UPDATE:

Sorry for the quasi-necropost, but I just stumbled upon relevant case law.  In reading Cecil Lynn‘s excellent recent article on Law.com, Drama & Destruction, that provides a great rundown of 2010 case law, I came across this case summary:

In Ross v. Abercrombie & Fitch, defense counsel argued unpersuasively that the defense had no obligation to search for or locate known documents that did not turn up using the parties’ agreed-upon search terms. 2010 U.S. Dist. Lexis 47620, at *11-14 (S.D. Ohio May 14, 2010).

I haven’t read the case, but this seems to support the proposition that a producing party’s obligation to produce relevant materials is not limited by the application of search terms.  In other words, search terms are not a shield to producing known, relevant documents. Thank you very much.

ORIGINAL POST (2009-09-18 12:22:33):

This isn’t exactly timely, but it’s been on my mind for months, and I wanted to share and get your feedback.

I had the pleasure of attending The Sedona Conference® Institute this past March in Philadelphia.  During the last day lunch, a particularly interesting conversation started up at my table.  I was sitting with people I had never met before and probably wouldn’t be able to pick out of a crowd now, but we managed to have a brief and interesting discussion about keyword searching and the obligation to produce.  I don’t remember how it started, but the terms of the debate were this:  Is there an obligation to produce responsive data that was not hit by negotiated keyword terms?  That is, you have positive knowledge that responsive documents have been excluded by keyword searching.  Are you obligated to produce them?

I thought, and still do, that there is absolutely an obligation to turn the documents over.  Keyword searching is a method for finding responsive documents that are mixed in with a morass of non-responsive documents.  But, if you have a collected group of documents that are responsive, there’s no need to dump them in the unsorted pile in the first place.  They should be set aside for production without having to be keyword searched at all.

I casually shared my opinion with the table and saw several nodding heads, but I was surprised to find that two gentlemen did not agreed at all.  Their view was that, if the keywords were negotiated, then the results are the results and there’s no obligation to turn over anything not hit by them.  One gentleman (a litigator, if I recall correctly) flatly said he would not turn over the responsive documents.  The other gentleman (a vendor, I think) rather snidely remarked something to the effect that ‘You wouldn’t tell the other side what to ask during depositions, would you?’  I agreed with that but thought it was a specious analogy.  Not wanting to ruin a  pleasant lunch with a heated debate, I let the discussion go, but it’s been eating at me ever since.

Compare the view of these two gentlemen to the view of those of us who use sampling techniques to test the accuracy of keyword searches.  When testing for false negatives (exclusion of responsive documents), many people are of the opinion that even one false negative requires that the whole pile of excluded documents be manually reviewed.  On the other hand, for the gentlemen at my table, it would make no sense to ever test keyword search results, because they wouldn’t produce any false negatives they found.  To me, not producing documents you know are responsive just because they weren’t hit by negotiated keyword searches is like using keyword searching as a shield.  That not only violates the principles of cooperation but amounts to bad faith.

So which is right?  Am I being naive and Pollyanna-ish, or do these two guys not get it?  Or maybe it’s a little of both?