Tag: mosaic theory

Blurry Lines, Discrete Acts, and Government Searches

I’ve written before about the “Mosaic Theory” some courts have recently employed to conclude that certain forms of government surveillance may trigger Fourth Amendment protection in the aggregate, even if the surveillance can be broken down into components that don’t fall under the traditional definition of a Fourth Amendment “search.” This has been applied specifically to high-tech forms of location tracking, where several judges have concluded that a person may have a privacy interest in the totality of their public movements over a long period of time, even though observing a person at any particular public place in a specific instance is not an intrusion on privacy. I’ve explained in that previous post why I find this reasoning compelling. Legal scholar Orin Kerr, however, remains unmoved, and suggests that divergent decisions applying the Mosaic Theory to government acquisition of stored cell phone location records effectively serve as a reductio of that theory:

To my mind, this opinion reveals the absurdity of Maynard’s mosaic theory. The analysis is all “look ma, no hands.” No one knows where the line is, or even what the line is. Sure, you could just count days of surveillance: perhaps 30 days triggers a warrant but 29 days doesn’t. But there is no reason the access to records has to be continuous. The government can skip around days, or get records from a few days here and a few days there. Who can tell how much is enough? No one knows what is revealing, because what is revealing depends on what the records actually say — and no one but the phone companies know what they say. So Judge Orenstein has to wing it, announcing that “he cannot assume” that the information would be revealing because it has breaks in time. But it’s not clear to me why the break in time matters: It’s the same net amount of data collected, so I don’t know why it matters if it was collected all at once or over several discrete periods. And how much of a break matters? If 21 days is too long, is 21 days with a one-day break enough? How about a 3-day break? One week? No one knows, it seems, not even the judge himself. [….]

There are some readers who will say that the cause of justice sometimes requires hard decisions, and that if judges need to make arbitrary calls like that, then that is what we pay them to do in order to enforce the Constitution. But as I see it, the oddity of the inquiries called for by the Maynard mosaic theory shows why it is not part of the Constitution at all. In Fourth Amendment law, the lawfulness of government conduct has always been viewed discretely: Each government act is either a search or it is not a search. Under Maynard, conduct can be a non-search if viewed in isolation but a search if viewed in context — but there is no guide to tell how much context is proper. If you want to say that certain conduct is a search, then just be direct and say it’s a search. That’s fine. But a mosaic theory, in which non-searches become searches if grouped a particular way, has no proper place in Fourth Amendment law.

Orin’s point about the seeming arbitrariness of these determinations—and the difficulties it presents to police officers who need a rule to rely on—is certainly well taken. The problem is, the government is always going to have substantial control over how any particular effort at information gathering is broken into “acts” that the courts are bound to view “discretely.” If technology makes it easy to synthesize distinct pieces of information, and Fourth Amendment scrutiny is concerned exclusively with whether each particular “act” of information acquisition constitutes a search, the government ends up with substantial ability to game the system by structuring its information gathering as a series of acquisitions, each individually below the threshold.

Let’s consider a concrete case involving location monitoring. Under the Supreme Court’s ruling in United States v. Karo, technological location monitoring does count as a Fourth Amendment search requiring a warrant when it reveals information about where the tracking device is located within a private place, such as a home. On this theory, if the police want to be able to pinpoint a target’s location with sufficient precision to be able to tell when he goes from the garage on one side of the house to the bedroom at the other end, they’ll need a full blown search warrant. If they just want to know the general area the target is in—which cellular tower the phone is closest to, for instance—a subpoena or another less demanding form of court order might be sufficient.

There are, however, several methods of determining a phone’s precise location by triangulation, using data from multiple cell towers—and many cell networks use these methods to provide location services. The records from any one cell tower only yield a very general radius within which each phone registered at that tower can be presumed to be located. Combine the records from the three nearest towers, however, along with some measurements of signal strength and timing, and in an urban area where towers are relatively densely packed, you can often pinpoint the phone within a few meters.

Let’s suppose, then, that existing doctrine would require a warrant if police plan to go to the phone company and say: “We want you to triangulate the precise location of this phone for us over the past month, including at times when our suspect was at home.” What a hassle! They’ve got an out, though: They can issue separate requests for the records from each tower, then combine the data and do the triangulation themselves. As long as each request “viewed discretely” doesn’t yield enough information to pinpoint the phone within the home, there’s no search!

I don’t mean to suggest that, in practice, police are likely to use this particular method to circumvent the warrant requirement—though I wouldn’t be shocked either. But I think the example illustrates a problem with Orin’s categorical insistence on making the binary search/no-search determination only with respect to isolated “acts” of government, when the government itself controls how its monitoring is distributed across discrete acts.

Here’s another example, and one where I think there is a very real possibility that investigators are able, in practice, to game the standards governing electronic surveillance. According to the Justice Department’s U.S. Attorneys Manual, a “pen register” (which can be obtained much more easily than a search warrant) can be used to obtain general information about the domains or IP addresses a target is visiting, but not what particular pages somebody is reading. The idea is that there’s a sharp Fourth Amendment distinction between the “content” of a communication—its “meaning or purport”—and the non-content transactional information, such as the phone number or IP address, which tells you something about who is communicating, but not what is communicated. But there’s a loophole:

This policy does not apply to applications for pen register orders that would merely authorize collection of Internet Protocol (IP) addresses, even if such IP addresses can be readily translated into URLs or portions of URLs. Similarly, this policy does not apply to the collection, at a web server, of tracing information indicating the source of requests to view a particular URL using a trap and trace order.

Emphasis added. Roughly translated, this means that the government can obtain records showing that I accessed (say) the IP address of a particular political Web site, but not which specific articles I was reading. However, they may be able to separately go to that site and request the transactional logs for each article, then search through those to determine which articles were sent to me.

It seems very likely that technology will increasingly permit this kind of multi-step searching, perhaps in ways we can’t yet predict. For all that Orin is right to worry about the practical difficulty of determining how to group discrete acts of information gathering, the consequences of dogmatically insisting on evaluating each “act” in isolation seem equally absurd if it implies that the government will have the practical ability to transform a Fourth Amendment “search” into an unregulated (or much less regulated) “non-search” just by breaking it into smaller pieces.