“Fooled” by Artificial Intelligence? A Case Study on Judicial Argumentation in Response to (Potential) AI Use in the High Court of Ireland

Abstract Emerging technologies have had a discernible impact on how judicial systems operate and how judges serve their functions. Improvements in software and the evolution of artificial intelligence tools may unlock the full potential for judicial systems to deliver more efficient and potentially more effective justice. The challenges posed by AI, however, can be just as significant. In the Republic of Ireland, the use of AI in legal submissions, along with its less-than-accurate results, has so far mainly been flagged in two cases, i.e. Coulston et al. v. Elliott and Elliott and Reddan v. An Bord Pleanála . In this work, a qualitative analysis of Coulston and Reddan through the pragma-dialectical framework [1, 2] led to uncover the structure of the High Court’s argumentation. As far as the alleged use of AI-generated materials on the losing litigant’s part is concerned, evidence shows that the Court deployed argumentation from coherence [3]. In both cases, the Court was observed to have resorted to the negative variant of the argument, according to which a solution put forward by a party is not required (or rather discouraged) by the coherence of the legal system. Moreover, both cases confirm that the argument from coherence is a form of complex argumentation including a main argument and a subordinate argument, whereby several reasons are put forward in defence of the standpoint that the case should be dismissed because of a putatively inappropriate use of artificial intelligence in the relevant submissions