27, 2005 1:37:44 PM

How can one not marvel at the faith, yes, faith in good governance that honest, trusting folks such as senor Justus demonstrate?

While I tend to be more along the lines of the founding father, distrustful of government in general and want strong rules and regulations that limit it, I can appreciate that folks such as Dave and I can agree to disagree. That's American baby!

Posted by: ice weasel | Dec 27, 2005 1:56:27 PM

There is a difference between being distrustful of government in general and naively believing paranoid theories without evidence.

Was the moon landing faked too?

Posted by: Dave Justus | Dec 27, 2005 2:23:18 PM

As for eminent computer scientists and mathematicians, my understanding is that the NSA has some of the best and the brightest already.

Indeed, they are the largest employer of mathematicians in the country.

Posted by: Allen K. | Dec 27, 2005 7:27:45 PM

Justus: I don't think we have the slightest evidence that any American citizen has been dragged of the street for coercive interogation and forced to confess to terrorism.


And no wrong has been done till an American citizen has been physically tortured? All the steps leading up to that, or all the things that can go ill from false positives, are fine? It is generally acknowledged that the administration has: wrongfully imprisoned and interrogated people (just not US citizens), explicitly pays attention to partisan affiliation in non-partisan activities (screening attendies at government events), has been spying on radical but non-violent American organizations, and this is the stuff we know about. I'm not saying this FISA scandal is the end of the world, but don't be so melodramatic as to say that harm hasn't been done.

Lastly, I generally thought the point of sep of powers, and warrants, and all that jazz, was that once the executive reaches the point where we DO "have the slightest evidence that any American citizen has been dragged of the street for coercive interogation and forced to confess to terrorism", then ummm, it's too late to turn back the clock.

Posted by: Tony Vila | Dec 28, 2005 3:22:52 PM

Isn't Padilla an American citizen?

Posted by: Maura | Dec 30, 2005 2:05:18 AM

I've read with pleasure. Maybe it's offtopic, but i just wanted to say, that it's really interesting to read everything this... You discuss here a lot of interesting things on different useful themes. Thanks for that =)

Posted by: Susan | Dec 30, 2005 6:25:01 AM

The comments to this entry are closed.