It’s been a while since my last post on the Volokh conspiracy. In 2021, I became an associate dean at George Washington and had no time to write. Last year I changed roles to associate dean and my portfolio shrank, so I was fortunate to have some time to return to scholarship and complete several papers. I will begin my return to blogging by writing a series of posts offering shorter versions of the key arguments in a recently completed article that I have now submitted to law journals, entitled The cost of justice at the dawn of artificial intelligence.
The article explores how changes in lawyer productivity affect the legal system and how legal actors should prepare for a future that may feature more or less productive lawyers than today, depending in part on how artificial intelligence develops. The article’s simplest claim is that the legal literature should pay more attention to the productivity of the legal sector, because changes in legal productivity directly affect the cost of legal services. Legal fees, in turn, impact how the legal system can carry out its critical mission of ensuring that similar cases are treated equally, while cases are treated differently when applicable legal principles require it.
The preeminent economic model for considering changes in the costs of goods and services over time is William Baumol’s cost disease. This model is perhaps best known because it provides an explanation for why the costs of college education and health care have risen faster than inflation in recent decades. The essential story is that these sectors have enjoyed lower productivity gains than productive sectors such as agriculture, textiles or technology, and so products have become more expensive. Arguments that a given market is suffering from cost disease can be contested. Perhaps, one might argue, the costs of education have risen in large part because the primary educational product—a professor lecturing in a classroom—has been paired with ever-more lavish complementary services, such as ever-better food and an ever-increasing number of administrators to help students navigate through college.
Whatever the case in each market, however, the cost disease story is virtually tautological. When markets benefit from technological advances that increase productivity per worker, costs fall relative to costs in industries where technology has been relatively stagnant. If costs in productive sectors decrease relative to costs in stagnant sectors, then costs in stagnant sectors must increase relative to costs in productive sectors. If we then control for the overall price level, which depends mainly on macroeconomic policy, relatively stagnant sectors must become relatively more expensive over time. We can argue in any market, including the legal sector, whether stagnation is occurring, whether apparent, material inflation is attributable to cost disease or quality improvement. But the very definition of productivity implies that if an industry stagnates, its products will become more expensive than other industries.
The article considers the implications of legal productivity both retrospectively and prospectively. The retrospective question is whether lawyer productivity has remained stagnant and, if so, how this has affected the legal system. The potential question is whether we should expect AI to increase the productivity of lawyers and therefore reduce the costs of legal services relative to other goods and services in the economy. These investigations frame the ultimate question of how legal actors might prepare for a world in which legal services might be considerably more or less expensive in relative terms than they are today.
These topics will be explained in future blog posts (and will obviously be fully developed in the article). For now, I will note two points from different parts of the article that highlight how considering changes in productivity, past and future, can be important for understanding the future of the legal system.
Over the past few decades, lawyers and legal scholars have observed a sharp reduction in the number of cases brought to trial, both in civil and criminal courts. This transformation has been called the “disappearing process,” and the name is not hyperbole. The percentage of federal cases that go to trial has declined by more than half in the two decades since academics began seriously discussing the dying process, and declines have also occurred in state courts and elsewhere . John Langbein argued in Yale Law Journal that the disappearing trial is attributable to changes in procedural rules, and while this may be a partial explanation, it fails to answer why the number of trials has declined so systematically, rather than just in periods of reform procedural. While commentators certainly understand that the high cost of legal services helps explain why cases settle, they have not considered whether rising costs over time, potentially attributable to cost disease, can explain historical trends. Disappearing processes are exactly what you would expect from simple transactional contracting models in an environment where costs are rising.
The potential point is that if the legal system developed based on implicit assumptions about the cost of legal services, changes in legal productivity could change the balance of power in legal regimes. For example, lawmakers could allow longer maximum criminal sentences than lawmakers think justice requires, if lawmakers believe long potential sentences are necessary to empower prosecutors in an environment where it is impractical for public ministries take many cases to court. Suppose, however, that AI significantly reduces the time required for pretrial work, for example by efficiently sifting through evidence, creating trial plans, and assisting in drafting briefs. Armed with a credible threat to take more cases to trial, prosecutors should demand harsher sentences in plea deals. Judges will disagree about whether such a change is healthy or harmful, but awareness of how changes in productivity can revitalize or disrupt established legal regimes is the first step toward conscious consideration of how the legal system should evolve following changes in productivity.
But has the law really remained stagnant? And if so, will artificial intelligence reverse this trend? I will address these questions in the next two blog posts.