eDiscovery Leaders: Martha Louks of McDermott Will & Emery

Each week on eDiscovery Leaders Live, I chat with a leader in eDiscovery or related areas. Our guest on December 4 was Martha Louks, Director of Technology Services at McDermott Will & Emery.

Martha and I talked about tying together structured and unstructured data, building scalable timelines from the information, and using that content to build a fuller picture of a case. We discussed Martha’s eDiscovery technology wish list, including applying advanced active learning to short messages, having early and ongoing discussions with the data, and analyzing sentiments in yet more ways. We also touched on deployment models, working with audio and video data, and the need for new standards to deal with the data coming from the different ways COVID has forced us to work.

Recorded live on December 4, 2020 | Transcription below

Note: This content has been edited and condensed for clarity.

George Socha:

Welcome to eDiscovery Leaders Live, hosted by ACEDS and sponsored by Reveal. I am George Socha, Senior Vice President of Brand Awareness at Reveal. Each Friday morning at 11 am Eastern, I host an episode of eDiscovery Leaders Live, where I get a chat with luminaries in eDiscovery and related areas.

Past episodes are available on the Reveal website; go to revealdata.com, select “Resources”, and then select “eDiscovery Leaders Live Cast”.

This week we have joining us Martha Louks. Martha is the Director of Technology Services at McDermott Will & Emery in D.C. She focuses on implementing high-value and efficiency-driven solutions to improve the delivery of legal services for clients; collaborates with legal teams to develop technology and workflow approaches that align with case strategies, reduce costs and improve efficiency; has extensive experience developing defensible processes that center on the tailored use of technology and experienced professionals to achieve results for their clients; and evaluates the efficacy of a wide range of technologies to determine which tools are best suited to a matter’s unique needs.

Martha has a B.A. from Albion College. (We were just chatting before this because it’s a place I’ve spent some time at over the years as well.) In terms of her work background, she’s been with McDermott Will & Emery for over 10 years, starting as a Technology Project Manager and leading up to her current position. She’s had a number of different roles, seen a number of different facets of how all of this plays out in a law firm, three years before that as an account executive and project manager at DTI, and then two years before that at Lord Bissell & Brook as a paralegal. So, Martha you’ve seen various different aspects in all of this.

Martha Louks:

Yes, I’ve touched a lot of different areas from being a paralegal – and actually I worked a lot in paper back then, which was an interesting experience – and then going into the vendor side, working at a law firm and it’s all added up to where I’m at now.

George Socha:

In preparation for this call, we had a discussion or two. I sent you a list of potential topics. But just before we got on you said, “Go ahead and surprise me”. So, surprise you I will. I’m going to ask you essentially the same question I asked in an earlier episode of Ron Best, which is: Set aside what exists, what doesn’t exist in terms of eDiscovery technology. Set aside what you think is possible versus what you feel is impossible, practical versus impractical. If you could have the ideal eDiscovery application or platform – think as broadly as you want, anything is possible – what would it look like?

Martha Louks:

That’s a big question. That’s a good question. Gosh, I have to think about that for a second. A lot of platforms have really nailed the review needs, the production needs. I think we’ve got a good handle on that as an industry. I don’t have a lengthy wish list on the review side. When it comes to things like short message data, I think there’s a lot of room for growth in the industry.

What I would love to see is a platform that somehow lets us collect and consistently format data versus like Slack or Jabber, Zoom chats, Microsoft Teams, which has seen enormous adoptions since COVID occurred. And being able to grab that stuff and work with it in a platform, so we can review it in a meaningful way. There are so many different data challenges associated with this. It needs to be consistent and repeatable yet also flexible, because we’re going to have to deal with an avalanche of this type of data in the next few years. I think this type of data is the big challenge facing our industry for the next, probably at least five years or maybe longer. This is going to continue to proliferate and we’ve got to find really good solutions for dealing with this in review. What would that look like in practice? I actually don’t know, George.

George Socha:

Well, let’s approach it this way. In this imaginary platform we’re thinking of, anything is possible. So set aside any limitations, what would you like to be able to accomplish?

Tie It All Together

Martha Louks:

We need to be able to leverage our natural language processing analytics to be able to effectively perform technology assisted review or these kinds of conversations but they’re weird, because they’re really short and they’ve got attachments. Are you breaking it down by a whole entire conversation from the beginning to the end? Is a day document or not? I think we’re moving out of the document world and need to think about ways of incorporating short message data, which is basically structured data. How do we make structured data work in a different database?

George Socha:

You want structured data and unstructured data to co-exist happily in this platform, so that you don’t need as a user, to think about which it is, you can go after what you’re looking for?

Martha Louks:

Correct. Then ideally, what you would want to do is seamlessly show a timeline of a person’s activity. If I’m collecting their phone and their chats and their emails, I can see maybe you’re talking with this person on email and then they take the conversation off to chat. People have a different way of communicating over chat, so you might want to see whether the sentiment changed or things like that. I would also want to know, are they talking on the phone? Maybe you can incorporate call logs or other features of data like that; there is an almost perfect record at this point of people’s electronic activity.

That we really need to do is harness the power of a platform to tie these sources together, so we can see one continuous line of activity and understand what they’re doing where they’re doing it because people are simply not working inside of one platform anymore, like email. That’s not the only source of communication, so we’re not going to get the full picture. The whole point of discovery is building facts, determining what the facts of the case are. It’s very difficult to do that unless you’re actually looking at all of the information in a way that ties it together. If we don’t find a way to work with this type of data, we may actually, I think ironically, find more manual review needs rather than less – if we’re not able to use technology effectively to comb through it.

George Socha:

With what we know is a growing volume of data, a growing variety of types of data, even different types of e-mail messages, different types of short messages – e-mail versus short mail, versus recorded video communications, versus all of that – you’ve got all of that complexity and richness.

At the beginning of a matter, what would you like to be able to do with a platform? Let’s assume you’ve got to the point where you’ve got some data, you’ve pulled it in, the system has made it all magically completely accessible so that you can work with it in any fashion you want. You’re there on one end, the data is there on the other end, how would you want to be able to interact with that data?

Apply Advanced Active Learning to Short Messages

Martha Louks:

I think eventually it would be really great if you could do a broad search across it, conceptual searches looking for concerning areas within the data, but being able to see activity. If I have a search result I would want to see what are they doing immediately before and after. I think what we’re going to end up needing is more like a contextual search result because if you’ve ever worked with some of these little short messages, you might run a search term, let’s say a basic search term, and it’ll hit on one message. You have no idea what they were saying before and after nor do you know what they’re doing with other platforms, so some contextual search would be really helpful to get started on some of these things.

I think it would be really cool if there was like a technology assisted review that would score the documents – they’re not really documents – it would score the communications at the individual message, maybe a range of messages level, and an entire conversation level. That way you can start slicing through the day.

It would be really cool if it automatically adjusted the scores in the algorithm based on how the attorneys are dividing these conversations to produce. If I’ve got a year of a conversation, maybe I only need to produce two weeks’ worth of it, that’s the only relevant portion. Am I doing technology assisted review on everything for two years? I probably need to, to determine what’s responsive, but it would be really neat to have just different ways of classifying these messages because there’s got to be some flexibility built in.

Have Early and Ongoing Discussions With the Data

George Socha:

So much of our industry’s attention has been focused over the past 10, 15, maybe even 20 years on trying to optimize the process of reviewing content for production: finding relevant content, weeding out privileged content and producing what remains. That, however, is only a small part of what lawyers do in the life of a lawsuit and what they need to do with content. What would you like to see lawyers be able to do with this technology through the life of a lawsuit?

Martha Louks:

It would be great if they could go in and just interrogate the data directly and say, “I’m looking for something that supports an argument like this”. If you could tell the system, “Here’s what I’m aiming for, find me documents that might potentially support what I’m trying to say”, that would be really cool. Almost like asking somebody familiar with the documents, “Hey, can you go and find me all the stuff that John Doe did related to issue one?” It’s the same kind of thing except you wouldn’t have to rely on a search term, it would bubble the documents up to you. It might be neat to put in even pleadings and things like that, draft versions, and see what it might do to auto-cite things or automatically imbed exhibits and things like that and things that are conceptually related to what you’re arguing.

George Socha:

If you are, at least conceptually, sitting on one side of the table, the computer broadly defined is sitting on the other side of the table, and you start posing it these types of questions, what would you like its responses to be like when they come back to you?

Martha Louks:

I think it may be great have a summary at the top that says, “We found x number of documents related to these thing within the set, here are the people that are talking, here are the companies that they’re mentioning, here are the primary concepts, and we think these five would be your best place to start. We’re giving you a hundred documents, but start with these five”. Then as you’re going through the results, it’s constantly updating that search, just like a continuous active learning, so that you are teaching the system and building on your original question and as more documents are fed in, you can continue to update it.

I think it would have to show the value of artificial intelligence – certainly classifying large volumes of data and using it that way, but the biggest thing that attorneys can use on the front end is seeing what the system has identified for patterns in the data. That gives you a smart place to start; you’re not just throwing terms out there and seeing what hits. To me that would be a really great way to build out a more fulsome type of analysis.

George Socha:

I think it would be wonderful to have those types of capabilities. Maybe we’ll get there. Who knows; we can try.

Martha Louks:

I think we might get closer.

George Socha:

We’ll jump in a slightly different direction – what about clean uniform standards for data storage?

Martha Louks:

The way I see it, the cloud is where this is all going. I think there is an appetite for that. As we use more and more analytics, we are going to need the cloud to rent compute power, it’s just going to require more and more resources. With that, I think a lot of storage would be managed on the backend, whoever is reading the software and however they manage it in conjunction with the cloud. We’re going into an interesting area where, like I was saying with the short message data, it’s structured data, it’s not really a document, so there’s no native file necessarily. I think some storage ironically may go away a little bit, just because there’s not so many files. Email isn’t really documents either, email is technically a database.

Deployment Options

George Socha:

E-mail is its own form of structured content. It’s got various fields of metadata, some required, some optional, and it’s got one area where there’s a lot of text or very little text. Who knows?

You mentioned cloud. There are, at least today, four deployment models for eDiscovery platforms: Entirely in the cloud, entirely on premise, a hybrid of the two, and mobile. What are your thoughts about the pros and cons of those four different approaches?

Martha Louks:

That’s a good question. With the cloud, the pro is just being able to rent the resources that you need, as opposed to purchasing and maintaining them all the time, and having a clean and simple interface where the back end doesn’t expose to the users, it’s a little easier to keep your data organized, at least the back end functions, while you don’t have a lot of clutter on your servers and people working in different ways. There are benefits to that. With that said, you’re also not working so much on the back end, you really don’t have a lot of the flexibility that you might have if you had an environment on your own to work in. It also makes it a little bit more challenging to work with multiple applications.

I think there is no silver bullet choice in eDiscovery. There are different advantages to various platforms. They perform really well at some things and in some other areas another platform might be better. I think we get the best results when we combine tools and that can sometimes be better if you’ve got an on premise environment where you can architect things to your specific preferences and workflows. It really depends on how you work and what your objectives are, the types of cases, what you perceive as complex needs and what the limitations might be if you went all in on a single solution. I think different organizations could have a good experience either away.

Work With Audio and Video

George Socha:

Text is a lot of what we deal with in eDiscovery. Text in email messages, text in short messages, text from word processing documents and the like. But there are other types of content out there that we at least theoretically need to be dealing with, especially audio and video. How do you see folks approaching those? How would you like to be able to approach those?

Martha Louks:

That’s a really good point. Especially, I don’t know how many more recorded video or audio things there might be to discover in the coming months and years. Cloud platforms are really good with this, they can farm out the data to speech-to-text technology and bring that back in and then you can do all the things that you normally do with text. Having that automatically transcribed and put back into the database, is something that we probably would need to incorporate as a standard for a lot of platforms if it becomes a major thing that we need to work through all the time. I think it’s helpful with voicemails attached to emails and it’s time consuming to have to listen to those.

George Socha:

So, getting those transcribed would help immensely. What about the video component, how would you like to be able to handle the video part of audiovisual content?

Martha Louks:

There is image recognition technology where you can do searches for things that it’s auto-identified in pictures and I think that could likewise be applied to video. You could see what’s in the video, so they wouldn’t have to watch the whole thing. Then, like many applications have now, which of course is very helpful, when you play your video in the viewer it synchronizes with the text below or the side, wherever it is, so you can both read and watch and make sure you’re seeing it at the same point. That would be a good common functionality to have integrated. Certainly some of these image recognition things would be really cool because the more information we’re able to extract automatically from these files, the more we can leverage our other analytic tool to classify them, so we don’t spend all this time dealing with junk.

Analyze Sentiments

George Socha:

I’m going to take a little leap to the side here. Part of what you were talking about earlier was sentiment analysis. How do you see that used most effectively? What would you like to be able to do with that, that you can’t do with that today?

Martha Louks:

Oh my gosh, sentiment analysis is one of my favorite things. It’s a great starting point if you are at the beginning of a case and you’re not sure what’s in the documents, you want to see “How strong is my case, how weak is the other side’s case, are we in a bad position?” You can start with high negativity documents, high pressure documents. Those are the best places to start. Sometimes it could be very positive documents, if people are maybe doing something that benefits them that’s wrong.

There are a lot of different things that we can do, but what sentiment does not currently do: It’s really bad at picking up sarcasm. In text, you cannot tell whether somebody is joking. A computer can’t tell that but we can, by reading it.

There is potential for being able to detect this and I think we need to incorporate emojis into the emotional analysis of the data. For example, I can identify sarcasm by finding like a negative statement coupled with a winky face or a smile emoji. Those two conflicting sentiments might indicate a joke or sarcasm. I do think emojis should really start to be incorporated into some of these analyses and it can really be helpful in finding and determining what people’s intent might have been at the time when they wrote that thing.

George Socha:

I wonder if even with the currently available tools, if you look at something where there’s both positive and negative sentiment within the same communication, if that can start… I’m thinking off the top of my head here.

Martha Louks:

No, I know right. We could try it.

George Socha:

That’s right.

Martha Louks:

That’s a really good idea. We should do that.

George Socha:

Maybe we could build a model on that. Right?

Martha Louks:

That’s right you could. I wonder if that would be effective. It’s hard to say, because it depends on how it’s parsing the information right now.

Coming Next: New Standards for New Data Types?

George Socha:

I bet we can try it, we can test it and see how that turns out. Well, we’ve covered a number of things. What should I have asked you about that I haven’t?

Martha Louks:

We’re at a weird point in our industry. I was thinking about it this morning George, what would we talk about, and I remember distinctly, a couple, maybe two or three years ago, I talked to a bunch of folks that said, “I feel bored in eDiscovery, it just feels boring”. I think that is not the case.

I see a lot of disruptions from the corona virus like everybody is seeing, and I think we need to be gunning for really cool and different solutions to deal with all these remote platforms – when people went remote they started using different platforms. I guess it’s really just about the disruption in the industry that’s coming up. I don’t know what’s coming, but I think it’s going to be very different. I think we’re seeing companies working on solving these problems and someone’s going to come up with some really good solutions.

And I also think, frankly as I’m talking out loud, thinking out loud, I do think we as an industry, we’ve got to come up with some agreed-upon standards for dealing with some of these new data types. We can’t control how they will be formatted when we collect them because that’s in the hands of the companies developing them, but like email comes out in a few typical formats. I think we’ve got to start focusing for now at the beginning. If we go through the EDRM left to right, when we’re collecting data, before we process it, we’ve got to get it into a usable format that once it’s processed will actually function and do all of those amazing things that we are talking about. The foundation for it is clean and consistent and agreed-upon formats and standard practices. I think we really, as an industry, need to have a conversation about how we’re going to meet the needs of litigation and deal with these data challenges and still make it usable and work with the technology we’ve already developed. We’re not quite there yet, so we’ll have to see how that unfolds in the next couple of years.

George Socha:

It gives us a great challenge to work toward.

Martha Louks:

Absolutely. Not boring at all.

George Socha:

Thank you Martha. It has been wonderful. Martha is Director of Technology Services at McDermott Will and Emery. I’m George Socha. This has been eDiscovery Leaders Live, hosted by ACEDS and sponsored by Reveal. Thanks for joining us today and please join us again next week when we talk with Luke Smith, who is CEO and Head of Customer Delivery at Legal Tech Innovations (LTI) in Belgium. Thanks again Martha.

Martha Louks:

Thank you George.