Laozi not Lousy: A Modern Day Argument for Predictive Coding
“Those who have knowledge, don’t predict. Those who predict don’t have knowledge.” – Lao Tzu
One of the myths of Lao Tzu (the father of Taoism) is that he was born as a grown man. He came out of his mother’s womb as a 62 year-old man with a full grey beard and long earlobes, both symbols of wisdom and long life in Chinese culture. If you’ve read the Tao Te Ching, or if you’re like me, the abbreviated “Tao of Pooh” (Winnie the Pooh, that is), you know that he then spent the rest of his life (900 years) teaching people to chill out, share and generally how to be happy through a change in approach. Change the way you see life and you will change the way you live life.
I feel like Predictive Coding is a lot like Lau Tzu – it’s been incubating for so long in our minds and review tools that when people do start using it, it too will emerge with a full beard and long earlobes. Why are so few people using Predictive Coding workflows? Why does everyone want to just talk about Predictive Coding? The reason is because it’s scary not putting eyes on every document and because the approach is cool, interesting, inventive, innovative and inexpensive (comparatively speaking). Everything which document review, traditionally, is not.
In every meeting that I go to, Predictive Coding is a big topic in the discussion. How do we do it? What tools and workflows do we use? What assurances do we give of risk mitigation? What are other clients doing? I talk about Predictive Coding pretty much every day. But much like Lao Tzu, when I talk, I mostly talk in parables or stories of what could be, if you only take a different approach. Until Predictive Coding becomes the norm, or at least more widely used, I might as well be spouting ancient Chinese proverbs.
Confucius says, “Man who uses Predictive Coding has lower review bill”
Confucius says, “Man who embraces Relativity Assisted Review sees many statistical model reports”
I was recently in a meeting in NYC with a large corporation to talk about e-discovery outsourcing, including forensics, processing, hosting and managed review. I got about 30 minutes into the meeting, finished with my fairly technology-centric pitch, when one of the women in the room casually mentions that they still mostly deal with paper.
(cue crickets chirping here)
I took a beat, switched gears (glared briefly at the sales person who brought me to the meeting) and answered their paper-related questions. We then talked about scanning, setting up war rooms for paper review and inventories, and sending review teams on-location to their archive warehouses, all the while trying not to be too technical because the people in the room were really dealing with discovery, not e-discovery. And yet, even these people, these behind-the-curve “paper” people, ended up wanting to talk about Predictive Coding anyway. I almost laughed out loud when they asked.
I love talking about Predictive Coding, and yes, that’s dorky, but I do. I think it’s cool and interesting and innovative and all the rest of that stuff, so I answered this man’s question – fully and politely, and as if I was talking to my mom. Because this was Ray Kurzweil talking to Beatrix Potter. This was city mouse talking to country mouse.
As LegalTech New York 2014 rapidly approaches, I find myself asking again – what will it take for law firms and corporations to embrace Predictive Coding? What do we need to do to get this train out of the station? These are questions I’ve been asking myself for years now and I think the only real answer is: change. We need to change. The fundamental approach to document review needs to be altered. People need to stop saying that traditional human review is the gold standard and they need to understand that TAR is the gold standard. Humans are still essential, yes, but the use of TAR is a game changer. People need to stop thinking about folders and start thinking about social networks and how they communicate today. They need to stop seeing the EDRM as a linear set of process steps instead of viewing it as a fluid, rapidly evolving and changing data and document management process. They need to stop thinking of a document as just a document and more as a piece of information. They need to start applying e-learning and gaming principals (i.e. the utilizing of fun and addictive game-like features and achievements to promote greater efficiency and productivity) to every step in order to better manipulate the data and gain an improved understanding of the project universe. Because data is everything today – what we have, what we don’t have, and the stories we are able to construct with every MB, GB, TB or PB.
If algorithms can group together similar documents for us – and they are good at grouping – can’t this only help? And if algorithms can group together documents that are basically the same, save a few differences (near dupes), don’t we want that? Predictive coding can be applied in the collection through to production phases and make us better all along the way. Is it risky? Maybe, but we can put safe guards in place to better protect against inadvertent production of sensitive or privileged materials and we can visually QC at every step to ensure “bad things” don’t happen. No one wants bad things to happen. No one.
I’ve been talking about Predictive Coding for five years now. And in those five years I have used several companies’ Predictive Coding tool kits and even helped design one for my own company and yet, very few clients have been willing to really utilize Predictive Coding for culling and/or review. Out of our many clients, only a few routinely apply Predictive Coding workflows. Now, granted, these are major clients and they make up 20 percent of our review projects, but the volume of clients is just not there for Predictive Coding. Despite all the reports hyping the technology! We do actually use Predictive Coding tools in most of our document reviews, but it is done as a prioritized review (train the system, propagate decisions then batch by Potentially Responsive or Not Responsive) or via triaged pricing workflows (much like prioritized review, but the Potentially Not Responsive results go offshore for half the cost). So the truth is few clients anywhere in this industry have embraced full-on Predictive Coding exercises. I’m afraid that Predictive Coding’s fate may be the same as Lao Tzu’s or the ill-fated Laser Disk from the 90s – exiled and replaced with a better model (note my Confucius sayings above were not Lao Tzu sayings).
Exploding data volumes and the shrinking of law firm staff are both really doing a number on the review phase of discovery. How are firms possibly going to get through millions of documents, one by one, consistently and accurately, and in 60 days? 30 days? 15 days? And how are they supposed to do it without spending millions of dollars? It’s crazy to think that this will even be possible with any approach other than the use of Predictive Coding. Pretty soon (2015? 2016?), Predictive Coding won’t be a question anymore, it will be the only answer. I’m ready to meet this 62 year old man.
Bring on the Laozi Review!