A Tiny, $25 Million Mistake (2014)

It depends on the contract.

Tbh, I would prefer if things government related were set in stone. No ambiguity, send the money, be done.

The private sector can work with customers and are reliant on customers for business. This makes it so compromise and being reasonable is incentivized by additional profit.

(lawyer here, but i don't make a living at it so i have no care :P)

Empirically, they are in fact scalable.

The number of contracts litigated in the situation you are talking, as a percentage of overall contracts in those situations, is ridiculously small.

(choose any metric, in fact, and you'll see they've scaled just fine)

Meanwhile, the thinking you are espousing turns out to be pretty much impossible in practice even if you wanted it to be so. I get the desire by engineers, and even the desire in some forms of contracts. But in the general case, in fact, it makes it much much worse than we have now. In one case, you have a judge looking at the spirit of the contract and making some reasonable decision about what should happen.

In the other, you have to have completed the often impossible task of thinking and codifying every edge case and possibility. If you get it wrong, too bad, so sad.

I know which i'd want.

One of the reasons we have judges in the first place is because the desire to codify all possibilities is not achievable. It's not a thing we sanely have achieved in thousands of years of law, contracts, etc.

The idea that because we are now expressing it in a bytecoded language and not words, it will suddenly become possible to achieve, seems incredibly wrong headed, because the language was never the problem. The contracts you can effectively express this way were never a problem from the language perspective. It definitely makes them more automatic, but making these contracts more automatic is basically "meh" on the innovation scale[1].

[1] Though i'm not crazy and certainly believe there are many many billions of dollars to be had in doing this.

A good example of this concept in action is actually the whole Ethereum / DAO fiasco, where the contract code actually contained a bug. Turns out the bug was not intended to be “binding”, and thus the Ethereum chain was forked to undo any damage of the bug.

So seems like even the crypto / smart contract community agrees with this sentiment to some degree, although forking an entire block chain might be somewhat of a crude interpretation of “wiggle room”.

That seems to imply we need two pieces of code for "perfect laws as code": 1) that which writes/defines the laws and 2) that which interprets the prevailing milieu and determines the application of law.

Since we're speculating anyways, why not implement the whole system we currently have? Assuming humanlike-AI (heh!), the following services are imaginable:

    1 AI Legislature
    2 AI Judge
    3 AI Lawyers

Govt programmers handle (1) and (2), applicants and respondents (or prosecution and defence) handle (3). AI Legislature determines laws according to citizen referendum and policy upon consultation with the the AI Judges. Each AI Lawyer makes arguments to the AI Judge who evaluates each against the law and jurisprudence to arrive at a judgement. Wiggle away!

As a final layer which arose out of the Luddite massacres of 2137-40, a group of rubber-stamping human judges at the top serving as final adjudicators and appellate court.

> Putting perfect laws as code is essentially a humanlike-AI-complete problem. That is, the necessary code would also have to implicitly and completely implement our intuitions about morality, and the mechanisms behind emotions and suffering.

I don't think this applies to all areas of law. Take tax law, for instance. Is there any moral or ethical or otherwise well-founded argument in favour of having tax laws interpreted less strictly than a machine would?

After reading the book Ordinary Injustice, I can assure you the US legal system is extremely far from humane. Our system is overworked, biased, and inherently breaking down in th invisible small courts of our country. Fairly enforced laws may seem inhumane, and I am sure there would need to be patches to any such system to refine the morality it reflects, but I am having a hard time imagining it being any less humane. Secondarily, you can simply issue a "fix" contract for 25 million extra. If the other business is operating in good faith, which they obviously were they would be just as willing to have an additional contract to make it right. I am not sure I am a fan of smart contracts or not yet. There can still also be overrides for humanitarian purposes.

Playing devil's advocate, allowing for 'wiggle-room' in laws encourages vague/hand-wavey/overly broad laws. Because the devil is always in the detail.

This allows for two key issues to manifest:

1) Selective enforcement 2) A non-expert can't know whether they are breaking any laws in advance of action

I am skeptical that we can get well defined laws such that algorithmic enforcement is feasible before say mathematical proofs of computer software is widespread. If we can't/don't practically do the latter, I'm not sure how we expect the former to become possible in a general sense.

On the other hand, maybe there is a useful narrow subset of contracts/contract law that is possible to implement. Aren't there some contract areas which are written with fairly extensive references to some set of uniform agreements?

I think there is room for both. Start with very simple contracts first.

> The only thing that is rock solid and predictable is contract written in programming code.

The only code that's 'rock solid and predictable' is code that's been formally proven. Are smart contracts doing that?

Further, as described above, even if they were perfectly written every time, they'd still be a poor replacement for human law. We are squishy, imperfect and inconstant, we have legal protections and legal wiggle-room for a lot of reasons.

There's not many use cases where a purely technical solution has meaningful impact on the world, though. The "code-as-law" will generally have to be supported by "law-as-law" at some point, and that can be overturned arbitrarily.

What good is having technical ownership of some private key if your government declares you have to give it over or face jail?

How does that differ from a standard legal agreement? At any time two parties can make a new agreeement, with or without a court battle. Your immutable record is documents filed with the government on public record.

That's true, but let's make it just slightly more complex: Our contract is to swap 10 EUR for whatever it is worth tomorrow in USD according to what "bloomberg.com" gives as the representative rate at 12:34pm.

Now, this is easily codified, but ...

a) Someone could hijack the "bloomberg dns"

b) Someone could hijack bloomberg's BGP

both the above would likely be defended by using https, but CAs have been hacked before, and there are questionable state controlled ones.

c) Someone could hack inside Bloomberg itself to give the wrong rate

d) Bloomberg might have a bug that is struck exactly at 12:34pm on that day, which shows the wrong rate.

All these 4 things may happen independently of our contract - Bloomberg is not a side to our agreement, and other actors might have reasons to try to make it give wrong data.

In "The real world", if the rate is really off (say, half what you expect), you'd look at Thomson/Reuters and Yahoo, realize Bloomberg is borked, and react accordingly - and a court would likely support that court of action.

This is the kind of wiggle room you want, because in the real world, it is impossible to lock down all of these improbable, weird, but possible, outcomes.

Contracts that are so simple that there's no potential need for wiggle room aren't very useful.

In real world contracts, the more money is involved, the more both sides try to protect themselves against such weirdness; and yet, courts are sometimes involved, either because interpretation is disputed (comparable to "code has underfined behaviour") or due to events not covered in the original contract.

There's an entire genre of stories about deals with the devil or supernatural items ("Monkey's Paw"), where the deal is binding - and the punishment is getting exactly what you asked for.

That didn't work out so well for Shylock.

It seems to me that the set of useful contracts you can write that are completely unambiguous is very small.

Consider an employment contract on a blockchain.

There doesn’t seem to be any way to do this without relying extensively on oracles that reflect fuzzy human judgement about what the terms of the contract really mean, let alone whether certain actions took place, let alone whether those met or violated the terms.

Aside from the entertainment value of fixed-in-stone contracts, there is also the issue (touched on in the article) that the result would slow the economy drastically.

> On the other hand, even on the blockchain it's not hard to make a contract amendable with unanimous consent.

This. Both proponents and critics of blockchain often seem to forget that it's just technology. Entities trading on it and performing other actions are still subjects to applicable laws, whether they like it or not.

After the last email said $425 million? Probably only in America.

> even on the blockchain it's not hard to make a contract amendable with unanimous consent.

Unanimous consent is nice when it happens, but we typically want the ability to make amendments without unanimous consent, too.

Concrete examples: I wouldn't want to use an electronic payment card that doesn't have some sort of chargeback mechanism. My mutual funds want the exchanges they operate on to have some mechanism for busting erroneous trades, and I'm glad that they do. When I check in at a hotel, they want a credit card on file, so that they can charge me if I steal the towels.

All of these mechanisms involve retroactively altering the financial terms of a contract in a way that causes one person to lose some money without their consent. For the most part, people agree that this is just and desirable, because they also value fairness.

That said, I agree on the point at hand -- I don't think that the example from the story is particularly relevant here.

How hard is "not hard"?

The example is the sort of thing where you strike through a number on the contract, write a new one, and everybody initials it.

> If they hadn't agreed and the issue had been brought to court, the $400 million number almost certainly would have been enforced.

In the Netherlands, I wouldn't be so sure about that. Law says that the person signing the contract needs to be fully aware of what they're signing. If they can show an e-mail which shows they were expecting 425, it's completely up to the lawyer and will depend on context of the e-mail etc.

Replying to self since I can't edit.

I may have been a bit too confident about whether the text would have been enforced. It actually is possible (in the United States) for a court to alter the terms based on evidence of a clerical error, if the evidence is strong enough.

https://www.robinskaplan.com/~/media/pdfs/its%20a%20mistake%...

https://chicagounbound.uchicago.edu/cgi/viewcontent.cgi?arti...

To summarize, there's something called the "parol evidence rule" which boils down to: if you have a final written agreement, and the text of the agreement is clear, you usually can't introduce extrinsic evidence (e.g. from the negotiation process) that the parties' intent was something other than what was written. However, there are a number of exceptions to that rule, and one of them is that you can introduce extrinsic evidence to try to prove the existence of a "mutual mistake", which can be grounds for "reformation" of a contract (i.e. correcting the mistake). A "mutual mistake" is when both parties are mistaken; on the other hand, a "unilateral mistake" (by one party) generally will not be corrected, unless there's also an element of fraud. Mutual mistake must be proven by the high standard of "clear and convincing evidence", as opposed to the normal standard in civil cases of "preponderance of the evidence", and in general there's a heavy presumption that the text reflects the parties' intent. But "scrivener's error" is often a valid basis for showing a mutual mistake. It depends on the quality of the evidence; also, the standards vary somewhat by jurisdiction.

> An example of a contract in which it is useful have wiggle room is hiring a company to build a website for you. The best process for building a website necessarily includes discovery and continuous decision-making about what to build, so a long and prescriptive contract that spells out every product feature on day one will most likely result in a crappy final product and/or a ton of change orders and amendments.

The discussion seems to be about allowing wiggle room in an otherwise 'non-wiggly' contract, which doesn't seem to be the situation you're describing. Your suggestion clearly isn't that the long and prescriptive contract should be drawn up and then repeatedly amended, but presumably is rather that the initial contract itself should provide a mechanism for easy changes; and contracts enforced by technology, like any others, can be created in this way.

I was really surprised by that when my friend bought his house. It's a sneaky way to get around the fact that the bank won't finance anything that's not secured by the house. So say you are buying a house for 200k, but after the down payment, you don't have the money to redo the kitchen like you wanted. No problem, the seller adds 3k to closing costs and hands you a check for 3k. Your bank hands you a check for 203k, You pay the seller, and suddenly you have 3k that your bank wouldn't finance directly to redo your kitchen.

Interesting, given that health insurance is notorious for the opposite (failing to pay for things that the purchaser might resonably expect to be covered).

What are these historical reasons?

>(Indeed, when it came to light that some insurance companies were just sitting on unclaimed payments for small-dollar life insurance policies, the regulators had them reopen the books going back sixty years.)

Doesn't this undermine your point? Insurers were plenty happy to sit on life insurance payments and not even attempt to figure out who to pay them to, thus betraying what you state as their #1 rule and jousting with regulators and the public over actuarially insignificant improvements in their loss rates.

>I realize technologists feel like insurance companies are malevolent entities,

No, it isn't that. I feel like every large company behaves almost always like a machine whose job is to print money, with no consideration to morals, ethics, or decency unless they believe the good PR from such will cause them to be able to print more money.

I believe this because it is how they (usually, not quite always of course) behave, as they are beholden to their investors to print money.

I was curious too, seems they filed for chapter 11 and are hoping that debt gets canceled as well?

https://money.cnn.com/2015/07/22/news/companies/caesars-bank...

Yup, that's how I found the link.

I'd sure love to see the EU try and assert extraterritorial jurisdiction over an American organization publishing news explicitly intended for American audiences...

I agree, just something I'm not used to seeing on the internet, I felt like it was just getting into it, and then it was over, and I start to wonder how it became an article in the first place...

Agreed, I for one appreciated the writing style that had a very delightful story to tell yet didn't wax on and on.

thank you!