Migrated from eDJGroupInc.com. Author: Barry Murphy. Published: 2013-02-14 05:23:46Format, images and links may no longer function correctly. This past Tuesday, I had a chance to participate in the eDiscoveryJournal webinar, “TAR: From 10,000 Feet To Ten Feet, Let’s Get In The Weeds” (thanks to Kroll Ontrack for sponsoring it). Without any hint of hyperbole, I can honestly rank it in the top two webinars I have ever been part of.  You know a webinar is good when you have a hard time getting all the way through the content and the audience fires a continuous stream of questions throughout.  Not only that, but also virtually all the attendees stayed until the end, which happened to be five minutes past the end time.

I think the webinar had several things going for it.  First, it was not some pie-in-the-sky portrait of how Predictive Coding can save millions of dollars in review costs.  Nor was it one of those events that examines the various case law rulings endorsing/denying the usage of Predictive Coding.  Rather, we treated Predictive Coding as a simple reality and explored how to deal with it in pragmatic terms.

Second, the webinar featured two Predictive Coding rock stars.  David Baldwin is a Litigation Support Manager at Choate, Hall, and Stewart who has used Predictive Coding in the real world.  He is also a master at creating PowerPoint slides that illustrate a point.  Karl Schieneman is a Predictive Coding expert – literally.   Karl was the expert in the recent Global Aerospace case where Predictive Coding was utilized.  The key to a good Predictive Coding webinar?  Get me out of the way early and let David and Karl talk through how to confront issues around sampling, measuring results, dealing with seed/training sets, and understanding what relevance scores mean.

The webinar did, though, go beyond how to confront the very real challenges that Predictive Coding presents in practice – we did look at what current adoption rates are and what most users are looking to get from Predictive Coding.  I was able to present a preliminary analysis of some of our survey data.  Just a reminder: the Predictive Coding survey is still open and will wrap up next week, so if you haven’t had a chance to take it, please do (you could win a $250 gift card and, at the very least, you will get a copy of the summary results).  One interesting nugget coming from the data supports a theory of Karl’s – that Predictive Coding is the next evolution in culling methodologies.

PC Used Mostly For Culling

PC Used Mostly For Culling

One could argue that Predictive Coding is really fulfilling on the promise of Early Case Assessment (ECA), but I believe this is the broader category of TAR (technology-assisted review) that is the new ECA.  Predictive Coding is one element of it, but so too are mechanisms like concept clustering and faceted navigation.  What is exciting, is that 2013 will see TAR gain more and more systematic adoption and we will stop arguing about whether methods like Predictive Coding are defensible and start creating best practices.  The recorded webinar will be up on the eDiscoveryJournal soon in case you missed it.

eDiscoveryJournal Contributor and eDJ Group Lead Analyst – Barry Murphy

0 0 votes
Article Rating