When some attorneys hear "technology-assisted review" (TAR), they may think of a machine completely taking over the document review process. But as experts at the "Choosing a Predictive Coding Approach—Predictive Coding For Dummies" panel at the ILTACON 2019 conference explained, that's far from the truth.

Even though TAR has evolved to become faster and easier to use, there is still a lot of manual work involved, as well a lot of reliance on vendor help. So while it may be easier to conduct e-discovery than ever before, it's still no walk in the data park. Here's a look at three insights that help shed light on state of TAR today:

|

TAR 2.0 The Way to Go?

The verdict is in for TAR 2.0. The advanced TAR process is synonymous with continuous active learning (CAL) in that it learns how to identify relevant documents by continually being fed varying degrees of responsive and nonresponsive documents, and it remains the preference of many attorneys and e-discovery practitioners.

"I like the tech better because it feels like how I'm interacting with [artificial intelligence] in my everyday life," said Catherine Casey, chief innovation officer at DISCO. She explained that TAR 2.0 is similar to the AI recommendation engines powering platforms like Spotify and Netflix.

Julian Ackert, managing director at iDiscovery Solutions, noted that speed is one of the biggest benefits with TAR 2.0. "You can get started right away… [and] you can begin to classify [documents] based on different characteristics a lot quicker."

Still, this isn't to say that TAR 1.0, which requires that subject matter experts first build a seed set of relevant documents to train the machine before TAR can be deployed, is completely outdated. "The thing to remember is TAR 2.0 isn't here because TAR 1.0 is wrong," Ackert said. While it's more labor intensive and expensive to use, at the end of the day, "there are still teams that use TAR 1.0." One reason for this, he said, was that the use of TAR 1.0 for document review may be allowed by some regulatory agencies, who have yet to approve TAR 2.0.

|

Deploying TAR Still Requires Manual Work

Despite being able to automate a lot of document review, TAR still requires some manual hands-on work. "The thing about TAR is that it's not 'button pushing' as much as people want to be it button pushing," Ackert said.

For one thing, TAR, whether 1.0 or 2.0, will vary in accuracy depending on the data set it's trained on, making manual curation an important requirement. What's more, e-discovery teams have to constantly decide what workflows and data they want with any TAR review.

"TAR is a little bit of a moving target, because when you first go to the documents, you have one view of what is going on, then you discover more stuff and you realize you're looking for other things," said Lia Majid, CEO and founder of Acorn Legal Solutions. She added, "TAR 1.0 and TAR 2.0 are different approaches, but even within TAR 2.0 there are different approaches."

Because of the many directions a TAR process can go, DISCO's Casey advised that, while anyone can try TAR, "it's good to invest in someone who understands it" as a guide.

|

Vendor Transparency, and Honesty, Matter

While TAR experts may be needed, the right ones aren't always easy to find. Doug Matthews, partner at Vorys, Sater, Seymour and Pease, stressed the importance of having a "really good partnership with your vendor" and being on the same page about what type of TAR process should be used. Yet this can be much easier said than done.

"Sometimes sales people will throw out terms because that's what they think you want to hear," Matthews said. He recalled a situation where his team wanted to use continuous active learning (TAR 2.0) on a document review, which the vendor said it could do. However, it turned out the vendor only had the capability to do TAR 1.0, much to the surprise of Matthew's teams.

"You need to dig deep into what the workflow will be when working with a vendor to make sure it's the one you want," Matthew said.

But it's not just process that clients need to be conscious of, but also TAR pricing as well. With vendors, "it's very difficult to compare prices, as a lot of them charge for different things" such as hosting fees and project management fees, he said. The best case scenario then, is to understand everything as much as possible upfront before diving in.