Comment by etherael | Hacker News Reader

Comment by etherael | Hacker News Reader

Parent Full thread

What's an interesting segue to this is that fundamentally the "language understanding" issues are mostly red herrings. The gold standard within the legal profession we are constantly led to believe is to pursue a standard of objective fairness and justice which would be better defined by deterministic, logical languages like programming languages rather than fuzzy, broad languages like English.

It is far, far easier to have laws written in code and execute test cases based on those laws before even investing any capital into a venture which may require edge cases of a given law, rather than the current scenario of hemming and hawing around a fundamentally subjective and fuzzy area and eventually subjecting your endeavour to the rigors of being a legal "test case" consisting of much ado about nothing and the outcome is as close to random as anything else.

When you reduce the legal profession to coding and writing test cases though, hell will freeze over the next day. Such a politically powerful faction will never allow themselves to be marginalised in this way, even if they claim to desperately be striving for exactly the kind of scenarios that this would allow; universal access to a completely objective justice system with etc etc etc. The extent to which this is actually impossible is somewhat depressing because it's something of an expose on the hypocrisy implicit in the claims of the legal system to actually serve the public, rather than itself.

But hey, that's life. If it ever does happen my wager is it will be because it is forced on them through market forces, the public at large simply uses alternatives venues for mediation and the legal system is forced to compete with it. I predict much artificial legal protection to prevent this from happening before it actually comes to pass, though.

Although there are some laws that can kind of be represented as computer code (for example, TurboTax captures a good portion of the tax code, and programming a 55 mph speed limit into a car shouldn't be hard) for the most part the law acts as a system for delegating to judges limited authority to exercise discretion. Ultimately, the decisions are made through that discretion, and they aren't entirely predictable. If you haven't already, take 30 minutes and read Oliver Wendell Holmes' "The Path of the Law" and see if you still think the only obstacle to mechanizing the law is the political power of the legal profession. http://www.gutenberg.org/ebooks/2373

I'll give it a look in, but yes, my general opinion is that discretion leads to subjectivity and unfairness. Unfortunately the attempts to reduce discretion have historically resulted in malicious arbitrariness, sentencing teenagers to sex offender status for sexting, silly three strikes laws ad et al. At the same time it seems to me that a more richly detailed granular objective system with oversight into the actual code to any interested parties with a clear and concise method for changing that code would be a far superior approach than either of the previously mentioned extremes.

You might want to look into the civil code system of law. It might appeal to you more than the English-derived system that I suspect you live under. Check out the Quebec Civil Code: http://www2.publicationsduquebec.gouv.qc.ca/dynamicSearch/te... (ignore the bizarre English translations of some concepts...).

You are correct, I currently live under a common law derived system, and my research into actually translating a legal system to code has led me to believe that the best place to start would be a civil law based system, though I had been mostly looking into scandinavian countries prior to this link. Thanks for that, just to be clear though, is civil code specific to Quebec or is it all of Canada on a civil law derived system? If it's Quebec only I'm guessing my grasp of English isn't much going to help me but if it's all of Canada I'm thinking this could be a much better starting point than the Scandinavian systems as I don't speak any of those languages fluently.

The Civil Code is just in Quebec. That's because that region of the country was originally owned by France and centuries later it's still French-speaking and uses a France-derived legal system for all but criminal matters.

There are many English speakers in Quebec (one of Canada's top law schools is located in the largest city in Quebec, which is known for being English-speaking...) and so you should be able to find a lot more materials in English than Scandinavian countries but I wouldn't know.

You're likely already aware of it but Louisiana also has the historical France link and civil code connection but I know much less about how that system works.

> attempts to reduce discretion have historically resulted in malicious arbitrariness

You're 90% of the way to the answer. Take into account Hanlon's Razor and it should become clear.

Unless the people designing the law are smart enough to take into account every single edge case, we're faced with a choice between gross miscarriages of justice through inflexible laws, and gross miscarriages of justice through incompetent or corrupt judges.

We know the people designing the law aren't that smart; and even if they were, the laws are already too complicated. The only solution is to strike a balance between the two types of errors that minimizes their sum.

Also, you're focusing on statutory law. There is a whole, great, big, huge area of law that is contracts made between two arms-length parties.

Very little of what I do as a lawyer involves interpreting or applying specific statutory laws.

How much of this is a characteristic of law, in general, and how much of this is a characteristic of the English common-law tradition?

In a civil-law system where laws and regulations are more minutely detailed, and judges have less discretion, would law-as-computer code be more feasible?

Perhaps _more_ feasible, but not feasible enough. In German law schools they make it sound as if there's only one legal interpretation for every case, but actually there are often competing ones.

Also, even if the legal situation is clear, there's still discretion involved to decide which claim ranks higher. E.g. Free Speech versus (something like) Right to Privacy in the case you can see at http://lexetius.com/1999,2324 . (If you don't speak German, Google translate might be good enough.)

To me the practice of common law and civil law seem more similar than their theory.

When I was a law student, I think as you do: the law could be better replaced with code.

Now that I am a software engineer, I don't think that any more.

Though law and code have a lot of similarities: http://skepticlawyer.com.au/2011/06/27/the-art-of-law/

What changed your mind?

Complexity.

The law isn't complex for the sake of it. The law is complex because the world is complex (cf. essential and accidental complexity). A single person is complex and hard to model reliably -- sure, you can get the first 90% right, but that's not enough. Then you scale that up to pairs of people, people in families, people in businesses and so on.

Law tries hard, very hard, to boil this intractably vast variety down into general, abstract systems. But corner cases frequently arise that don't quite fit the current rules. The rules need to extended by analogy, or broken, or turned inside out, or ignored, or not ... in order to make it work. Every once in a while enough of these little oddities accumulate that a new generality can be perceived.

But, like any model, the first 90% is easy. Everything else is recursively more difficult as you go into sub-sub-sub-sub categories where millions of dollars or people's lives and freedom can turn on a tiny distinction that only humans could, currently, tease out.

I did not come to think of law this way while studying law. It was when I began writing software that I realised how complex the world really is. How difficult it is to change established, complex systems.

The computer is a useful brute, but not too bright. Explaining all those corner cases, all those once-in-a-turquoise-moon exceptions in nauseating detail, is a useful lesson in humility.

But to sacrifice the gains in the first 90% for the sake of the remaining 10% just so that the remaining 10% is handled in a way that is maybe possibly better than what we might have if we went the other route? That sounds like a questionable benefit for a huge cost.

You do have a point though, perhaps a hybrid approach is best.

The people in the 10% feel otherwise.

One of the classic moments in the history of common law was the emergence of Equity. It arose because the strict, predictable rules of Common Law as it then existed were leading to what were clearly unjust outcomes.

Sometimes no specific rule can capture the correct action. You need room for judgement. It bugs the everliving daylights out of me that so much of law is undecidable and informal, but it is so out of historical demonstrations that a balance between robustness and strictness is required.

And the people in the 90% that don't want to pay lawyers 500$ an hour? That's what I mean, as far as I can see I don't follow as to why that does not necessarily dwarf the concerns of the remaining 10%, and also it's not even a granted that this would actually happen. The emergence of equity in common law is one thing, but common law is administered by humans, a system written in code to which anyone was free to contribute and which was administered by computers would be far less susceptible by nature to human error.

The only scope for human error in such a system is the process of coding itself, and massively distributing this and opening that process as much as possible makes any bugs in the system shallow by comparison, as well as easily demonstrable by test cases rather than actually requiring some clueless politician, lawyer or judge to make a terrible decision and someone else to suffer the consequences for it.

The act of writing programs, and discovering a bottomless well of bugs in those programs, is what has made me respect lawyers all the more.

I do think some tools and practices can be adapted. I think, for example, that expert systems -- whether classical or hybrid a la Watson -- could make a big dent in that $500 figure. That is definitely pursuable.

But humans inject errors, it's what they do best. And outside of a properly formal system (and per Godel, these aren't perfectly usable anyway) the best way to identify errors is ... with another human.

You might say that judges have been performing software inspections for much longer than we have.

Don't get me wrong, law is as slow as frozen honey. But the problem domain is heinously complex and I expect that any software system would reflect that complexity. Given our profession's iffy track record on creating items of massive size and complexity, I lack your transformative enthusiasm.

I am what you might call a cautiously optimistic conservative. Or a pessimistic radical. Not sure. One blogger I host talks about "radical centrism".

>a standard of objective fairness and justice which would be better defined by deterministic, logical languages like programming languages rather than fuzzy, broad languages like English.

Can you formally define concepts and events occurring in a "fuzzy, broad" world in a useful way? It is kind of amusing to imagine that someone out there is sitting around with a recursively-enumerable objective statement of what constitutes fairness.

To the extent that applications defining the class to which you believe your behaviour exists and allowing you to test it with a high degree of specificity, yes. There have been explicit ventures along this exact path previously in a more limited extent, that being financial products having python based code that would allow the investors to put in a set of assumptions and see what the product model would do under those circumstances.

Fairness itself is a fuzzy concept, sure, but writing clearly delineated code that deals with reality in a concrete fashion with the goal of being fair is the only way that fairness even enters the equation.

>To the extent that applications defining the class to which you believe your behaviour exists and allowing you to test it with a high degree of specificity, yes.

I'm operating on the assumption that you dropped the word "belongs" somewhere in there.

>There have been explicit ventures along this exact path previously in a more limited extent, that being financial products having python based code that would allow the investors to put in a set of assumptions and see what the product model would do under those circumstances.

In the same way that there have been explicit ventures along the path of the axiomatization of mathematics, yes.

>writing clearly delineated code that deals with reality in a concrete fashion

Show me some clearly delineated code that deals with the workings of a single sheet of A4 paper in a concrete fashion, and then we can consider the law. Fairness is not a fuzzy concept, it is a human concept, humans as much a part of the universe as any other. It is not really harder to measure fairness than to measure the electromagnetic interaction.

We spoon-feed our problems to computer systems so that they may analyze the parts which we cannot; your precept of a computerized legal system is assuming a priori the lack of spoon-feeding, since corruption can enter the picture here just as easily as in the old system, cf. LulzSec. Thusfar there are two pieces of secure code in existence: seL4 and qmail, and I'm not sure about the former.

Reply on news.ycombinator.com