Robot Lawyers and Inequality Part I: The Robotic Rule of Law

For a while, I’ve been worrying about the distributive effects of the invasion of legal practice by technology. But Frank Pasquale is much better at worrying about those issues than I am, and a recent post of his on concurring opinions partly replying to mine has prodded me to think about it a little more loudly.

On a naive view, AI and tools of other computational lawyering are unmitigated boons to the poor. More supplier competition lowers prices! Lower prices benefit the poor! Greater access to legal services! But the real story is much more complex.

There are two distinct worries we might have about the impact of these developments on the poor. First is a worry that Frank raised: people might use these services to oppress the poor, either by themselves accessing cheap legal services and using them to unequally enforce unequal rules against the poor, or by using these technologies to impose barriers to things the poor need. Second, it might have a substitution effect, driving out service providers who do serve the poor (or at least the lower middle class), leaving them with worse legal services than they were getting before.

Thus, this is the first part of a two-part post on the subject. After the fold, some more thoughts on the first of those issues. Later today or tomorrow, some more thoughts on the second.

The First Worry: LegalTech Oppression

Frank tells a grim story of landlords who use legal tech to automatically evict tenants when they fall behind on rent: no longer need the landlord personally take the trouble to show up in court, or even think about whether she or he wants to give the tenant a break---legal tech may be automating the way to the ultimate mean old landlord story: the moment the rent is a day late, out the door you go. And we can proliferate these examples: right now, powerful actors use the legal system to oppress people in lots of ways, and how much worse could it get if they could do so cheaply and with no personal effort?

For example, there’s lots of techno-utopian talk going around right now about computational contracts. The idea there is that if you can fully represent a contract in a computer, you can automate its enforcement. Performance happens automatically, and if the human-controlled parts of performance get breached, penalties happen automatically as well. (In a way, we can think of DRM encryption as a primitive form of computational contract in this sense.) Which sounds great, until you imagine the horrible adhesive contracts that corporations impose on us all unilaterally every day. All of a sudden, computational contracts sound really bad. It might be even scarier than Larry Lessig’s worst fears: not only is code law, in that it can architecturally constrain behavior, but code also acquires the capacity to invoke law too, and back up its electronic walls and gates with the traditional enforcement tools of the law, to wit, people with guns and real-life cages for those who disobey.

Frank also touches on the flip side of this, namely oppression by those who are supposed to be helping. Instead of going to the welfare official who might be able to make a decision, you go to the computer who has no discretion or mercy at all. Frank quoting Bret Scott on fintech: “If you fail to interact exactly correctly, you will not make it through the digital gatekeeper, which – unlike the human gatekeeper – has no ability or desire to empathise or make a plan. It just says ‘ERROR’.“

If we think human gatekeepers in our system have the capacity to empathize or make plans (having spent a few years doing legal aid work, I honestly have my doubts---I interacted with a lot of really heartless welfare officials, but let’s suppose that I got an unusually bad lot)*, this seems like a bad worry.

Solutions? Well, as you may know, I’m a rule of law person. (Shameless plug again: go buy The Rule of Law in the Real World.) And so naturally, all of this sounds to me a lot like the worries expressed classically by folks like Morton Horwitz about the replacement of discretion to treat humans like humans with the bureaucratic and mechanized process of law:

I do not see how a Man of the Left can describe the rule of law as “an unqualified human good”! It undoubtedly restrains power, but it also prevents power’s benevolent exercise. It creates formal equality—a not inconsiderable virtue—but it promotes substantive inequality by creating a consciousness that radically separates law from politics, means from ends, processes from outcomes. By promoting procedural justice it enables the shrewd, the calculating, and the wealthy to manipulate its forms to their own advantage. And it ratifies and legitimates an adversarial, competitive, and atomistic conception of human relations.
(Morton J. Horwitz, The Rule of Law: An Unqualified Human Good?, 86 YALE L.J. 561, 566 (1977))

In a way, the prospect of LegalTech oppression is the ultimate rule of law, on a naive conception of the rule of law according to which it just means “everyone obeys the rules that have been set down.” And if you’re an old-school, vaguely authoritarian, rule of law person, that kind of sounds like a good idea. People should pay their rent, and if they don’t pay it, they deserve eviction; we should obey our adhesive clickwrap contracts with Facebook, and if we don’t, well, bring on the legal Terminators. People are only entitled to welfare benefits if they meet the rules set out in the statute, and if they don’t, the system ought to deny them. Worse, on this story, what the discretionary, pre-rule-of-law approach really means is favoritism and discrimination---maybe the person of a racial background that the welfare official likes gets a nod past the rules, while others do not.

But I actually don’t think that’s what the rule of law is. Rather, the ideals of the rule of law are fundamentally asymmetric: they demand we control the powerful in the aid of equality. In the first instance, that means controlling the direct use of government power, but it also means controlling the private use of government power, including through the disproportionate rights to access government power that private law gives some, but not others, through tools like property and contract.

And that suggests that the real rule of law answer to LegalTech oppression is twofold.

First, it is the substantive law, not necessarily the tools, that must give way. Sometimes, the computational contract favors rather than oppresses the little guy. Think about how nice it would be if we could genuinely hold corporate malefactors like airlines, cable companies, cellphone companies, and the like to providing the services they promise to provide! Instead of fighting the technology that might make enforcing the obligations of the powerful easier as well as enforcing the obligations of the powerless, we need to restructure the substantive legal rules that make it possible for the powerful to lawfully wriggle out of their obligations (like repeal the outrageous consumer protection law preemption provision in the Airline Deregulation Act, for starters). The problem is not the power to enforce the law, but the fact that the law is unjust.

Frank’s eviction story really brings this home. Mass evictions didn’t start with some snotty startup full of landlord-loving brogrammers. It started with foreclosure factories run by the big banks that had humans doing ten-a-days.

Second, not all technologies are neutral, and this includes not just the kinds of technologies that are mediated by computers, but also more foundational kinds of LegalTech, namely the legal innovations that lawyers make without the aid of programmers. (For example, we can describe the practice of bundling and securitizing home mortgages as a technological innovation in law... and obviously not a neutral one.) And so, just like we work to control the legal innovations cooked up by corporate lawyers, we also ought to work to control the legal innovations cooked up by programmers. Maybe, for example, we should pass laws requiring landlords to show up in person (for some sufficiently tricksy definition of “in person” that can meaningfully constrain corporate landlords) in eviction suits, to put a few barriers between them and algorithmic evictions. Perhaps we ought to use Rule 11-like constraints to more carefully police the bulk-filing of lawsuits by large entities.

But: we should be working to restrict access to these tools on the part of the powerful, while enabling their use by the powerless.  If we care about unequal access to justice, we ought to at least try to choose who gets to lower their legal costs, rather than imagining that uncontrollable tech leads to immediate all-or-nothing change. And if we think we don’t have the power to control the distribution of legal tech, that the forces of politics will ensure that technological innovations go first to the rich and powerful with an interest in using them to oppress, then how could we think that we have the power to stop the innovations from happening altogether?

------
* My least favorite public benefits case: I represented a client before the SSA on a disability claim. The claim was so obviously legitimate that the ALJ called me before the hearing, said “why are we wasting our time?” and sua sponte ruled in our favor. The client continued without benefits for months on end, and in the face of increasingly frantic bureaucratic advocacy from me, until, days away from pulling the mandamus trigger, I found a secret phone number and managed to get ahold of someone in authority, who discovered that a clerk had simply not bothered to enter my client’s claim in the system. If this is the human touch, please, bring on the damn robots!