Few people have had as much of an impact on modern eDiscovery as Judge Shira A. Scheindlin. As District Judge for the Southern District of New York, her series of groundbreaking decisions in Zubulake v. UBS Warburg in the early 2000’s helped create modern eDiscovery practice, shaping a generation of judges and litigators. She’s remained a leader in the field, even as recent amendments to the federal rules have moved away from the regime established in Zubulake. Judge Scheindlin retired last year, after 22 influential years on the bench.
Logikcull recently spoke to Judge Scheindlin about access to justice, preservation and proportionality, and the future of eDiscovery, including the court’s role in evaluating ever-changing technologies. A transcript of the first part of that discussion, edited for clarity and brevity, appears below. Part II, on the Supreme Court's recent Goodyear decision, sanctions, and more, is available here.
Logikcull: I'm sure you're aware that there's a very small and declining percentage of civil cases that go to trial in the federal system. Meaning, most of the disputes are settled in the discovery stage if not in summary judgment. To what extent you think that discovery system is fashioned to and capable of ensuring that justice is served when so few cases actually make it to trial?
Hon. Shira A. Scheindlin: Only two to three percent of federal civil cases go to trial. Most cases are either disposed on motions to dismiss, motions for summary judgment, or they settle. Assuming it's not dismissed on motion, the parties will settle based on the discovery that has been revealed during the course of the litigation.
In this country, discovery is very robust. People have to respond to a party's request for information. If they don't, they don't at their peril.
I believe that most parties do not hide things; they actually disclose what they're required to disclose. Everybody has a sense of what the evidence is and then they make a decision that they really don't want to face the risk of trial or what they predict the outcome of the trial will be and so they settle. So, I do think that our discovery system is capable of leading to just results.
"I do think that our discovery system is capable of leading to just results. That doesn't mean every settlement is a just result."
That doesn't mean every settlement is a just result. Sometimes, I’m very aware, defendants settle merely to avoid the continuation of litigation or the risks of a runaway verdict. They make that business decision. They feel like they did nothing wrong at all, but the settlement makes more sense that proceeding. There are certainly those entities in corporate America that think that the discovery practice is too robust, too expensive, and it leads to unjust results.
So, there you have it. It depends on whose ox is being gored. From the plaintiff's perspective, learning everything they can about the case and then using that to discuss settlement is a good thing. From a defendant's perspective, sometimes they don't even want to go through the process. It's too long and too expensive and they rather reach a settlement early without full discovery.
Logikcull: Speaking of that, producing parties, as you know, often complain about the costs of preservation and discovery. The new revisions to the Federal Rules of Civil Procedure bend towards those concerns, in the recent amendments. With respect to the preservation obligations in particular, how do you as a judge weigh the cost burdens that data-rich producing parties face with requirements that requesting parties are entitled to broad and liberal discovery? How do you go about that balancing?
Scheindlin: Well, I think you're referring to the proportionality requirement that now defines the scope of commensurable discovery. Up until these recent amendments, a party was entitled to information that was relevant to a claim or defense. Before that, it used to be relevant to the subject matter, but that was narrowed some years back to relevant to a claim or defense. Now, that's not enough. Relevance is not the only test of scope. Now, it has to be both relevant and proportional.
So, in general, proportionality is meant to take into account the burden of expense that you referred to with respect responding to discovery request.
Your question asked me, with respect to preservation obligations, how does this weigh out with regards to data-rich producing parties. I don't think it will help too much. There are comments in the advisory committee notes to Rule 26 scope of discovery that say proportionality even applies to preservation, not only to search and production. But I suspect that's not going to be brought to the court's attention very often.
"Parties are going to be fearful of not preserving everything that they think is relevant, without too much concern to proportionality."
Parties are going to be fearful of not preserving everything that they think is relevant, without too much concern to proportionality. The proportionality is going to come into effect when the discovery request comes in. They'll have to decide what to search and what to produce. I'd think they'd be hesitant not to preserve when they reasonably anticipate litigation or when litigation is actually brought. They'd be very hesitant to make a proportionality determination without having the ability to get a court ruling. When litigation is reasonably anticipated, it's long before there is any court in the picture or any lawsuit in the picture. So, they have no judge to turn to; they have to make that decision for themselves.
I suspect that business entities will continue to broadly preserve until there's a lawsuit, until there’s a document request. Then, if they think a request is not proportional, taking into account the factors of Rule 26(b)(1), then they can make an application to the court to render discovery based on proportionality.
Logikcull: With regards to eDiscovery in particular, one of the big challenges related to proportionality decisions is the limited visibility that courts have into the actual cost of eDiscovery and the eDiscovery technologies that are at work. So, judges may face situations where, for example, a producing party claims they can review and produce a huge amount of data by a given a discovery deadline but they will need very expensive technology to do so -- perhaps, utilizing a proprietary “black box” algorithm. In such instances, how do courts ensure that the information off of which courts are basing their proportionality decisions is accurate?
Scheindlin: You know, I think they do it the same way they do it on any disputed issue and any lawsuit.
"A court cannot be an expert in every subject matter that comes before the court. So, they listen to what witnesses have to say and they judge credibility."
A court cannot be an expert in every subject matter that comes before the court. So, they listen to what witnesses have to say and they judge credibility. If there's really a dispute -- in other words, if a producing party says, "In order to do this, it will cost me X hours of manpower and X amount of time,” and the other side says "That can't be right, judge. We're prepared to challenge that," then there'd probably be a hearing at which each side will call witnesses to explain what the process is and how costly it is. They might even bring in an expert to testify on their behalf, each side. Then, the court would listen and the court would judge credibility, make a determination based on the evidence.
Now, the judge is not an expert in computer science and the judge is not an expert in high technology of today. So, the judge does her best having heard the evidence. I think without it, if there's no challenge and the producing party says "here's our cost, here’s how much time, money, and manpower it'd take", if the other side doesn't really contest that, then the judge would make a proportionality analysis assuming that what was presented to her was accurate. If the other side contests the representation, then there would probably be an evidentiary hearing.
Logikcull: Let's say there is some sort of objection and there is a hearing, do you think that the Daubert standard should apply to determinations of whether specific search technologies are used, to claims about search technologies’ abilities? For example, would Daubert or Rule 702 apply to expert claims about predictive coding?
Scheindlin: Well, yes. If each side were to proffer an expert on the subject, then any expert of any type is always subject to a Daubert challenge. Again, if each side hires an expert to testify to these issues and neither side challenges the other expert under Daubert, I don't think the judge would do that sua sponte. The judge would simply listen to the two experts and have to determine which expert she believes, or believes more than the other. If there's a challenge under Daubert where one side says, that person's expert is not qualified or hasn't looked at the facts of this particular effort or hasn't used peer-reviewed studies or hasn't tested it -- whatever the challenge may be, I think the court has to determine under Daubert whether to let that expert's testimony be evaluated.
I don't think it's as hard a question as I know some judges who have been writing about whether Daubert should apply or not [think it is]. The point is, I don't think the judge is going to do it sua sponte.
Logikcull: How do you think the courts’ gatekeeping function may evolve as courts are forced to consider more advanced technologies such as artificial intelligence?
Scheindlin: The only answer, again, is what always happens in courts: They will become educated on those issues, partly through education & training that judges are always receiving through the Federal Judicial Center on the federal side, but also because parties will have the obligation to educate the judges so they can make good decisions in that area. That may include contested hearings, as I've mentioned earlier. I realize technology is changing and it's advancing faster than the judges know, but there are ways to educate the court.
Logikcull: Is there enough continuing education for judges around, particularly, discovery issues and technology? Do you think that is something that is adequately handled by the system where parties will advocate in a way that informs the court best? Or, do you think that there could be more opportunities for judges and courts stay on top of the latest development on their own volition?
Scheindlin: I think the federal courts are very lucky to have a robust training and education program for all magistrate and district judges. They do have the opportunities regularly to receive training in these areas.
I don't think that is as true on the state court level. The state courts obviously, first of all, there are 50 of them; secondly, some are wealthier than others and many don't have the funds to really do the kind of advanced training that one would need to understand difficult technology.
It's a different answer depending on federal or state court. Of course, most litigation is in the state court. We who came from the federal system think we have a monopoly -- we have anything but. Most litigation is in the state court. So, that remains an issue.
"[I]f you can't get decent training and education through your court system, then really it's up to the parties to educate a judge..."
But if you can't get decent training and education through your court system, then really it's up to the parties to educate a judge and if they don't do it well, I think the judge can stand on that, saying “I'm sorry, but you have not made a case for why this is not proportional, [why] I shouldn't allow it,” on one hand, or “You haven't made a case for why you need it,” on the other hand. So, if you have the burden of proof, you lose. I think parties are pretty aware that they have the burden to educate the court - particularly with technologies that are new and the court has not seen before.
Logikcull: There has been an explosion in data these past few years, particularly with more and more connected devices. You have the growth of the Internet of Things, for example -- your smart speakers, phones, your Fitbit tracking you all the time. There's been some worry that this data growth is going to add more and more burden to litigation in terms of discovery, ESI, production, review.
On the other hand, there’s an argument that technology is going to make it that much easier to get through all of the extraneous information that's gathered and focus on the hottest documents and data points. Where, if anywhere on this scale, do you see yourself? How do you think technology and data growth are going to operate in the discovery context?
Scheindlin: There's a different answer for the two words you used, data growth and technology. With respect to data growth, there are always more and more sources that collect more and more data. I've seen incredibly interesting statistics about the growth of data. It doubles every year. It quadruples every 5 years. It's just enormous. So, I think that makes the discovery process much more difficult.
We know there is a lot of controversy now about collecting data from personal devices of employees or even the company-enabled devices that employees get. It's really hard because people use their devices for business and for personal [reasons]. They're all intertwined and mixed up and you have to be able to separate them. I think the big topic is personal devices and company devices and, of course, social media that's all over those devices. There is more and more data, more privacy concerns, collections concerns, possession, custody, and control concerns. That’s one side of the question.
On the other hand, on the technology side, as technology improves, collection, search and review should become less expensive.
"Data growth is huge and it adds up to a lot more data, a lot more collection, search, and review, but technology helps us do it quicker and cheaper."
Initially, things are more expensive. The first television cost a lot of money. But once the years go by and it's mass-produced, and it's miniaturized, it becomes something everybody can afford. That's probably what's going on with the use of technology for collection, search, review and even production. All of the costs are actually going down and technology is not a threat and not a foe of being able to do decent search job.
There have been a number of opinions that say perfection is not required. One does the best under the circumstances, [considering] both costs and time-efficiencies. We may not find every last tidbit, but most of that stuff is irrelevant anyway.
So, the answers differ. Data growth is huge and it adds up to a lot more data, a lot more collection, search, and review, but technology helps us do it quicker and cheaper.