Data Mining as Search: Theoretical Insights and Policy Responses - Discrimination and Privacy in the Information Society - page 321

opposed to visceral feelings) for limiting governmental power. This theory notes
that searches are found to be illegal when they are a powerful tool government
should not be entrusted with (at least without various forms of judicial
supervision). Again, this rationale applies naturally to searches of the home and
self, as well as wiretapping of communications.
When considering the use of data mining for automated predictive modeling, one
can easily argue that government should not be entrusted with such a powerful tool
without being closely scrutinized. Data mining can potentially turn even seemingly
benign factors into a powerful mapping of an individual's persona and insights. For
that reason, ex-ante judicial (or other forms of scrutiny) must be applied.
The challenge of applying this theory to the data mining context and finding
that a privacy interest was compromised is that the analysis here discussed uses
information which was collected lawfully by government. Therefore, the power of
government was already examined and limited when information was collected.
Accepting that a search-related interest might have been compromised in the data
mining context calls for accepting a non-trivial argument: at times the knowledge
provided by the analysis of the sum of the dataset goes beyond the value of the
parts of the dataset previously collected, when viewed on their own. If this is
indeed true, then the fact that the governmental actions were reviewed by courts at
the data collection stage is insufficient. Additional scrutiny is required at the data
mining “search” stage. Given the enhanced ability of data mining tools to engage
in broad, automated and predictive tasks, this argument seems quite convincing.
Data mining transforms small segments of information into an overall “mosaic” of
human behavior.
The provocative notion that many, seemingly innocuous, bits of information
which were collected lawfully should be treated differently in the aggregate is
slowly gaining recognition in US courts which examine the notion of “search”
(although it has yet to be accepted into Fourth Amendment doctrine). Most
famously, this issue is fiercely debated in the context of location-based data
(which is currently easily collected by mobile phone operators and other GPS
devices), while questioning whether there is a difference between collecting
limited and vast amounts of such data. For instance, in a controversial opinion, the
Federal Court of the D.C Circuit chose to restrict governmental collection of
location-based data over an extensive time period while promoting the “Mosaic
Theory.” 4 This finding contradicted previous cases which found that individuals
have no privacy in GPS information which pertained to their actions in the open.
4 U.S. vs. Maynard, 615 F.3d 544, 562 (D.C.Cir. 2010), cert. granted , 131 S.Ct. 3064 (2011).
For a critique, see Orin Kerr, Applying the Mosaic Theory of the Fourth Amendment to
Disclosure of Stored Records , T HE V OLOKH C ONSPIRACY (Apr. 5, 2011, 4:54 pm),
http://volokh.com/2011/04/05/applying-the-mosaic-theory-of-the-fourthamendment-to-
disclosure -of-stored-records. Several courts have taken the opposite position and allowed
for these forms of surveillance. Cf. United States v. Hernandez, 647 F.3d 216 (5th Cir.
2011) (holding that government's use of hidden GPS to track defendant's movements was
not an unconstitutional warrantless search); United States v. Cuevas-Perez, 640 F.3d 272
(7th Cir. 2011) (holding that placement of GPS tracking unit on defendant's vehicle did not
violate Fourth Amendment).