Sunday, 11 December 2011

The Ready Relevance Reckoner

You don’t know what predictive coding is?  Well that may not be too surprising, since one thing that almost all the contributors to an October 2010 Survey on Predictive Coding (published by the eDiscovery Institute) agreed on was that there should be a better term than predictive coding to describe what they do. 

That is a shame, since predictive coding technology promises to be the most important advance in tackling the mountain of electronic documents that we litigators now face when doing discovery. (Sceptical contrary view here)
In a nutshell, predictive coding tools enable the lawyers to review a sample of the overall document collection and decide whether each document in the sample is relevant or irrelevant.  The tool analyses the reviewers’ decisions and then evaluates the rest of the collection, giving every document in the collection a relevance ranking. 
Each tool has a slightly different approach, for instance some are more iterative, but that’s the gist.  The hope is that since the vast majority of any given electronic document collection is irrelevant, most of the documents can with a high degree of confidence be excluded from human review – an approach that chimes very nicely with the emphasis in the English procedure rules on proportionality.
So what can we do about the name?
My own flight of fancy is the title of this post.  But it is hardly a serious generic term, though you never know – relevance reckoners, anyone?.  But even that doesn't convey the message that these tools enable the user to identify irrelevance, rather than relevance.  Since the key point of the technology from the human point of view is the ability to start with a subset and end up discarding most of the documents, maybe a suitable generic term would flow from that – sample-driven culling, perhaps?  Irrelevance cullers? Garbage detectors?
Hmm, maybe we’re stuck with predictive coding after all.  It has to be better than one of the alternatives suggested by a contributor to the eDiscovery Institute Survey: ‘Prognostic Document Profiling’.

[Updated 1 May 2012.  Predictive Coding now seems to be morphing into Computer Assisted Review.  Not perfect, but much better.]