June 18, 2013
Predictive Coding and Technology-Assisted
Review: Hype or Need?
Over the last few years, predictive coding has led the way as a hot topic.
Recently the new buzzwords have been
technology-assisted review and
computer-assisted review.
While these are nice words to throw around, questions arise: How many cases
have any of these been applied and what are the metrics of savings and time to
the client? How many cases actually require the use of these items?
I can recall that this was the way of future document review, but we find
only a few years later that predictive coding has taken a long time to grow.
This could be from various factors: hype information about the benefits and
risks, though it could be an important part of your e-discovery toolkit but you
will need to determine whether predictive coding is right for your particular
case and weight the cost savings from a finical point, compare what the cost
would be to implement vs. not to implement, your dead line and
resources.
In a recent article, “
7 E-Discovery Takeaways from CEIC,” three items stood out to
me: (1) humans will still play a vital role in document review, but technology
will assist in reducing the number of documents; (2) applying text analytics can
also reduce the cost and population of document review; (3) make sure you have
an agreement with opposing counsel on which technology you will choice. Here are
thoughts from the article’s author, Sean Doherty:
1. Technology-assisted review, computer-assisted review, or predictive coding
technologies will not replace human review, according to David Cohen, partner at
Reed Smith. He said that document review technologies will still need humans to
train systems and provide quality control. The undertow to the message: document
review is not a career path.
2. Applying text analytics to document review reduces the cost of document
review from dollars to cents per document. According to Cohen, human review, on
average, can cost $1 to $3 per document. With Equivio, Reed Smith charges clients 3 cents
per document after the system is trained, said Cohen. When asked how the firm
calculates the cost to train a technology-assisted review system, Cohen said
that Reed Smith charges clients by the billable hour.
3. If you can't get an agreement with your opponent on the use of
technology-assisted review, said Cohen, don't forge ahead on your own (see Kleen Products v. Packaging Corp.).
But there are still three use cases for TAR that don't require an agreement with
your opposing party: use it on the production set your opponent delivers to you;
use it to provide quality control to manual or human review; or use it to
accelerate human review where automated review systems can identify highly
relevant and irrelevant documents, which can be grouped or batched to speed
review.
My advice to end users is to consult with your internal support team, whether
it is the IT department or litigation support. Outline your goals and objectives
for each project prior to making your selection on which technology can benefit
your goal.
It is easy to become overwhelmed with information or want to use the newest,
best, and greatest technology in the market. However, you must consider that the
use of higher-end technologies come at a much higher cost. Consider the
following while making your choice: (1) remember that sometimes the most
cost-effective and efficient solution is also the most straightforward; (2) have
a well-defined scope of the data you are going to process and review, because
the days of "collect and process everything" are long gone; and (3) plan and map
out your strategy for the production and review.
A successful production and review features the following elements:
-
Communication and planning
-
Meet and confer
-
Bring in the experts (internal resources, vendor, review team)
-
Outline: scope, goals, deadlines
-
Daily progress and quality control
checks
Another helpful practice is to reduce data prior to production and review.
This procedure includes the following:
-
Strategic collection of data
-
Strategic filtering of data
-
Keeping your ESI in a native format
-
Apply text analytics
—Robert
Grande, codeMantra, Plymouth Meeting, PA
http://apps.americanbar.org/litigation/committees/technology/news.html#01