Information Cascades and Intelligence Oversight

Sen. Blumenthal opened a recent surveillance oversight hearing by hammering an important point from the Privacy and Civil Liberties Oversight Board’s thorough report on the NSA’s telephony metadata program: It was only this summer, after the disclosure of the program by Edward Snowden, that any judge on the Foreign Intelligence Surveillance Court penned a detailed legal analysis justifying the program. As David Cole pointed out in an exchange near the end of a previous hearing, this fact seriously undermines the familiar talking point that “15 judges” have endorsed the program’s lawfulness. (This also, incidentally, puts paid to the argument that Congress somehow “ratified” the FISC’s reading of §215 by reauthorizing it: Even those members who knew of the program could not have been aware of the interpretation supporting it.)  It is, thus, perhaps more accurate to say that at least one FISC judge, having heard only the government’s arguments, concluded that it was lawful, and that subsequently 14 others saw fit to renew that order.  It would be comforting, of course, to believe that each subsequent judge reviewing an application for a bulk records order conducted an independent, de novo review of the government’s legal and empirical claims in support of the program.  But it seems rather more likely that, after the first few orders had issued, most of those subsequent judges would have deferred to the judgment of their colleagues and simply signed off on what now appeared to be a routine order.  Put yourself in the position of, say, the sixth judge to consider such an application—perhaps someone relatively recently appointed to the FISC.  Do you really want to invest countless hours of research and thought in an opinion concluding that five of your colleagues have thoughtlessly authorized an illegal program for several years?  Remember, nobody is going to hail your bold stand for civil liberties if you do this: The only people who will see that opinion are those same colleagues and the government lawyers whose request you would be rejecting.  Or do you instead sign off on the renewal on the assumption that if five colleagues have already approved it, there simply can’t be much serious question about its legality?

If this sounds like a plausible story, it’s because intelligence oversight gives us the ideal makings of a classic “information cascade“: It’s what happens when a series of agents make sequential decisions, each (like good Bayesians) factoring in the decisions made by their predecessors, but without necessarily having access to the detailed information or reasons behind those decisions. Think of a line outside a club or restaurant: Without knowing anything about the quality of the establishment, I might observe that one has several people lined up out front and take this as strong evidence that it’s better than those without.  If many people think this way, then each person who joins the line seems to add to the weight of evidence—the line grows so long that the place must be very hot indeed. If the first few people in line are there because of their direct knowledge of the establishment, of course, the first person to jump on the bandwagon may well be making a perfectly reasonable inference, but as more join, a feedback loop process magnifies the apparent social “evidence” in the club’s favor until it has become completely untethered from any direct assessment of quality. Of course, if other prospective patrons have their own information about the quality of the establishment, the feedback loop can be broken: I will not be fooled by the line outside a place I already know to be terrible.  But when direct information is limited, this sort of social evidence may be the most salient type available.

The extreme secrecy and compartmentalization of information about intelligence set the stage for just this type of information cascade. One famous historical example was deliberately engineered by J. Edgar Hoover, who had persuaded F.D.R. to approve national security wiretaps, with the caveat that they be “kept to a minimum” and directed primarily at aliens. Hoover later apprised President Truman of his predecessor’s authorization decision—but without the crucial caveat, with the result that Truman signed off on far broader surveillance, on the false premise that he was only extending the existing authority. But calculated deception is hardly necessary.  You have the FISC conducting nominal oversight, but with no real ability to actively police the intelligence community’s use of its authorities, a track record of being substantially misinformed about the details of intelligence programs for months or years at a time, and strong potential for internal cascades of the sort outlined above even on the legal questions that are its purview.  Their approval nevertheless seems to reassure most members of the intelligence committees that there’s little need to expend time and resources digging beneath the intelligence community’s sanguine self-assessments. In effect, you have a hall of mirrors reflecting the executive branch giving itself a thumbs-up, but if the lighting is dim enough, it may pass as “rigorous oversight by all three branches of government.” 

About the Author(s)

Julian Sanchez

Senior Fellow at the Cato Institute, contributing Editor for Reason magazine. Member of the editorial board at Just Security. You can follow him on Twitter (@normative).