Microlawsuits

The speed at which a lawsuit is decided depends on the number of judges and their productivity. A judge is a scarce, treasured, exotic piece of human capital; a person who can speak with the voice of the common law. Judicial productivity has many factors: the raw efficiency of the judge; whether your court system is efficient; whether it is rigorous about timelines; whether it has integrity’; whether it is competent.¹ You have faster courts if you have higher judicial output per capita.

How much faster can you make courts? If you want the resolution time to be days or months, you need to increase judicial output per capita to levels which permit that resolution time. The complexity of the dispute doesn’t matter that much. When an ordinary lawsuit takes ten years, it is not because a judge has spent her fifties racking her brains for a solution to your property dispute. It is because your legal system is not built for fast resolution.

Now, let’s make it faster. Imagine that you have AI systems that put cheap intelligence at your disposal, that can read the record and deliver instantaneous results. You have near infinite judicial output. The per unit cost of resolution is near zero, and resolution time is measured in seconds.²

This is not incredibly far off, because we already have instant resolution in low-stakes cases — these are usually dispute resolution procedures employed by companies to avoid lawsuits. When you demand a refund from a food delivery app, you are engaging with a dispute resolution process that is often faster and better than a consumer court.

But you wouldn’t use that process for anything important. Imagine a lawsuit, decided in seconds, about the boundaries of Sudan. The problem with this is not that people do not believe it is possible to interpret the boundaries correctly in seconds. The problem is that an incredibly fast resolution one way or another would seem like an illegitimate process tothe litigants. Resolution in a few months with a lot of agonising about the difficulty of the decision? Great. Resolution in a second? You have succeeded in producing a decision, but utterly failed at resolving the dispute – your litigants are raising their armies even as your AI judge’s HAL 9000 voice reads the judgment.

Courts are a branch of government, and so need legitimacy and authority to perform their functions. Even supremely efficient governments may lose the mandate of heaven. In Tiers (1984), Judith Resnik offers the example of a trial judge in New York who decided whether to sentence a defendant to 20 days or 30 days based on a coin flip. There was massive outrage and the judge was censured. Resnik asks:

What was so offensive? The coin flip produced an outcome, inexpensively and quickly. Moreover, the judge’s critics did not claim that the decision itself incorrectly reflected either law or fact, or that the time ultimately to be served by the defendant was unjustly long or short.

The complaint was about process. The coin flip offended this society’s commitment to rationality. The open embrace of chance as determinative was frightening. The decisionmaking procedure was also uncomfortably commonplace; people flip coins to decide who must wash dishes or go first in a game. State-backed orders to incarcerate are more seemly if made in a way that differentiates them from those involving dishwashing. The import of the judge’s decision, that a person had to spend many nights in a cell, was not reflected in the procedure. The community’s outrage expressed feelings that some decisions should be treated specially, perhaps garbed in ritual, and certainly made to appear rationally and carefully chosen.

Everything about the judicial process, including certain binding constraints on speed, are imposed by the need to ‘construct the authority of the court’ and ensure ‘a certain, careful process in each case.’³ This is a requirement of solemnity, and to our monkey brains, doing things slowly means doing them solemnly and carefully.

There is a common law principle – de minimis non curat lex – which means that the law does not concern itself with trifling matters. When the claim is for less than a certain amount, it’s not important enough for the legal process. When a breach of contract is only a technical breach and doesn’t cause harm, the law refuses to intervene. You don’t waste the common law’s time on a ten dollar dispute. This is partly because judicial time is scarce. But partly it is also because it is hard to feel awe for an institution that deals with trivial things. One purpose of de minimis is to construct and maintain judicial authority.

To summarise, until you invent AI judges, your binding constraint is judicial output. After you invent AI judges, your binding constraint is sociology.

You need to now construct the authority of the AI judge. Aiming for AI judges who instantaneously provide decisions on par with the best human judges is a good start, but it’s ultimately unsatisfying. I recommend massive overambition. The quality of our legal decisions is presently affected by political and socioeconomic considerations; and also limited by our status as slightly-advanced chimps.

Instead, we should demand perfection from the AI judge. I am kidding, but only a little bit? In Taking Rights Seriously (1977), Ronald Dworkin invents a judge, Hercules, of “superhuman skill, learning, patience and acumen” who was capable of an internally coherent, logical understanding of the legal system.⁴ Posed with a legal dispute, he would construct an internally consistent theory of the state, of government, of law, and produce a decision that was consistent with that theory:⁵

You will now see why I called our judge Hercules. He must construct a scheme of abstract and concrete principles that provides a coherent justification for all common law precedents and, so far as these are to be justified on principle, constitutional and statutory provisions as well.

We might not accept an instantaneous decision from a machine if it were only a judge. But perhaps we might accept it from a machine that was the common law itself. As for me, I would tell Hercules to pretend to take at least a week on the case anyway, and hire a (human) PR consultant. And bodyguards. You know, just in case.