Artificial Intelligence and Role-Reversible Judgment

Abstract

Intelligent machines increasingly outperform human experts, raising the question of when (and why) humans should remain ‘in the loop’ of decision-making. One common answer focuses on outcomes: relying on intuition and experience, humans are capable of identifying interpretive errors — sometimes disastrous errors — that elude machines. Though plausible today, this argument will wear thin as technology evolves.

In this Article, we seek out sturdier ground: a defense of human judgment that focuses on the normative integrity of decision-making. Specifically, we propose an account of democratic equality as ‘role-reversibility.’ In a democracy, those tasked with making decisions should be susceptible, reciprocally, to the impact of decisions; there ought to be a meaningful sense in which the participants’ roles in the decisional process could always be inverted. Role-reversibility infuses the act of judgment with a ‘there but for the grace of god’ dynamic and, in doing so, casts judgment as the result of self-rule.

After defending role-reversibility in concept, we show how it bears out in the paradigm case of criminal jury trials. Although it was not the historical impetus behind the jury trial — at least, not in any strong sense — we argue that role-reversibility explains some of the institution’s core features and stands among the best reasons for its preservation. Finally, for the sci-fi enthusiasts among us, role-reversibility offers a prescription as to when the legal system will be ready for robo-jurors and robo-judges: when it incorporates robo-defendants.

Keywords: Artificial Intelligence, Humans in the Loop, Criminal Procedure, Judgment, Jurisprudence, Democratic Theory

JEL Classification: K10, K14