Velkommen til Lørn.tech. En læringsdugnad om teknologi og samfunn med Silvija Seres, Sunniva Rose og venner. 

Silvija Seres: Hello and welcome to today's edition of Lørn.Tech. Today we're speaking English because of a very nice guest we have. We will be talking about law tech, or regulatory technology. And our guest is Dan Katz, professor at Chicago's Illinois tech and a co-founder of a company called Lex predict. And one of the best thinkers on this digitalization of law. Welcome! 

Dan Katz: Thanks for having me. It's nice to be here.

Silvija: It's a brave of you to just jump in into a podcast with a bunch of unknown. So, Dan. This is actually an area that is more unknown to most people than some of the other topics that we talk about, which might be virtual reality or biotech. So, will you tell us a little bit about what you do? 

Dan: Okay. Sure. First of all, I'm a professor. I teach at a at a law school in the United States. That's the Law School of Illinois Institute of Technology. As part of that I'm teaching feature lawyers how to better leverage technology to solve problems for their clients, and how to invoke technology to make the legal system work better for everyone. I also have a company that I founded called LexPredict and that's focused specifically on using prediction models and methods in law. So, sometimes people have a dispute with someone, and they go to a lawyer and they say, you know, maybe I want to sue them. And the lawyer evaluates their claim and then says well, look, I don't think you're going to win on this or maybe you would win. And that's kind of the historic way to do it. The alternative is to try to use an algorithm to do that prediction. And that's now what our company focus on.

Silvija: So, you can help me explain my problem in a structured way, and then you can give me some sort of probability for winning a case, based on past cases? 

Dan: Yeah, I mean, we don't deliver it like we have a robot do that for you. It's more like we take data on a particular type of cases and then try to forecast what would happen in a new case for a for a client or customer.

Silvija: How does that work? 

Dan: Well, I mean it works in the way that a human does today. Based on their experience in these prior matters, they're actually using their mind to process that. And so they're looking for key factors about your case. And then they're using that to kind of render a judgment when they sit down with you and try to explain what's going to happen. What we're doing is trying to formalize that in a statistical model. And without getting too much into the math of it. The basic idea is that there are patterns in data that that we can surface using models and methods, rather than exclusively using a human or a professional. 

Silvija: I'm an algorithm is person. And it's a fascinating question whether law is algorithmic at all. Because I have the feeling that first of all the more senior a lawyer is, the more adamant they will be that technology doesn't bother them, because what they do is truly relational. And then on the other side, I also have a feeling that there are so many different human ways to interpret both past setup and the current situation, as there are human lawyers. Can you actually find statistically relevant patterns? 

Dan: You absolutely can. If it was so unpredictable there would be no basis on which for a human to render judgment. So, what we're trying to do is get at those very things that they're looking for. Certainly, there are there are matters in the universe for which there's it's no data no way to get at it, but I would say that's a challenge whether you use an algorithm or a person. So, one of the things that models teach us is sometimes when you shouldn't be too confident. And yes, of course, there are things that we can't reach for these approaches. But if you looked at today, there's almost historically, in law there's been very little of this done. So, the idea is that there's plenty of opportunity even if you can't do everything for these approaches. Another area we work in is in the area of contracts. So, you know, lots of people in their lives have touched a contract in one form or another. Maybe they least the piece of property. Maybe they rented a car. In one form or another they've touched a contract. Employment agreement. All these different sorts of instruments. So, they've touched these things. We're trying to build systems and their many people working in this area to better understand and process contracts. 

Silvija: So for a Layman, like myself. A legal contract is usually almost unreadable. It is legalese. And if somebody can help me identify the central parameters, and the central decision points in a normal language. Usually there isn't all that much in the contract, but the language surrounding those center points is very extensive. Is that something you can, or the technology can help people with it? 

Dan: It's not something our company does as much of that particular. But there are a narrow range of companies working on that exact type of problem. Because it is a common problem people faces. They get these agreements, and then they are really not sure what to make of it all. And you could see where if somebody could give you some guidance. That would be helpful. Now, you could pay a lawyer to do that. But a lot of times people would say “All that's going to be very expensive. They're going to make it take forever. I just want to get some basic information so I can make a decision”. And we see this a lot with people running a small business. The cost of lawyers is really too high to justify using them in the to an extensive amount. So, the question is am I to completely do it on my own, or am I going to pay for an expensive lawyer? There's Something in between and it's that in-between space that a lot of people in this kind of legal technology spear or kind of positioning themselves. 

Silvija: So what are your favorite two three examples internationally of exciting products? 

Dan: It is different products for kind of different market segments. So, we sometimes called the enterprise space which would be large business interacting with one another. Just to kind of break it out a little bit, and then the kind of stuff that's we call the middle market. And then there's sort of retail facing things. That is kind of for everybody. We were talking briefly earlier about one that has gotten a lot of coverage, which is called “do not pay” but it's an application that allows people to contest parking tickets. So, they get a ticket and they feel like they shouldn't have, and it's not just parking tickets. They try to move beyond that. Some of that was called is it using artificial intelligence, is it not. I don't think that's really to me the critical question. I think the critical question is does he bring some clarity for a regular person when they look at a process and they have no idea how to even contest it. So sometimes people just end up paying these fines or tickets cause it is too much work. Even if I'm actually innocent. I'm just going to pay because I don't really know how to go about it. So, law is full of complexity. And the problem is for most people who consume it. It's they don't know what to do with it. So, I think of legal technology is sitting in there. It's creating simplicity where there's a lot of complexity today. So that's one example, there's several others that we could point out too.

Silvija: You'll mentioned also Legal Zoom and Rocket Lawyer. Can you say a couple of words about these? 

Dan: Sure. These are two of the bigger companies in legal Tech globally. They own that brand, is a Us brand, but now they're kind of going all over the world. And the basic idea is that they help you do some very simple processes, like file a trademark application. Help navigate and getting a company incorporated and things like this. And some focus on just sort of regular problems folks have, but also small businesses. 

Silvija: Is there actually an argument to be made that this can also become to available, and too cheap? Are people going crazy and incorporating businesses because it's so easy now, or do you feel that it really works as it should?

Dan: I'm sure there's a criticism one can find. I mean, I can go to Twitter anytime and find a criticism. I think lawyers don't like it because they can point out these instances where, you know, a person who didn't know necessarily what they were doing. Use these tools and didn't really achieve a good result. But a lot of other people are saying. Well I don't have any real good way to do this otherwise that's economically rational. Because it's just way too expensive. And there I get kind of certain and clear price for what I'm getting. I think LegalZoom has done, multiple millions of business and corporations just as an example. So, they sort of got it down to a fairly streamlined process.

Silvija: I also think for certain groups of people that find it very difficult to navigate. For example, services for refugees, or where banks need to do a better customer service. I think many of these things really could benefit from more transparency and automating the easy parts and then letting people do the hard parts. What's your optimal kind of division of labor here between a good lawyer that knows how to use his tools, and a good set of tools?

Dan: What we do to now is we just sort of bundle the routine and the kind of complex together. And then we have lawyers doing things that are really not something you need to go to school for a bunch of years to learn how to do. And that's just makes the price too high. So, to the extent we can automate elements of what they're doing while still retaining the kind of high-end expert for the kind of parts that really require high end reasoning. Lots of things are not that complicated. If you knew a lot about them. The problem is you don't do it that much. You don't really know. So, a lawyer might say with some of these problems, “These are pretty routine things for us”. They aren't routine to the person who's never done it before. It feels like climbing Mount Everest in terms of complexity. 

Silvija: So I have two AI related questions for you. One is, there is also more and more tools that incorporate artificial intelligence and some level of machine learning. And there has been criticism that very often they expose and expand on the past biases. So certain ethical profile will get the worst conditions for insurance, or more likely to do this or that. Better or worse than the human. Because humans of course have our biases as well. I guess part of the question is, if we automate this process which really needs a human touch at all the critical points. How do we make sure that it stays fair? And what's the artificial intelligence of fairness? 

Dan: Well, I appreciated the way you structured the question. Because it's not typically how its presented. It's presented that “Hey we can't use algorithms because they're biasing”. And then you look back at the existing process today, and it's far from perfect. So, my own view is that we can only put them side by side and compare them. And in an ideal world we would use the algorithm to de bias human decision-making and try to essentially set the two against each other. That would be the ideal state. We do have this challenge that we don't want to bake in these sorts of biases that already exists on the one hand. It's definitely the case though. I don't think the public is really ready for a mature conversation about this topic. They complain about bias in the justice system, you suggest an algorithm and they lose their minds. So, should we just go back to this process that you've already identified doesn't work very well? People have oddly more faith in these human-centric processes and they probably should. 

Silvija: I think people are generally very scared of artificial intelligence in terms of who controls it. And we need to teach them how it is balanced. 

Dan: Yes. There are different types of methods out there. One of things is just even doing these methods. It starts to expose some truths that we are necessary so comfortable with about how things work today. That's one, and then two, you know, on the spectrum of methods in artificial intelligence, some are true black boxes, and some are not. So, I think the public's concern which has been manifested as a political concern is that, we have a black box that you don't really understand how it works, and it's rendering a judgment and it's not particularly explainable in a deeper sense. And that terrifies people now, we don't know what's going on in your cognitive apparatus in your brain when a person renders a decision. We have notions, we have theories about what people are doing. But we don't really understand. So, I think again in that whole conversation is actually exposing what's true today, which is we humans are the ultimate Black Box. 

Silvija: Yes, and our Consciousness. 

Dan: That's right. So. I think it remains to be seen where all of that conversation goes across different societies.

Silvija: I think you touched on some really central starting point for actually the whole series of this podcast we make. We have to start from a position of some knowledge. Because otherwise, it's just fear. 

Dan: Yes, that's right. And I mean to me, I think the proper thing is to start collecting data and measuring what goes on in these processes. Most lawyers’ performances are actually evaluated rigorously. Something bad happens, they say “you know things happen, it´s unpredictable”. In medicine for example. I think is much more rigorous, than measuring things that lawyers are about measuring in their quality and performance. 

Silvija: I want to ask you a question. We are a little over time, but I have to get your opinion on this. Fundamentally the law system changes very slowly. It takes a big case, and then it takes a long deliberation, and then it takes maybe several layers of this legal hierarchy before you actually get to change. Especially if it's a constitutional thing. The world changes super-fast. How do we deal with these kind of to speed set up? Just a concrete example: We've seen a couple of examples with the autonomous cars actually killing a person. It's the fraction, of the fraction, of a fraction, of the people who killed people in the car. Yet everybody is now worried because we don't have a regulation ready. And the same thing will happen when the robot judges do something wrong, or when the robot surgeons do something wrong. And with GDPR I think we've gone completely crazy about what we protect, and we don't know how to protect the models that big data companies have of us. So how do we speed up the regulation for this new world we all live in?

Dan: To me it's one of the it's probably the biggest challenge we have in politics. If you just accept the basic proposition that the rate of change is increasing, then these models we have for delivering legal rules or regulation really don't keep up with the pace of the world. And I think there's this tendency to fight the last war. And our regulators are sort of backward-looking. Meanwhile the world keeps moving forward and arguably at a faster pace than ever. And I don't have any fantastic answers, but I do think it's a fundamental issue. If you think about how should we design government’s, it's going to have to be some method to adapt more rapidly a changing world, but also to make sure that our fears and causing us to crush something that could also save us. I think about lots of these technologies are both undoing in our Salvation all at once. I wonder if people understood what the internet would bring if they would have allowed it to even come to pass. I'm thankful that enough that politicians weren't savvy enough to really understood what it was. Otherwise we'd never have it.

Silvija: But I think it's a beast that changes. It was made for one thing, and then we use it for something else. But that's with every technology. 

Dan: That's right. There are a range of challenges. I'm fairly optimistic about our ability to deal with the challenges. But for my students who are going to be lawyers, and some of them will go on to be regulators and judges and things like this. I would first of all like them to have a deeper appreciation for these technologies. Because then they can actually make reasoned judgments about how they balance interests, by both allowing things to happen. But also, trying to prevent the truly horrifying risks. 

Silvija: So you are based in the US. And now you're in tiny Norway. Are there any advantages or kind of disadvantages for this kind of technology for a much smaller country? 

Dan: Laws are a very language center field. So, I think one of the one of the biggest challenges, it's not how large or small Norway is. I think it's how common a language is. So, to the extent of technology is going to be built around language. This language is not spoken in a lot of places and that’s challenging. At an infrastructure level across lots of problems, not just legal Tech. But any technology that needs to process and understand language is part of whatever it's doing. That's a thing the government should try to subsidize, because a bunch of tech isn't going to exist in this market unless somebody's going to bear that one time cost of getting these natural language tool kits to work here. And I if you look at some of the Spanish legal tech companies, they're going to nail the model for Spain and go to all Latin America. They have a huge scaling opportunity that isn't present here. 

Silvija: Cool. what's your one sentence summary or main point about legal Tech? 

Dan: I think I think people have legal needs, and they need a more cost-effective way to solve their legal needs. And technology can help facilitate that in my view. But it's up to entrepreneurial lawyer types and technology types to bring it to life. 

Silvija: Thank you so much for having this conversation with us, and for teaching us about law-tech. 

Dan: Okay. Thank you. It's great to be here. 

Silvija: And thank you for listening. 

Du har lyttet til en podcast fra Lørn tech, en læringsdugnad om teknologi og samfunn. Følg oss i sosiale medier og på våre nettsider lørn.tech.