CatalaLang/catala: Programming language for law specification

This project seems to implicitly assume that a formally specified code of laws, where statutes can be interpreted largely mechanistically, is a good thing (and by extension, that the existing system of human interpreters with discretion and margins of error is a problem to be overcome).<p>I don&#x27;t disagree with this assumption outright, but it&#x27;s certainly not obvious to me that it is correct, and the authors appear to present no arguments supporting the same.

For years I&#x27;ve tried to convince my lawyer friend that something exactly like this would be great to have, and then it turns out to have existed probably all the time.<p>I think this is truly awesome.<p>Every law should be written in a language like this, and presented publicly with syntax highlighting and consistent formatting rules.<p>Then it should be made part of the school curriculum to learn the law language.<p>I believe it would greatly improve everyone&#x27;s ability to read laws and be confident about their understanding of them, which would be a huge boon for society.

It would be interesting to also &quot;weave&quot; in test cases. The workface of logic statements is exactly where bugs are introduced.<p>Especially around temporal events, and that goes to formal models (and even more bugs).<p>Typically, if there is a rule around height, there would be at least three tests¹: one taller, one equal to, and one shorter. (Without types or something, then also negative, null, and max&#x2F;min boundary inputs too.)<p>So you could have tests based on timelines, like<p><pre><code> Given a regulation is passed in 3 months And parties are prevented from exercising B But &quot;17 tons&quot; of waste are dumped anyway And ... When ... Then ... </code></pre> Having a model checker integrated would be a boon. Maybe we could have DevOps-like pipelines in formally-verified legislature (or at least the encoding of language to code).<p>¹ <a href="https:&#x2F;&#x2F;en.m.wikipedia.org&#x2F;wiki&#x2F;Equivalence_partitioning" rel="nofollow noreferrer">https:&#x2F;&#x2F;en.m.wikipedia.org&#x2F;wiki&#x2F;Equivalence_partitioning</a>

I think what I&#x27;d rather see is a standardize test suite format for laws that spells out the intentions.<p>Once I lived in a state that proposed a very simple anti-child porn law with good intent, but it was <i>too</i> simple. It read sort of like &quot;anyone sending explicit pictures of minors from a cell phone will be guilty of conveying child porn&quot;. It was written in the proper legal jargon, but wasn&#x27;t a whole lot more detailed than that. I called the sponsor of the bill and asked if that meant if my hypothetical daughter sent a naked picture of herself to her boyfriend, then wouldn&#x27;t she be a felon under his new law? He had an &quot;oh, crap, that&#x27;s not what I meant!&quot; reaction and ended up withdrawing the bill so it could be re-written. (Aside: I felt pretty good about that. Props to the legislator for being quick to understand and respond appropriately!)<p>Imagine if that were handled like program code, with a test like:<p>* This law does not apply to minors sending pictures of themselves.<p>That would do a few big things:<p>It would make legislators be clear about what they mean. &quot;Oh, we&#x27;d never use this online child safety law to ban pro-trans content from the Internet!&quot; &quot;Great! Let&#x27;s add that as a test case then.&quot; I confess that this is a deal breaker: politicians don&#x27;t like being pinned down like that.<p>It would probably make it easier to write laws that reflect those intentions. &quot;Hey, that law as written would apply to a 15 year old sexting her boyfriend! The code doesn&#x27;t pass the tests.&quot;<p>Future courts could use that to evaluate a law&#x27;s intent. &quot;The wording <i>says</i> it applies to 15 year olds sending selfies, but the tests are explicit that it wasn&#x27;t meant to. Not guilty.&quot;<p>I&#x27;m sure this couldn&#x27;t happen for a hundred reasons, but I can dream.

Something like this got passing mention in Greg Bear&#x27;s [1] book <i>Moving Mars</i> (1993) [2], under the name <i>Legal Logic</i>. The (human) Martians used it with AI assistance to formulate legislation for their newly independent society.<p>For those who don&#x27;t know, Greg Bear was a well-known SF author who died less than a year ago. His passing was discussed here at the time [3] [4].<p>He was one of the authors that influenced my youth a great deal, and I particularly remember this aspect of Moving Mars as catching my imagination, so will be interested to read what Catala has to offer.<p>[1] <a href="https:&#x2F;&#x2F;en.wikipedia.org&#x2F;wiki&#x2F;Greg_Bear" rel="nofollow noreferrer">https:&#x2F;&#x2F;en.wikipedia.org&#x2F;wiki&#x2F;Greg_Bear</a><p>[2] <a href="https:&#x2F;&#x2F;en.wikipedia.org&#x2F;wiki&#x2F;Moving_Mars" rel="nofollow noreferrer">https:&#x2F;&#x2F;en.wikipedia.org&#x2F;wiki&#x2F;Moving_Mars</a><p>[3] <a href="https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=33679668">https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=33679668</a><p>[4] <a href="https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=33675708">https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=33675708</a>

(self plug) They published a paper describing the language <a href="https:&#x2F;&#x2F;hal.inria.fr&#x2F;hal-03159939" rel="nofollow noreferrer">https:&#x2F;&#x2F;hal.inria.fr&#x2F;hal-03159939</a>, and here&#x27;s a short video summary of it: <a href="https:&#x2F;&#x2F;youtu.be&#x2F;OiaFTFSAa1I" rel="nofollow noreferrer">https:&#x2F;&#x2F;youtu.be&#x2F;OiaFTFSAa1I</a>

programmers love to propose using &quot;programming language&quot; or similar for law<p>But this fails to realize that _ambiguity (in some ways) is a fundamental important part of law_.<p>This is because the world itself is fundamental ambiguous (in some ways)&#x2F;clear cut.<p>Naturally not all ways of ambiguity are wanted.<p>But you can be sure that with &quot;code as law&quot; the ways loopholes are abused will get worse in my opinion.<p>I would even go as far that some many laws should be more focused on what should be upheld then the details how (which is fundamental less clear cut&#x2F;more ambiguous).

I don&#x27;t think there&#x27;s much of a problem with actually reasoning about a law&#x27;s text that a computer can help solve. The complicated bit is weighing equities, which still requires humans and lawyers.

As someone who speaks Catalan, I find the name collision with the language referred to in the article most unfortunate.

The naming choice is really unfortunate. It&#x27;s like naming a programming language &quot;français&quot;, or &quot;Deutsch&quot; (or &quot;English&quot;).<p>From the bottom of the readme:<p>&gt; The language is named after Pierre Catala<p>I&#x27;d suggest changing it to PierreLang then.

Laws would do well to follow the rules of software. Small modules with clear responsibilities with an emphasis on readability and test cases that are run <i>before</i> you go to prod, for example. Testing is expensive so I understand why the legal system would rather just push their code and fix bugs when they see them in the wild. The collateral damage for people caught up in real life test cases is tolerable, especially when it’s someone else footing the bill.<p>Linting and type checking the existing codebase would also be more helpful than rewriting everything in a new language. Enforcing size constraints on vocabulary and word count. Cross referencing between different legal systems. Throwing out dead laws that are no longer executed in prod. Profiling the efficiency of existing laws to find hot spots.<p>There’s little incentive to do this when the current system is run by a cadre of highly trained legacy COBOL programmers. I’d pick a very small part of the system — incorporate a new city and start from the ground up — and take it from there with the clear eyed expectation that a full rewrite is going to take a century.

Also see Logical English, a &quot;Programming Language for Law and Ethics&quot;<p><a href="https:&#x2F;&#x2F;virtuale.unibo.it&#x2F;pluginfile.php&#x2F;1273247&#x2F;mod_unibores&#x2F;content&#x2F;0&#x2F;Logical%20English%20for%20Law%20and%20Ethics_new2.pdf" rel="nofollow noreferrer">https:&#x2F;&#x2F;virtuale.unibo.it&#x2F;pluginfile.php&#x2F;1273247&#x2F;mod_unibore...</a>

I was a bit confused by the fact that my first language is Catalan, which in Catalan is spelled Català. So yeah, imagine someone proposing a language specification for the law called English.

Ambiguity is a feature, not a bug. I used to spend a lott of time on business process automation, and even in those more structured and restricted settings trying to codify procedures most often fails. The reason is that reality has (a) so many edge cases that it very rapidly devolves into chasing down an ever diminishing ROI, (b) is unknown by the middle management and business analists, those that would have the authority to construct and sign off on it, and (c) relies on intelligent people applying creative pragmatic solutions to keep the business running and straightjacketing those into inflexible automatons is the most surefire way to sink the ship.

Related:<p><i>Catala: a programming language for socio-fiscal legislative literate programming</i> - <a href="https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=24948342">https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=24948342</a> - Oct 2020 (37 comments)

I&#x27;d be interested in seeing something like this for verifying game designs &#x2F; new game rules given an existing design

I think having a &quot;linter&quot; for laws can be beneficial. It can help producing laws that are easier to read and understand.<p>Having a &quot;compiler&quot; for laws can help identifying conflicts between different codes of law. e.g.: Imagine having a compiler error when a law is unconstitutional from a logical standpoint.<p>But verifying the &quot;business logic&quot; (e.g.: what is the spirit or intent of the law?) of the law will remain a human intelligence task.

I’m really interested in this project as another way to computationally express and use law. For more color on the project, here is a demo and discussion “Idea Flow” session I did with the project leads: <a href="https:&#x2F;&#x2F;law.mit.edu&#x2F;pub&#x2F;ideaflow8" rel="nofollow noreferrer">https:&#x2F;&#x2F;law.mit.edu&#x2F;pub&#x2F;ideaflow8</a>

My hunch is, in any sufficiently large rule set, there will be inconsistencies. Handwaily think Gödel, or just the need for bounded domains in DDD.<p>Humans (or, well, AI) is needed to cope with inconsistencies.<p>That said, pointing out the fact of existence of inconsistencies could be very valuable. But a system needs to embrace them, not fight them.

This paper definitely has thoughts on the matter <a href="https:&#x2F;&#x2F;law.mit.edu&#x2F;pub&#x2F;interpretingtherulesofcode&#x2F;release&#x2F;4" rel="nofollow noreferrer">https:&#x2F;&#x2F;law.mit.edu&#x2F;pub&#x2F;interpretingtherulesofcode&#x2F;release&#x2F;4</a><p>&quot;As the Rules as Code movement gains momentum, questions are starting to be asked about the performance and practical effects of expressing law computationally. This article examines the strengths, weaknesses, and new opportunities of engaging with these emerging systems.&quot;<p>The Future of Coding podcast covered it recently <a href="https:&#x2F;&#x2F;futureofcoding.org&#x2F;episodes&#x2F;065" rel="nofollow noreferrer">https:&#x2F;&#x2F;futureofcoding.org&#x2F;episodes&#x2F;065</a><p>The abstract says, &quot;Software code is built on rules. The way it enforces them is analogous in certain ways to the philosophical notion of legalism, under which citizens are expected to follow legal rules without thinking too hard about their meaning or consequences. By analogy, the opacity, immutability, immediacy, pervasiveness, private production, and ‘ruleishness’ of code amplify its ‘legalistic’ nature far beyond what could ever be imposed in the legal domain, however, raising significant questions about its legitimacy as a regulator.&quot;<p>It&#x27;s a complex paper&#x2F;topic that I personally need more time to grasp before throwing my opinions around too heavily. But my first, knee-jerk reaction so far is that moving laws into code is a bad idea. Specifically, as the paper says, &quot;...code by its very nature tends toward a kind of strong legalism. This is the case regardless of the intent of the programmer, however vicious or virtuous that may be.&quot;<p>The &quot;strong legalism&quot; inherent in code means &quot;the sovereign’s exercise of power is de facto legitimate, and thus not open to question.&quot; Not to be reductive, but that ain&#x27;t good.<p>I feel we&#x27;ve seen evidence of this path already, with (easily refuted, but somewhat common) claims like &quot;data can&#x27;t be biased&quot; (for example). The tendency to blindly follow a computer&#x27;s dictate with, &quot;Well, the computer says this is so, so it must be so.&quot; is strong in our society at times, I think.

Interesting that their “money” type doesn’t allow fractional cents or track the currency. And the only collection type is a fixed-sized list. I can imagine a lot of other useful base types that would be useful for laws about the real world, such as physical units like meters.<p><a href="https:&#x2F;&#x2F;catala-lang.org&#x2F;en&#x2F;examples&#x2F;tutorial#The%20Catala%20language%20tutorial-Catala%20values" rel="nofollow noreferrer">https:&#x2F;&#x2F;catala-lang.org&#x2F;en&#x2F;examples&#x2F;tutorial#The%20Catala%20...</a>

The Catalan language name is written Català.

Law is a mess, in part because its authors take shortcuts. For example, from the first example on CatalaLang&#x27;s README.md:<p>&gt; If the property was acquired by gift [and various conditions apply], then <i>for the purpose of determining loss</i> the basis shall be such fair market value. [emphasis added]<p>I <i>think</i> (and I&#x27;m not a lawyer or a tax expert) that this means that the basis of an asset can have a different value for the purpose of determining gain or determining loss. Wow, basis isn&#x27;t just a number, although one might not notice this if one didn&#x27;t read the six emphasized words.<p>But the Catala code seems to completely ignore this. Oops. I filed an issue:<p><a href="https:&#x2F;&#x2F;github.com&#x2F;CatalaLang&#x2F;catala&#x2F;issues&#x2F;514">https:&#x2F;&#x2F;github.com&#x2F;CatalaLang&#x2F;catala&#x2F;issues&#x2F;514</a><p>In a real use case, I imagine that substantial refactoring of the parts that <i>consume</i> basis might be needed when one notices that the basis is not a number.

Naming is hard Catala is cardsharper (cheat) in russian.

If anything, I&#x27;m grateful to this article for exposing me to this Bill Gates deposition. I had no idea he was so repugnant and incompetent liar. Sometimes, in my mind, I&#x27;m trying to justify the clearly bad actions by big corporations by saying something like &quot;maybe they don&#x27;t know&quot; or &quot;maybe they are in a tight spot between many bad choices&quot;, but here it&#x27;s loud and clear that it was deliberate. He and his buddies knew what they were doing, they knew it was wrong and they didn&#x27;t care. Neither had he the guts to own up to his wrongdoing.

That&#x27;s really cool!<p>I was puttering around with the idea of a ricardian compiler for legalese, basically a <i>decompiler</i> for something like this that could compile a legal text into clear logical rules. This would aid in proof checking for legal documents to ensure that they&#x27;re compatible with existing law, that there are no (unintended lol) loopholes and the like. It would <i>also</i> be useful if you wanted to create self enforcing legal documents that can be enforced deterministically by machines, such as collateralized agreements, and finally, even though someone would still need to know legalese, it could make the development of such agreements easier for people and lower the bar tremendously.<p>I wonder if anyone has built anything like that, if these guys have, or if anyone has built other interesting ricardian compilers.

I was expecting Lojban, or something similar. I remember someone proposing Lojban as a tool that could improve the clarity of legal and contractual writing.<p><a href="https:&#x2F;&#x2F;en.wikipedia.org&#x2F;wiki&#x2F;Lojban" rel="nofollow noreferrer">https:&#x2F;&#x2F;en.wikipedia.org&#x2F;wiki&#x2F;Lojban</a>

Would be so funny if all the compute moves from AI to finding loopholes in the programmatically defined law.

Although I agree on principle, the closest thing we have to &#x27;formal law&#x27; are smart contracts, and already billions of dollars have been stolen from bugs in these, despite barely anyone using them. I have some reservations for basing our entire legal system on code.

I would have a much easier time understanding government legislation if everything was provided in such language. I tried to compute my taxes by hand a month ago to see whether and how much money it would save if I enabled &quot;loon middeling&quot; (some Dutch law about income). But I couldn&#x27;t figure it out. The explanations provided were ambgiuous in some subtle way, leading me to incorrect assumptions. In the end I did figure it out by reverse engineering a free third party calculation tool (which also was not correct, but putting their insight and my insight together made something that came close to the number on my belastingaanslag).

This seems like a worthy project but I think its landing page should define the scope of what it aims to do in more concrete terms. Otherwise it runs the risk of being seen as overambitious and open ended without a concrete problem to solve.<p>E.g. its not clear if there is an explicit or implicit ontology against which the validity of any codification can be checked.

This is interesting, and not to criticize, but I wonder if transformer model&#x27;s accuracy in interpreting law will obviate the need for something like this.<p>It would be interesting to train Large Language Transformer Models to generate this code for you based on the text in the laws. This way you have a deterministic testable output, without risk of hallucinations.

What are the authors&#x27; goals, what is their intended purpose? I can&#x27;t find a mission statement on their website.

Something to keep in mind is that the courts are not necessarily trying to determine the truth, but rather create a place to allow two parties that represent different interests duke it out. Not always what the courts are used for, but it&#x27;s a different mentality than science or programming.

This project is doing code -&gt; text, right?<p>But then, the first line of description in Github says:<p>&gt; Catala is a domain-specific language for deriving faithful-by-construction algorithms from legislative texts.<p>This reads like &#x27;text -&gt; code&#x27;, which is the opposite of what this project seems to be doing.

I&#x27;m actually in court later this month pertaining to some stolen traffic cones. Dismissed my lawyer just yesterday due to a dishonest swagger. I&#x27;ll be preparing some documents using this after I prototype using a few frameworks. Will keep you posted.

This is really cool! But I guess it’s biggest drawback is being unable to deal with case law?

Super interesting. I also think we would win with a kind of versioning system for laws, including a definite objective for a law from the time of it&#x27;s creation, and constraints under which it should be questioned again

this is peak hacker news &#x27;solve human conflict by using technology&#x27;

Started something like this years ago with a company that ended up pivoting to a slightly different direction after a while. glad to see something in open source space.

No way would this ever be a good idea. Language changes too much, be it human or machine. This is why you need flexibility, not a rigid structure, in making law.

without formal training, one key thing I picked up is that most public understanding of legal concepts diverges from court understanding because law follows <i>logical</i> and&#x2F;or gates<p>so “and” isnt a list of accepted criteria, it is a list of things that must be simultaneously satisfied<p>but its only using logical gates <i>most</i> of the time<p>this is a good step in showing that. not a panacea but a good step!

I had so much trouble getting catala to even compile that I got frustrated. It’s a piece of academic code that’s very much abandoned

It’s good that they made it not turing complete but they should probably also force an upper bound on law complexity

A software for people. For dictating human behavior. Is that what we&#x27;re looking at here?

How is this different from Prolog?

Doesn’t TurboTax have some kind of DSL they use to encode all of the tax rules?

&quot;We&quot; will interpret the law for you and the judges, and &quot;we&quot; are not suspicious at all of having a hidden agenda to replace &quot;the law&quot; by &quot;how we see the law&quot; to benefit ourselves.<p>Is this a joke?

Will it help?<p>On the one hand, I think it would be fantastic, if you had automated tests for the law. For example, when German politicians introduced the &quot;hacker law&quot;, you could have pointed out that &quot;This new law would break the &#x27;security researchers need to be allowed to do penetration testing&#x27; test&quot;.<p>On the other hand, &quot;Brexit is in conflict with the Good Friday Agreement, we need a solution for Nothern Ireland.&quot; was known without machine readable laws and test, but politicians ignored it anyway.<p>Maybe what&#x27;s needed is a law that outlaws test-breaking laws and requires politicians to fix the tests first, but I bet that would just result in a lot of &quot;commented&quot; tests.

Laws are often written in most vague language to allow wide interpretation, especially laws regarding treason, communism, foreign agents, anti-war speeches and such things. Programming language won&#x27;t help here.

Can we use it to compute taxes ?

Too have fun with laws&#x2F;rule systems take a look at nomic

Lawyer here.<p>Oh, this again. I suppose this looks relatively harmless, but I&#x27;m always wary of &quot;law is like computer code.&quot;<p>The impulse to think this can strongly solve any real problem <i>in the law</i> is intuitively attractive, but I strongly predict this mostly never happens; it&#x27;s the law&#x27;s job to be intensely practical in the face of hard-edged &quot;computer-like&quot; rules.<p>If anything, you get goofy confusion about what things &quot;are?&quot; My go-to on this is always the &quot;smart contract&quot; -- which can be useful little bits of automated robot money moving code, but emphatically are neither &quot;smart&quot; nor &quot;contracts.&quot;

What problem does this solve? It appears to add precision where it&#x27;s mostly already clear - perhaps it can enforce some kind of rigor... but then like the example given uses &quot;fair market value&quot; as a term which I&#x27;d expect to be the kind of thing that&#x27;s in contention, rather than any of the actual &quot;logic&quot;, and it doesn&#x27;t help with that.<p>The reason we have courts and lawyers is because of the need for interpretation beyond just writing good logic, so I don&#x27;t see how this can really do anything. Or is it for something else?

Completely unrelated to Catalan (Català), the language spoken in Catalonia (Catalunya). I think if someone wants to google a question about this, &quot;catala language beginner hello world&quot; won&#x27;t help them much.

Very cool. Pessimistically, I think that having a clear, understandable view of legal text so that people can navigate the law safely is against a lot of entrenched interests.

Thought this was realated to the Catalan western romance language. Naming could confuse some Spanish and French users.

I know that naming is hard, and it has already been mentioned in the comments here but... I can&#x27;t believe that somebody named a programming language with the exact same name as an existing natural language spoken by millions of people.<p>It just seems like a bizarre decision that can&#x27;t be a benefit at all and can only have negative consequences. Just googling things about it is going to be hard. Why immediately create potential problems for yourselves when you can choose a name that&#x27;s not an issue?