The Future of Predictive Coding (Part II) – Caveats Revealed
Reprinted with permission.
The Future of Predictive Coding (Part II) – Caveats Revealed
See Part I :The
Future of Predictive Coding - Rise of the Evidentiary Expert?
Maggie Tamburro on
Is predictive coding poised to take the e-discovery world by storm, as the courtroom catches
up to technology?
Following its approval by the U.S. District Court for the S.D. of New York in the seminal
case Da Silva Moore v. Publicis Groupe, Case No. 11-cv-01279 (S.D.N.Y. April 26, 2012), predictive coding has been
garnering major attention as a potential e-discovery game-changer.
Recently we posed some compelling
questions about the future of predictive coding, and whether its recent acceptance in Da Silva Moore signals the
specialized evidentiary experts. This week we’ll continue the discussion (as promised).
Silva Moore: An Update
At last glance, Da Silva Moore is starting to resemble a big screen drama.
On June 15, Judge Peck issued a
denying plaintiffs’ motion that he recuse himself from the case due to an “appearance of partiality.” Plaintiffs
claimed Judge Peck’s prior activities and statements in support of predictive coding showed bias in favor of
Then on June 29, Judge Carter issued an order granting plaintiffs’ motion to amend their
complaint and granted
conditional class certification under section 16(b) of the Fair Labor Standards Act and for claims under the
Equal Pay Act.
So the saga unfolds.
Although predictive coding holds great promise
as a way to decrease costs and burdens associated with traditional manual review of electronically stored
information, its future remains uncertain as its fate continues to play out in the courtroom.
commentators have weighed in on what exactly predictive coding entails, we thought we’d do something different –
focus on what it isn’t.
Da Silva Moore: The Caveats
1. Predictive Coding Is Not a One Way Street
it has now been officially court-endorsed in Da Silva Moore, it’s important to note that the judge did not order
the parties to use predictive coding. Rather, the parties had agreed to its use, but then disagreed on
As Judge Peck noted in his February 22 opinion, for those planning to use predictive coding
or other type of computer-assisted review, it is recommended to first advise opposing counsel of the intention to
use predictive coding, and then try to reach agreement with opposing parties about its use. If an agreement cannot
be reached, “consider whether to abandon predictive coding for that case or go to the court for advance approval.”
At its best, predictive coding requires both parties’ cooperation, active and ongoing participation, and good
faith during all phases of the document review process. For example, in Da Silva Moore, the protocol required that
defendants provide opposing counsel their seed set, even including non-responsive documents, in order to
demonstrate the computer is being appropriately trained. Additionally, the process must be transparent, for
example, opposing counsel should be able to see how each document in the seed set is coded.
Predictive Coding Is Not One-Size-Fits-All
Nor is predictive coding right for every case. Rather,
Da Silva Moore aptly demonstrates that it should be used only in the appropriate case. Predictive coding may make
proportional and cost-efficient sense in a case like Da Silva Moore, which involved review of over 3 million
some estimate that for every 340,000 pages of information preserved for litigation, only one is actually
used). Arguably, however, many instances exist where traditional manual review is more appropriate.
factors come into play in determining whether predictive coding is the right discovery technique for any given
case. For example, how many documents must be reviewed? What is the amount in controversy? Who are the parties,
what are their resources, and what is their respective bargaining power? How important are the issues in the case?
The bottom line: Is predictive coding the best method of ensuring accuracy and completeness in proportion to the
“value” of a particular case? Judge Peck referred to many of these considerations in his reference to the “proportionality
doctrine” used in analysis of discovery requests and set forth in Federal Rule of Civil Procedure 26(b)(2)(C).
The concept of predictive coding is not to achieve perfection (which, as Judge Peck points out, isn’t required by
Federal Rule of Civil Procedure 26
anyway), but rather to utilize a technique which is better than existing alternatives in terms of completeness and
accuracy, and makes sense from a cost perspective.
3. Predictive Coding is Not a Case of Machine
As Judge Peck cautioned, predictive coding is not the replacement of humans by
machines. Above all, predictive coding still requires human management – involving intuition, decision making,
exercise of judgment, input from the parties, and intervention at various phases of the process. Thus, the
ultimate process and end result will only be as good as the decision making criteria of the individual or
individuals involved in supervising it.
Although Judge Peck’s opinion endorsed the use of predictive coding
for certain cases, the stickier legal questions arise not from examining the technology, but in examining the
process – how that technology is utilized, how the methodology and protocols for use are determined, and how the
parties agree (or disagree) regarding its use. As Judge Peck stated in his February 22 opinion, “[I]t is the
process used and the interaction of man and machine that the court needs to examine.”
4. Da Silva
Moore Does Not Hold that Predictive Coding is Per Se Reliable
Given the high rate of human error
often cited in traditional linear manual review, some argue that
predictive coding is actually more accurate and therefore likely to be more reliable than manual review.
However, challenging the reliability of the predictive coding – or more appropriately challenging the reliability
of the predictive coding process – should occur on a case by case basis, with opportunity for judicial review at
For example, in Da Silva Moore, plaintiffs challenged the reliability of the predictive
coding methodology agreed to by the parties and approved by Judge Peck. Although Judge Carter ultimately adopted
Judge Peck’s order approving it, he clearly left the door open for a future reliability challenge as the process
continues, stating, “[I]f the method appears unreliable as the litigation continues and the parties continue to
dispute its effectiveness, the Magistrate Judge may then conduct an evidentiary hearing.”
Da Silva Moore has captured the attention of the legal community and provided valuable
roadmaps as predictive coding begins to go mainstream and undergoes continued scrutiny by courts.
Do you think
predictive coding and the methodologies approved in Da Silva Moore will continue to be upheld? What legal pitfalls
do you think will grab attention along the way?
This article was originally published in
BullsEye, an expert witness and litigation news blog published by IMS ExpertServices. IMS ExpertServices is a
expert witness and litigation consultant search
firm, focused exclusively on providing best-of-class experts to attorneys. We are proud to be the choice of nearly
all of the AmLaw Top 100.
For your next expert witness search, call us at 877-838-8464 or visit our