relevant documents. Nevertheless, judicial “management” is often imposed in
precisely such circumstances, before
the question is ripe for review.
» I’m A Judge, So I Know EDD. Even
if there were no legal barriers, there are
substantial practical risks associated
with judicial activism in e-discovery.
Any judge with significant tenure on
the bench likely conducted discovery
exclusively in paper form while in practice, combing through files in a warehouse; reviewing documents manually
for relevance, privilege, and confidentiality; and Bates-stamping and photocopying those responsive to a request.
In short, judges generally lack the
technical expertise to micromanage
EDD. As U. S. Magistrate Judge John Facciola (District of Columbia) (one of the
more technically savvy members of the
federal bench) has observed, “Whether
search terms or ‘keywords’ will yield
the information sought is a complicated
question involving the interplay, at least,
of the sciences of computer technology,
statistics and linguistics. . . . Given this
complexity, for lawyers and judges to
dare opine that a certain search term or
terms would be more likely to produce
information than the terms that were
used is truly to go where angels fear to
tread.” United States v. O’Keefe, 537 F.
Supp. 2d 14, 24 (D. D.C. 2008).
Perhaps because judges are comfortable with linguistic searches from using
computer-assisted legal research tools
such as Westlaw and Lexis, some have
not heeded Facciola’s caution. In a case
involving a claim under the Fair Labor
Standards Act, one court vetted dozens of keywords submitted by the parties, evaluating whether search terms
like “time,” “overtime,” “poultry and
egg,” or “chicken council” would yield
responsive documents. Helmert v. Butterball, LLC, No. 4:08CV00342, 2010 WL
2179180, at *1- 6 (E. D. A rk. May 27, 2010).
Yet, there is no indication whether
the judge’s choices were ever quality
tested to determine if they resulted in a
more efficient search than the keyword
lists submitted by either party.
The judiciary’s lack of technical
expertise is even more pronounced
when it comes to technology-assisted
review tools that do not depend on
familiar keyword strategies. Few vendors are likely to be willing to share the
details of their technology with a court
for fear of divulging proprietary information to competitors, but even if they
did, there is little chance that the judge
could comprehend the algorithms.
While it is no doubt entirely appropriate to permit a party to choose predictive coding or any other form of tech-
At some point,
morphs into judicial
nology-assisted review to collect and
review data, whether the tool selected
will ultimately produce reliable results
is a determination that a judge is ill-equipped to make in advance. See
DaSilva Moore v. Publicis Groupe, No.
11 Civ. 1279, 2012 WL 607412, at 5-6, 8
(S. D. N. Y. Feb. 24, 2012).
Judges lack not only technical expertise but also knowledge of the facts
that will determine whether a particular discovery protocol will be successful. Again taking the example of a document search, a judge cannot anticipate
the most efficient keywords without
intimate knowledge of the target entity:
how its employees use abbreviations
and codes, what words are associated
with others in responsive documents.
And, because any search is an iterative
process, a judge’s ruling at the starting
point has little impact.
Thus, for instance, a judge might
opine that one particular seed set of doc-
uments is better than another for pur-
poses of initiating a search that utilizes
predictive coding or some other form
of “machine learning.” But because the
search tool will be “reinstructed” on the
basis of the results of each successive
round of searching, the judge’s input is
ultimately diluted to the point of irrel-
evance. To have real influence on the
search process, judges would have to
be involved at every step, which would
strain judicial resources and patience.