I had half forgotten until last week’s Cato forum how much I enjoy a good live debate, but it does invariably mean you spending the next couple days thinking of points you wish you’d made—either because they didn’t occur to you, or because there just wasn’t time to get through everything you scribbled down while listening to the other speakers.  Here are a couple of central ones.

First, the claim that adding checks or civil liberties safeguards will “tie investigators hands” is a sort of stock answer that needs to be evaluated case-by-case, rather than as a blanket claim.  Suppose, for instance, that you had an automatic “expiration date” for information obtained by National Security Letters.  In other words, suppose all the records entered into the database for an investigation were set to expire six months from the time when they were last accessed, unless they were somehow flagged or pertained to an individual specifically targeted in an open investigation. This would alleviate a big civil liberties concern—that data about innocent people is sitting around, waiting for god-knows-what use might be made of it, either by an unscrupulous individual or as part of some unsavory policy—but aside from whatever initial programming would be required to implement it, need not actually demand any extra effort of investigators.  The data they’re actually making use of or flagging as important is preserved; other stuff is not.

Similarly, when you talk about adding additional layers of bureaucracy, you need to compare it to what’s already in place.  So, for instance, when you talk about raising the evidentiary standard for a 215 “tangible thing” order (for instance), you’ve got to bear in mind that applicants are already putting together a statement of facts to bring before a judge.  A higher standard might narrow the realm of what can be gotten, but not actually add much bureaucratic overhead at all. The question in that case isn’t how much extra time or effort the investigators are obliged to expend on applications, but whether the extra information would have made an important difference.

Second, there are a lot of different levels at which investigators are constrained.  People who are accustomed to debating public policy, quite naturally, focus on the macro-constraints at the level of statute. The actual operational constraints on “connecting the dots,” however, are the specific policies and practices in place within the organization. By the time you get down to the rules governing a real live investigator, the statute has been filtered through executive orders, Attorney General’s guidelines, and a range of more specific rules flowing from Office of Legal Counsel opinions, recommendations from the Office of Enforcement Operations, institutional culture, etc. etc.  Now, during our debate, Richard Samp made a perfectly sound point, which is that even if you can say folks at FBIHQ screwed up in (say) the investigation of Zacarias Moussaoui, screw-ups are going to happen, and you need to consider how the formal rules might make those more or less likely, or more or less harmful, even when two different sets of rules would give you the same result assuming perfectly competent officials.  That’s absolutely right. But very often—and certainly in the Moussaoui case—it’s often the lower-level regulations that have the stronger influence on that.

The problem is that when we get to public debate, these complex lower-level rules are a lot more difficult for wonks outside the organization to evaluate—and indeed, they may even be classified. In the wake of a big intel failure, you can be sure that people are looking at how to fix the lower-level rules as well as the macro-level statutory constraints. And the mistake that’s easy to make from the outside is to only look at the macro-change that might improve matters without considering all the other changes that are also being made.  I have no doubt there’s better intel sharing between agencies and divisions now, but it’s difficult to say how much of that we ought to attribute to PATRIOT as opposed to changes in lower-level guidelines.  Obviously, these things are related. There will be things you can’t do at the microlevel without fixing the macro-constraints.  But you need to avoid the mistake of thinking that micro-fixes somehow require macro-changes. Often the best way to “untie” investigators hands will be at the lower level. In fact, I’ll propose this as a sort of general theorem of intelligence oversight:

Civil liberties constraints are generally best imposed at the macro-level; improvements in investigative efficiency should first be sought at the micro-level.

What exactly do I mean, and why is that a good rule of thumb? As defenders of PATRIOT powers like to point out, there are safeguards other than judicial oversight. Certain investigative tools need to be approved by high-ranking FBI officials, or even by the Attorney General. There are all sorts of rules about what can be done under what circumstances that you won’t find in any statute book. But these micro-restrictions are generally more difficult for members of the general public to be aware of, their practical impact is difficult to assess, and they are relatively easy to change. What’s crucial here is a basic difference in incentives and knowledge. FBI special agents and even Justice Department attorneys are supposed to be first and foremost concerned with how to conduct effective investigations, and they are likely to have detailed firsthand knowledge about how specific rule-sets make this easier or more difficult. We don’t expect them to have a clear sense of how their actions might adversely affect civil liberties, and still less do we expect them to be political philosophers mulling over big-picture questions about the long-term structural implications of carrying out certain kinds of investigations or creating large databases. Rather, our system creates a kind of division of labor: Investigators are supposed to focus all their energies on how to best solve cases and hunt down bad guys within the rules, while “neutral and detached magistrates” take the broader view and balance that goal against competing interests. And a good investigator shouldn’t necessarily be focused on this larger picture, just as a good parent will ultimately care about what is best for his own child, even when this is different from what (at the policy level) is better for all children. We actually don’t want parents who all take this cool, detached big-picture view: We want them to have a kind of specific commitment to their own particular children that would be highly undesirable in a neutral policymaker who looks out for everybody’s interests equally.

Now, to be sure, sometimes you need a macrolevel change to enable the efficiency-enhancing change you want to make at the microlevel. But often you don’t. And in general, the information and incentives within the executive branch will tend to create a system of internal rules that adapt to improve investigative efficiency in the face of new circumstances.  In other words, assume there had been no PATRIOT Act; you can rest assured that there nevertheless would have been changes within the executive branch in response to their own diagnosis of what went wrong before 9/11, though of course the statute might still have limited how drastic those changes could be. Those changes will necessarily tend to be less visible, and it will be too easy to fail to take them into account, both prospectively (when assessing what kind of statutory changes are needed) and retrospectively (when assessing the impact of those changes).  This is especially important when we’re trying to talk counterfactually about the effects of policy changes. That is, suppose Method X is relatively unrestricted. What you will doubtless get are a lot of investigations that lean heavily on Method X, and you’ll end up with stories about the successful ones in which Method X plays a significant role. If you assume that the alternative to Method X is that investigators just throw up their hands and give up at the point where Method X was used in the actual investigations, it will probably seem absolutely critical.  But of course, had Method X been more restricted, investigative practice would probably change in other ways to compensate—and it’s a harder question whether the alternatives would be appreciably less effective.

Over time, I think it’s a sound assumption that microlevel rules and practices will tend to adapt to new information and changed circumstances (both in the world and in their available tools) to optimize for efficiency. We should probably not make the same about optimization for civil liberties protection—not, again, because we think badly of the investigators, but because an FBI SAC is not supposed to have the same focus as an ACLU attorney. This does not, of course, mean that policymakers working on the macrolevel should focus exclusively on civil liberties and ignore security and efficiency—there is little realistic danger of that in any event. But when we’re talking about “balancing” these interests at the policy level, we ought to bear in mind that in the multi-year gap between macrolevel statutory and policy changes, there are lots of folks within the executive branch focused on optimizing for efficiency and security at the microlevel.  There will be strong internal pressure to make security-optimizing microlevel tweaks that are permissible within the macrorule framework; corresponding pressure to make civil liberty–optimizing tweaks is much weaker. All of which is to say, policy makers should not “strike a balance” as though macropolicy is the only moving part. So when our opponents in the debate say we focus excessively on the civil liberties side of the ledger, I think a fair response is that this makes a certain amount of sense if you think of policy makers not as designing the full system of intelligence investigation in all its particulars, but as establishing a framework within which we can anticipate substantial adaptation and change with a strong bias toward optimizing for security.