Amin Ebrahimi Afrouzi

Karl Llewellyn’s critique of the canons of statutory interpretation led to a decline in their use for several decades. His critique, however, faced sustained resistance from some corners of the academy and the judiciary. Although this resistance has had only a selective uptake, it animated a gradual revival of the canons and brought the state of scholarship to an impasse that is for the most part partisan. In this article, I examine the semantic canons from a deeper level and argue that a universal assumption about them is false. Said assumption is that, although not dispositive, the semantic canons at least offer some reasons in favour of or against a candidate interpretation. Inclinations to rely on the semantic canons are also based on this assumption, although it is an assumption that the critics of the canons also share. I argue that this assumption is false because the semantic canons are a class of rules that are by nature not reason-giving. This provides a new ground against giving the semantic canons deliberative weight in questions of statutory interpretation.

In a recent article, Kiel Brennan-Marquez and Stephen E. Henderson argue that replacing human judges with AI would violate the role-reversibility ideal of democratic governance. Unlike human judges, they argue, AI judges are not reciprocally vulnerable to the process and effects of their own decisions. I argue that role-reversibility, though a formal ideal of democratic governance, is in the service of substantive ends that may be independently achieved under AI judges. Thus, although role-reversibility is necessary for democratic governance when human judges are on the job, it may not be so when AI judges replace them. One broader implication for normative evaluation of disruptive technologies follows: formal and substantive ideals that often align must be independently examined in the evaluation of disruptive technologies. This is because these formal and substantive ideals may no longer align under the factual circumstances that come to govern when such technologies are deployed.

The epistemic value of a political procedure—such as democracy or a civil trial system—depends on how well it performs in arriving at decisions that are correct by some independent standard. A core assumption in the literature on epistemic democracy is that boosting the epistemic value of such a procedure makes it better overall. Even though this assumption seems innocuous (and hence has not been discussed in much detail), we will argue that it is not beyond the pale of reasonable disagreement. For it is possible to increase the epistemic value of a political procedure in ways that give rise to egalitarian objections.

The rich evade conviction more often in criminal trials than the poor. They also win more often in civil cases against the poor. Given that money buys better lawyers and better lawyers are instrumental in winning in adversarial trials, the rich have a structural advantage in laissez-faire trial systems. Such inequality is concerning. In a landmark article, Alan Wertheimer argues that we should level down legal resources in civil cases on the basis that doing so increases the adversarial system’s accuracy—that is, its chance of reaching correct decisions. In a more recent article, along similar lines, Shai Agmon also advocates that, given some constraints of adequacy, legal resources should be leveled down in both civil and criminal cases. This article aims to show that such arguments fail because leveling down legal resources could decrease a trial system’s accuracy, making it worse by Wertheimer’s or Agmon’s own criteria.

This paper argues that an AI judge is conceptually undesirable and not just something that lies beyond the state of the art in computer science. In a nutshell, even if an AI system could accurately predict how a good human judge would decide a particular case, its prediction would be the product of correlations between such factors as patterns of syntax in bodies of legal texts. This approach of AI systems is insufficient for basing their output on the sort of rationales that are expected of valid judicial decisions in any desirable legal system. Thus, by their very nature, AI systems are incapable of providing valid legal decisions in any such system.

A longstanding orthodoxy holds that the Mohists regard the promotion of li (benefit, 利) as their ultimate normative criterion, meaning that they measure what is yi (just, 義) or buyi (unjust, 不義) depending on whether it maximizes li or not. This orthodoxy dates back at least to Joseph Edkins (1859), who saw Mozi as a utilitarian and an ally of Bentham. In this paper, we will argue that this orthodoxy should be reconsidered because it does not square with several passages from the Mozi. That the Mohists place a strong weight on the promotion of ‘li for the whole world (tianxia zhi li, 天下之利)’ is uncontroversial. We argue, however, that in certain cases the Mohist moral calculus diverges in its rationale or outcome from li-promotionalism. This position rejects the orthodoxy by showing that Mohism and li-promotionalism are not entirely coterminous.

The Mohists develop the concept of yi 義 to denote what is morally right in a normative sense. We argue that this concept has, as one of its necessary conditions, a requirement to not harm others. Additionally, we will show that the motivation of developing this concept is that it can be both universalized and publicly agreed upon, thus serving the Mohists’ endeavor to overcome human conflicts that make the world chaotic and unlivable. We argue therefore that the Mohist notion of gongyi 公義 is best captured by the requirement of “not harming.” Moreover, we shall show that the Mohists themselves committed to this requirement as the first principle of all their core doctrines. Finally, we argue that due to the justifiability and universalizability of the requirement of “not harming,” the Mohists’ commitment to it as a prerequisite normatively justifies their other values in a way that was unprecedented at the time.