pointer to reflections on NSA surveillance of USA citizens

[jrtom]

Amygdala: Blue in the Face

This is mostly a placeholder in case I come back to this later, but this blogger suggests that the reason why Bush & co. didn't get the warrants was that they were doing large-scale pattern analysis on the communications of tens of thousands of people (or more) . . . thus making acquiring warrants impractical at best.

This kind of analysis is precisely what I do in my research. I have no doubt whatsoever that I could get a job with the CIA or NSA to simply continue doing what I've been doing. Let me be clear: I don't think that there's anything ethically wrong with the research qua research; the evil, if any, is in how it is used.

But it still itches me.

Comments

Right. Evil, that is.

[fdmts.livejournal.com]

I don't think that there's anything ethically wrong with the research qua research; the evil, if any, is in how it is used.

To this I would add: The evil is in the trampling of civil liberties required to perform the monitoring, and the brazen audacity of gathering the data in the first place. Once the data is gathered, certainly, there's no additional evil incurred by the poor math geek who finds some eigenvectors in it (or whatever the cool kids do with data these days).

Reminds me of the conundrum of researchers who could do really good stuff with the primary data gathered from the human experimentation of the Nazis in ...

Crap. Godwin. I lose.

Re: Right. Evil, that is.

[jrtom.livejournal.com]

I would say that there can also be evil in the use of the results of the analysis. Especially because the results are almost always equivocal, and policy makers tend to (want to) interpret probabilities as certainties wherever possible. Or, on a different level, "we think there's a 75% chance you'll receive a communication from someone affiliated with al-Qaida within the next six months, so we're going to threaten you and your family with GBH unless you cooperate with us".

In some ways I feel as though what I ought to be doing, research wise, is explore methods of social network analysis that require as little data as possible, so as to not be encouraging large-scale data collection of this sort. (My research has been--and, I hope, will continue to be--public, so I can't prevent various three-letter agencies from being ordered to misuse it.) But certain very useful metrics that I use in my models--some of them indeed related to eigenvectors--would probably be really skewed by, say, 99% missing data (and that's still surveilling 1%!).