Conferences / Symposia, Contract Attorneys, Document Review, Federal Judges, S.D.N.Y., Technology

Live at LegalTech: That’s a Wrap, But the Predictive Coding Debate Is Only Heating up


LegalTech New York finished up earlier this week. I survived with only a minor case of technology loop, although my iPhone was begging for mercy by the end.

The conference was frenetic, to say the least. There was a lot going on, regarding a cornucopia of technological topics and tools to help lawyers. As expected, the biggest hype revolved around predictive coding and computer-assisted review.

The legal technology world has been buzzing about this stuff for a while now, and we have covered it on these pages several times before. (Here and here, for starters). At the conference, attendees got to hear from the naysayers, the enthusiasts, and everyone in between. Several panels helped explain exactly what the technology means on a practical level. And no, cyborgs will not be stealing all the contract attorney jobs any time soon.

One of this week’s highlights was a lunchtime panel featuring two prominent attorneys and a New York magistrate judge. The discussion helped clarify, demystify, and define the terms that have been making headlines (even in the New York Times) for a good part of the past year. Is computer-assisted review as scary as it seems? Of course not.

Let’s see what the panelists — and at least one irate audience member — had to say….

The panel, which consisted of Wachtell Lipton counsel Maura Grossman, Magistrate Judge Andrew Peck (S.D.N.Y.), and Jackson Lewis partner Ralph Losey began by examining all the problems with human review.

Grossman explained that humans — a.k.a. contract attorneys locked in a room — usually only find about 70 percent of responsive documents. They also turn out a significant number of false positives.

“It makes us uncomfortable,” she said. “It makes me uncomfortable. So we pretend it’s perfect.”

Losey and Judge Peck agree. In Monday’s Quote of the Day, we mentioned their strong opinions about older keyword search technology.

“We are going to price ourselves out of business as a court system if we don’t change,” Judge Peck said.

For those new to the party, Grossman gave a thorough explanation of what computer-assisted review actually means.

Basically, the computer comes up with document review seed sets and gives them to humans, who say yes or no, or tell the computer to systematically alter what it finds. Grossman compared it to visiting the optometrist.

“A or B. B or C.” You continue choosing the better option until they are so similar, she said, that you can no longer tell a difference.

Grossman said it can bring the discovery down to one-tenth the price of what e-discovery often is. And when/if errors happen, the types of errors are different than human errors. Technology makes systematic errors, she said. The problems can be fixed by telling the computer to change its formula for finding documents. Humans make random mistakes, which are harder to consistently eliminate.

Obviously, not everyone is convinced.

During the question-and-answer period at the end of the presentation, someone in the crowd said, “I’m very disappointed in this presentation. We were promised a debate, but we have three for the machines and zero votes for the humans.”

At first it seemed like a joke, and the panelists were waiting for a punchline. The man’s view was very different from what everyone had heard onstage thus far. And he had legitimate concerns.

In a Tuesday morning panel that included in-house counsel from DuPont legal and the Dow Chemical Company, as well as Ari Kaplan and representatives from Dell and FTI technology, the general consensus was that despite all the hype, attorneys are still nervous about the “embryonic nature” of predictive coding. Concerns are less about the cost, and more about the (entirely reasonable) fear of f***ing up a major case. Legal departments are more comfortable with using it for internal investigations.

The discussion should not revolve around whether courts have approved it, they said, because courts never approved keyword searches and people used them anyway. It is simply a matter of comfortability. Whatever technology counsel decide to use for discovery, they need to feel comfortable enough to defend it in court.

We shall see where this goes. Predictive coding is definitely not going away. If any of our readers have opinions about computer-assisted review — is it the panacea? Will it destroy lots of jobs and ruin everything? — let us know in the comments, or send an e-mail to cdanzig [AT] abovethelaw [DOT] com with “Predictive Coding” in the subject.

Now if you’ll excuse me, I’m going to go sit in a dark room with no electricity and unscramble my brain.

Earlier: Quote of the Day: Keyword Searching? You’re Doing It Wrong
Live at LegalTech: No More Boxing Robots

4 comments
(hidden for your protection)

comments sponsored by

Show all comments

Our Sites

  • Above the Law
  • How Appealing
  • ATL Redline
  • Breaking Defense
  • Breaking Energy
  • Breaking Gov
  • Dealbreaker
  • Fashonista
  •