Sunday, June 1, 2014

Unsurprising NSA story of the week

The New York Times reports that the Snowden documents reveal that the NSA is assembling a vast database of face photos.

No one should be surprised. Storage is so cheap that even for a commercial enterprise, it's better to save data once you have it than the throw it away -- no telling how it might come in handy later on A spy agency can start with a pretty good database -- photos of everyone who has ever gotten a passport or a visa, for example. And then it can mine the web, and cooperate with friendly governments abroad, who have their own local databases.

Is this creepy? You bet, though it is not completely obvious why. After all, there is nothing more public than your face. You walk around showing it to lots of people every day. Ordinarily, it's a pleasant surprise when someone you knew long ago recognizes it.

Two things about that. First, remember the koan: More of the same can be a whole new thing. There has always been a chance that someone would recognize you in a cafe while you were traveling halfway around the world. That is very different from an expectation that surveillance cameras in Istanbul will, through some intergovernmental cooperation, trigger an alert to a US agency that you are there -- and by the way, who is that woman you are there with? But that may come to be the norm, if the technology goes unregulated.

The other thing that I have always found creepy about face databases is the opportunities they present to influence behavior. Way back in 2005 Jeremy Bailenson at Stanford showed that voters' choices could be shifted significantly if they were shown photos of the candidates with a subliminal level of the voter's own face morphed into one of them. People tend to vote for people who resemble themselves, even when they don't realize (none of Bailenson's subjects did) that the effect has been achieved by subtly altering a photo. Think of the commercial and propagandistic uses of a tagged database of the faces of most of the world's population.

The article notes that Congress has done nothing on this front. I'll bet they won't, either, but they should.


  1. It's time to buy stock in the manufacturers of false mustaches and such.

  2. NSA/Google (etc) are automating, commodifying, and globalizing a remarkable computational capacity that humans have long been known to possess:

    Half a Minute:
    Predicting Teacher Evaluations
    From Thin Slices of Nonverbal Behavior
    and Physical Attractiveness

    Nalini Ambady and Robert Rosenthal (1993)

    The accuracy of strangers' consensual judgments of personality based on "thin slices" of targets' nonverbal behavior were examined in relation to an ecologically valid criterion variable.

    Consensual judgments of college teachers' molar nonverbal behavior based on very brief (under 30 seconds) silent video clips significantly predicted global end-of-semester student evaluations of teachers.

    The findings are surprising and provocative. They suggest, first, that our consensual intuitive judgments might be unexpectedly accurate, and second, that we communicate — unwittingly — a great deal of information about ourselves.

    Not only do we possess the remarkable ability to form impressions of others, as Asch (1946) suggested, but, perhaps more remarkably, the impressions that we form can be quite accurate.

    A Modest Proposition  Any one university can improve the quality of its entering class — while simultaneously cutting admissions costs and reducing faculty work-load — by requiring solely the submission of a picture … grades, personal essays, letters of recommendation, and even names not required … on the grounds that supplemental information (including names) can be data-mined starting from the picture, such that reasonably accurate predictors of student value to the institution (including post-graduation donation potential) can be distilled automatically, quickly, and cheaply.

    As Weyland-Yutani market-analyst Carter Burke puts it:

    This is clearly an important market-optimizing capability we're dealing with and I don't think that you or I, or anybody, has the right to arbitrarily exclude it.

    After all, how can any globalized corporate enterprise — including universities — responsibly and competitively refrain from market-optimizing practices that are feasible, effective, and well-grounded in cognitive science?