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

Silvija Seres: Hello and welcome to this edition of Lørn.tech. Today's topic is law Tech or reg tech. My guest today is Malcolm Langford, professor at University of Oslo in law and social science. Yep. Welcome Malcolm! 

Malcom Langford: Thank you very much.

Silvija: Malcolm is from sunny Australia and he speaks wonderful Norwegian, but we thought we'd do this in English because then we can use it maybe internationally.

Malcolm: Sounds good.

Silvija: So Malcolm, what brings a sunny man like you to a country like this? Why are you in Norway doing law? 

Malcolm: Well, there's the public version and the private version to why I am here. The private version is perhaps more related to some other attractive features of Norway. The public version is that I did my PHD here in law. But I also have a background as an economist and I also work with Statistics. And it's the work with Statistics that has taken me into Fields like legal technology, text analysis, machine learning.

Silvija: You started off with a combination of statistics and law?

Malcolm: Yes. And I was Doing more traditional increasingly political science and economics sort of analysis with Statistics. Like regression analysis. 

Silvija: Why is that a cool thing to do in Norway for a foreigner? Where you doing law back home?

Malcolm: Thing about Norway and all European countries is that you don't have to do another degree before you do law. So pretty much all Norwegian lawyers for example only got a law degree. But in Australia and America, you have to do another degree before you become a lawyer. So, we have double competence, which allows us more easily to move between fields to do law and political science, law and chemistry, law and engineering for example. So that statistical background has taken me into you know, researching wider judges behave in certain ways. Does law have an effect in society? Can we statistically measure that for example? So, it's doing more likelihood analysis. You have statistical research on Norwegian judges for example. Are they guided by ideology in the same way as American judges? 

Silvija: Do you have any conclusions that you can share?

Malcolm: The research from Bergen shows actually that the American model doesn't really work in Norway. Ideology doesn't really seem to explain judicial behavior. Which government appointed you at the time? But possibly professional background does matter. So, if you worked in public administration, you may be more likely to side with the state in a case between an individual and state. So that's sort of traditional political science analysis of judicial Behavior. 

Silvija: There's always human bias. 

Malcolm: Yes. And then we have the new way of computational legal studies, which is trying to predict for example, how judges will behave and what the legal outcome might be. So right now we're running a workshop. We've just had a paper where authors have predicted the outcome of a European Court of Human Rights decision. Just from a two-paragraph statement of the facts.

Silvija: Could any workshop like that predict things around Trump and all the stuff happening with you know approvals of covenants and so on. I just I still can't decide if law is such a rational thing or not.

Malcolm: So, we had a paper today actually on the relates to Brett Kavanaugh. You could predict that he is a fairly polarized, perhaps ideologically driven judge by the type of language that he's used in the last 12 years in his judgments. 

Silvija: You have to read through these articles automatically? 

Malcolm: We are using a machine learning, or text analysis software to look at the language patterns of judges. And some judges use very much polarized language. Either on the right or the left. 

Silvija: Using superlatives and..?

Malcolm: It can be adjectives is often particular types of nouns. A particular type of language for example in freedom of expression cases or gun rights cases or civil rights cases. And from that you can also predict which government was likely to appointed them, and also to a large extent their ideological bias. So, we know from this type of research the types of judgments that Brett Kavanaugh is likely to come with in the future. If you look at Trump's first nominee Gorsuch. The statistical research was pretty clear that he was very to the right. Then there are some in the US and even some Norwegian journalist that said he was more moderate. But the statistics were pretty clear, and he's pretty much been pretty hard to the right. So, the statistics was pretty much right. There's always going to be variations. He had one judgment where he went with the Liberals, but everything else has been as predicted. 

Silvija: Listen Malcolm. There is a way that you and I met as judges at the legal hackathon by wonderful Merete Nygaard. And one of the first exchanges we had there was about huge Norwegian company, intersecting computer science and law. And you said something revolutionary happened fifty years ago, and then is happening now again. Could you please expand on that here? 

Malcolm: So Norway can be very proud that it has one of the true Pioneers in what we might call computational legal studies. And also, legal technology, given that he also was behind love data. So, almost 50 years ago, a young pionér was setting up the Norwegian Center for computers and law. He was reading science fiction, so he had a sense of what was coming compared to most. He began thinking about how in fact the logic and the Machinery of computers could help us with solving legal problems, storing Legal Information. Building artificial intelligence. He has an article in 1986 about how we might set up artificial intelligence systems to decide on social security cases around pensions in Norway. 

Silvija: Was this unique? 

Malcolm: The first articles on artificial intelligence in law come out around the early 1950s, but they're very few. So, he's in a clearer world class. The first major PHD really trying to use this method only comes in 1986. 

Silvija: Where is it line drawn between science fiction and these scientific articles? You could argue that much of Science Fiction was dealing with these kinds of things. 

Malcolm: When we want to think about what we can do in the future we need what is called sociological imagination. We have to consider how the future might be different and then how can we get there and science fiction has shown itself to be quite important in imagining new things, and then we try and build it. And sometimes it works, and sometimes it doesn't. So, in that despect lawyers need not only start to improve their digital literacy and start to learn computational skills, but they also need to read a bit more science fiction in order to improve their creativity and Imagination. 

Silvija: But as a computer scientist, and I remember Jon Bing with his hair and his sandals, and I'm still surprised how precisely and focused he went into the area of law. Because for a mathematician like myself the area of law seems like it's all about the narrative, and it's about how you deliver the narrative, and it's about the language you use. To me it seems like a very non-mathematical, non-computational area of our lives. But you are telling me I was wrong?

Malcolm: Well here is a great Paradox about law. On one hand law is exactly like computer science. We use if and then. A Computer Logic constantly in the way we talk about how we do law. So, if a car hits you by the side of the road and then you fall over what is the legal rights that you have in that case. And then, if this happened in this jurisdiction, at this time and so forth. So, in the 1980s seeing the computer like logic and law people like Jon Bing and others which as Susskind try to set up computer models to reflect that process. But they found two problems. So, one is it gets very complex. You have very long decision trees of “if” and “then”, even for a simple case like that, because there's so many variables out there. The second thing is this literary language like human-like part of law, which you can't control with a computer model. And that's why we have this new revolution in law and artificial intelligence. Because we've partly solved two of those problems. Firstly, with computational power and machine learning. We can predict what those decision trees might actually been doing. Sometimes they're in a black box, but now we can actually in an increasingly large number of cases actually predict that chain of legal reasoning. The second thing is we using text analysis. We're starting to understand textual patterns and identify when something is logical or perhaps when it's more emotional. Get when it actually is rationally based and when it's more ideology and so forth. We're still a long way to go. But that's why law and computer science are increasingly coming back together again. 

Silvija: So, when we were on this Hackaton I remember there was one of the examples that was claiming they do artificial intelligence. And then I remember we both looked at each other and you ask the question. Well, how do you do machine learning? And they said, oh, it's very advanced, we have a lot of statistics here and there. What was the answer that you were hoping to hear? What would be a good case of machine learning in illegal app? 

Malcolm: Well, the first thing I want to hear is: what is your corpus or data that you began with? How did you choose your training set? To identify a set of data. Then what were you trying to predict? So, take the law of European human rights. You are trying to predictive is a violation or no violation by the court. And then you run your predictions and to set up the parameters for your model. Coefficients according to whole lot of variables that you can actually see. And then how well did the model do. And then, maybe experiment or look at the data and maybe adjust the coefficients in certain ways to see if it predicts better over time and then test it on new data. For example put your dataset in two and test your training setresults on the later results. So that's the sort of basic machine learning process that I would have liked to hear about. Including which program did you use, did you use Python? 

Silvija: So there is a geeky side to Malcolm as well? 

Malcolm: Yes. 

Silvija: So what's your research about? you talked about these language analytics and the statistical predictions. Is that the core of it, or can you help us understand? 

Malcolm: Okay. I've been trying to understand for example things like judicial behavior. So, I've been doing more statistical analysis on that. I've been doing network analysis on how lawyers and arbitrators’ network and control certain markets like arbitration markets. I've been doing experiments on the effect of the Norwegian Court of public opinion, for example, but now moving into more real computational legal studies. We just had a workshop yesterday on developing a robot lawyer for the European Court of Human Rights. So how can we actually help people make complaints to the European Court of human rights in Strasbourg. 95% of the complaints was dismissed immediately.

Silvija: Because there is no not enough time to deal with them? 

Malcolm: No. Because they filled out wrongly, they are using the wrong arguments. They are not putting the facts in a right way. And many of these cases are actually filed by lawyers, but they don't know the European convention on human rights. So, what we're trying to do is combine a sort of expert design system. We have lawyers who know the system set up the forms in a way, that gives you a good chance of success. But then using the machine learning results that we're getting from different projects and different partners. To help prompt applicants put in arguments and information that we know will help them get through the system at the court.

Silvija: So what you say is that we can make really good robot lawyers, if we specialize them in a niche? and then and then we still have the task of applying the right one to the right task and connecting them as humans. 

Malcolm: Like every other area of society, we can use artificial intelligence for very specific tasks.

We still a long way from using it for multiple tasks, and that's why most legal tech projects have to be fairly specific. And often in areas where you've got lots of usually big data behind you, or it's some sort of routine task. And I think in the next 10-20 years will see a complexification. 

Silvija: I'd like your clarification on the point I've heard before, and I want you to say that you agree. I heard that Norway has uniquely good public data long series, clean data, cross connected with the among other things as personal unique number Etc. Do you see that as a big advantage in this space for Norwegian law at least?

Malcolm: So my prediction is basically, and we can already see that in the U.S. A lot of legal technology is coming out of the private sector. That's where the money is. There's a lot of data being collected whether it's eBay or Facebook, Google and some of the other legal Tech projects are over time. In Norway and Europe, we see much less legal technology startups in the private sector, but it's increasing. But what we see also and almost a greater amount of action is in the public sector. And we're already seeing in various areas in Norway automatization, migration law, taxation laws are increasing in Social Security 

Silvija: Actually bureaucrats are working really hard on automating parts of their jobs? 

Malcolm: Exactly. So, there's lots of data and there's lots of turnover to. It is lots of people involved, lots of applications and so forth. We see also customs officials trying to use AI to flag which people should be stopped at the border when they're driving in. So, I think the public sector will be that the key innovator in Norway. And that's both good and bad. The good side is it's going to have financing behind it. There may be a public interest involved, but there's a question of, where is legal tech for everyday people? And do we have enough resources also going for that. So, I would argue the Norwegian state has a duty to start financing legal tech. That also is for every day. And it's also.. you've just been denied for example, a residence permit or denied pension or disability benefit. And you're now facing a robo decision-maker often in the department in Norway. So, you also need support on the other side. You either need a lawyer who understands these new automated processes, or you need some sort of legal Tech or Robo lawyer on your side. So, that's where we are in Norway. And on the other end of this spectrum is China where the state is driving a lot of AI, and you can see what happens if it's very centralized. 

Silvija: Yeah, so Malcome. We need to round off. If you were to summarize in the sentence or two, what is the perfect lawyer in the future? What are you training your students to be? How do they use the tools? What's the what are the most important skills?

Malcolm: So 500 years ago, if you're a lawyer you had to know multiple jurisdictions. Because we had a complicated sort of, medieval world. You had to deal with Roman law, French law, Norwegian law and so forth. Today we're partly back again with that with globalization. You have to deal with multiple jurisdictions. International regional and national law. But when we now move to the to the future, I think the task of the lawyer is to deal with these new forms of jurisdiction. It's no longer Roman law or French law, but it's this transnational law. It's these new technological spaces. Every day on our on our mobile phones were agreeing to contracts which were written in Silicon Valley for example. And part of it may be under South Korean or Chinese law and so forth. So, you have to deal with a whole range of new jurisdictions. Virtual jurisdictions, you have to deal with process, which is a highly technological.

So, I think every lawyer in the future needs to have basic digital literacy. They have to understand basic statistics, basic machine learning processes, how to understand sort of highly technological legal processes are driven, and then we need 20% of lawyers who are also computer scientists. And who actually have double competence in both. Because what I worry about is that computer scientist will design the legal technologies of the future. I think we need a partnership with computer scientist. And therefore, we also need lawyers who are trained in that who understand the legal side of things. 

Silvija: Amen. Malcom. Thank you so much for your time. We learned a lot about both Jon Bing, which I think we need to celebrate much more in this country than we do. And then on the future of law. 

Malcolm: Thank you very much. 

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.