Should courts accept e-discovery productions generated by concept searching? 

That’s the question addressed in When E-Discovery Is Put to the Test (Pennsylvania Law Weekly, 14 May 2008). This article takes a deep dive into a potential Daubert challenge to using concept searching. Some lawyers will find it tough reading but, as I argued in Lawyers’ Duty to Learn E-Discovery Search Techniques, that may be the price of admission for some litigators.

This article adds to an important professional discussion about e-discovery tools and it is worth reading. I take issue, however, with the author’s analysis, which contends that courts cannot evaluate concept search tools because the tools are proprietary. I see three issues with this assertion:
1. In my experience, only computational linguists can understand advanced concept search algorithms. I’ve tried reading academic or vendor literature on how the algorithms work and cannot get past the first paragraph. (Let the record show I have three years of college math major classes plus I worked professionally as an econometrician.) So even if you could get at the source code, it’s not easy to evaluate. Plus, my past conversations with computational linguists and software developers suggest that they typically do not agree on the efficacy of algorithms.
2. Even if everyone agrees an algorithm works, that does not mean the software works. You still have to show that it was translated into error-free computer code. It’s a rare piece of complicated code that is error-free.
3. Even if the code can be vetted, it may have to be vetted on a specific operating system or even on specific hardware. And modern software often interacts with other pieces of code – does all associated code need to be examined as well? (Along these lines, remember the problem Applied Discovery experienced.)

So the proprietary nature of the software seems only a piece of the problem. I would try to side step the whole Daubert issue . In my view, Judge Waxse’s comments at a Legal Tech session – legal system requires reasonableness, not precision; plus costs must be reasonable – offer a better framework for thinking about the most appropriate tools for e-discovery.