Don’t TAR and Feather Technology in E-Discovery

March 11, 2015 Jennifer Latzo

In the three years since the watershed decision in Da Silva Moore v. Publicis Groupe, technology-assisted review (TAR) has transformed from being a merely “acceptable way to search for relevant” ESI into “black letter law that where the producing party wants to utilize TAR for document review, courts will permit it”—according to a new opinion from U.S. Magistrate Judge Andrew J. Peck. However, as his opinion acknowledges, transparency remains a hot-button issue.

In Rio Tinto PLC v. Vale S.A., subtitled “Da Silva Moore Revisited,” Judge Peck approved the parties’ stipulated TAR protocol. However, instead of merely signing off on the protocol, he addressed the open issue of transparency “because of the interest within the e-discovery community about TAR cases and protocols.”

According to Judge Peck, court opinions and the discovery literature are split on the issue of whether parties should share the seed set of documents they use to train the TAR algorithm, which is likely to contain non-responsive documents. He did not rule on the issue here, because the parties had agreed to share all non-privileged documents in the control sets.

Our take on sharing the seed set used to train TAR:

In Da Silva Moore, Judge Peck recommended that counsel in future cases should “be willing to discuss, if not agree to, such transparency” in sharing seed sets.

Parties have been hesitant to adopt TAR as a money saving tool because sharing seed sets in their entirety has confidentiality implications involved in permitting easy access to non-responsive (and potentially sensitive business) information to opposing parties.

In Rio Tinto, however, Judge Peck tempered his view on sharing the seed set, sending the clear message that this was not the only way to confirm that the training and review were proper. He suggested three alternatives to sharing the seed set: (1) calculating a “statistical estimation of recall at the conclusion of the review,” (2) assessing “whether there are gaps in the production,” and (3) performing “quality control review of samples from the documents categorized as non-responsive.”

Judge Peck also stated:

“It is inappropriate to hold TAR to a higher standard than keywords of manual review. Doing so discourages parties from using TAR for fear of spending more in motion practice than the savings from using TAR for review.”

This is an important point—since no other form of analytics (email threading, concept searching, etc.) is held to a higher standard of discovery.

Thanks to this opinion, counsel now have support for the position that sharing the seed set is not always necessary when other options for validation are available. Judge Peck was able to provide at least some clarity on this topic and reshape the post-Da Silva Moore discussion about TAR.

Sheila Mackay is vice president at Conduent. She can be reached at

About the Author


Previous Article
The Madness of Big Data

To take a break from pure e-discovery, we wanted to focus on something fun. Now that Selection Sunday is be...

Next Article
Are China’s New Data and Cybersecurity Regulations a Wolf in Sheep’s Clothing?
Are China’s New Data and Cybersecurity Regulations a Wolf in Sheep’s Clothing?

As the Year of the Sheep gets underway, it is already apparent that the Chinese government will not be livi...