Catellani, P. (2026)
Paper presentation at the 8th National Conference on Legal Psychology, held at LUMSA University
Rome, 11-13 May
Abstract
Counterfactual Reasoning as a Tool for Debiasing in the Analysis of Judicial Cases (Patrizia Catellani, Mauro Bertolotti, and Marco Piastra)
Counterfactual reasoning (“If… then…”) is frequently used in judicial decisions, both spontaneously—following processes of causal attribution and liability based on this type of mental simulation—and in a more structured manner, in accordance with the practices and conventions adopted in the legal field. Numerous studies in cognitive and social psychology have examined the biases associated with the use of counterfactual reasoning in the analysis of legal cases, highlighting how spontaneously generated counterfactuals tend to focus on specific actors and actions, thereby influencing subsequent evaluations. Other studies on the communicative use of counterfactual reasoning have shown that, similarly, counterfactuals can be strategically employed by attorneys, judges, expert witnesses, and others to guide and influence attributions and evaluations in a legal case. In this paper, however, we will analyze the use of counterfactual reasoning for the opposite purpose: to broaden and make the processing of information on legal cases more flexible and accurate, including through the use of generative artificial intelligence. Some research has shown that generating or being exposed to counterfactuals encourages a tendency to adopt different perspectives and to evaluate events more accurately.
Starting with a case of medical malpractice, we will demonstrate how the generation of counterfactual scenarios based on the modification of various antecedents can be integrated into the decision-making process. We will also show how this process can be aided by artificial intelligence, which automatically generates such scenarios based on specially designed prompts. The discussion will address the potential of this approach and its limitations, including those arising from the use of generative language models trained on texts that may contain other sources of bias.
Keywords: counterfactual reasoning; communication; artificial intelligence; debiasing