Bail decisions present a fundamental challenge for data-driven decision support systems. When bail is denied, the counterfactual outcome of whether the defendant would have appeared in court remains unobserved. As a result, historical bail data embed structural label indeterminacy: future decisions are influenced by past decisions whose outcomes are only partially knowable. Building automated systems on such data risks introducing bias and reinforcing feedback loops. This raises a core question for machine-learning systems intended to assist judicial actors: how should cases in which bail was denied be treated during model development? In a case study of bail decisions from the Unified Judicial System of Pennsylvania, we evaluate five contemporary approaches to handling label indeterminacy across three machine learning models, including a novel label imputation method motivated by the dynamics of bail decisions. Each method relies on unverifiable assumptions, yet all influence the models' predictive behaviour, sometimes even more so than the choice of model itself. Explainable AI analysis further reveals that these effects extend to the models' internal decision-making processes as well. Finally, we consider the notion of label indeterminacy from a legal perspective and assess the legitimacy of these approaches in the context of bail decision-making.

Artificial Intelligence is increasingly applied to the field of law, and has the potential to increase access to justice. One particular movement that is gaining traction is that of agentic AI, wherein AI agents, based on Large Language Models (LLMs) can take autonomous actions. In particular, multi-agent approaches in the legal domain remain largely unexplored. In this paper, we investigate multi-agent deliberation methods for legal reasoning tasks using LLMs. We explore multi-agent deliberation (MAD) and introduce two novel multi-agent frameworks inspired by courtroom procedures and legal argumentation. Our experiments on both legal and non-legal benchmarks reveal that multi-agent frameworks achieve comparable overall performance to baseline large language models, but produce significantly distinct answers. Notably, these approaches can successfully solve cases that the baseline fails to address, and vice versa. We conduct a qualitative evaluation and highlight scenarios where multi-agent frameworks outperform monolithic approaches. For example, multi-agent approaches appear better suited for answering questions that require critical thinking from multiple perspectives. Our work positions multi-agent systems as a promising direction for AI in the legal domain, while demonstrating the potential of law-inspired multi-agent approaches for deliberation.

Legal reasoning is complex and multi-faceted, requiring a broad set of skills. By employing domain knowledge from legal experts, we design five elements that can be included in prompts for large language models that could aid in legal reasoning tasks. We use additional legal guidelines, 1-shot prompting, dictionary definitions, knowledge representations of legal articles, and IRAC-style prompting. We investigate the effect of each prompt element on the model’s performance on a legal entailment task. Certain prompt elements can improve performance, depending on the context and the model. For the smaller models, increasing the number of prompt elements improves performance on average. For any particular combination of model and sub-task, only using a subset of the prompt elements seems to work best. For the most advanced reasoning model we evaluate, using a selection of prompt elements increases average performance across all evaluated sub-tasks. Results indicate that the problem space of the legal entailment task may be too large for a single model and prompt. In future research, we therefore aim to investigate the capabilities of an ensemble of specialized models.

In this position paper, we advocate for the development of conversational technology that is inherently designed to support and facilitate argumentative processes. We argue that, at present, large language models (LLMs) are inadequate for this purpose, and we propose an ideal technology design aimed at enhancing argumentative skills. This involves re-framing LLMs as tools to exercise our critical thinking skills rather than replacing them. We introduce the concept of reasonable parrots that embody the fundamental principles of relevance, responsibility, and freedom, and that interact through argumentative dialogical moves. These principles and moves arise out of millennia of work in argumentation theory and should serve as the starting point for LLM-based technology that incorporates basic principles of argumentation.

Machine learning is increasingly used in the legal domain, typically operating retrospectively by treating past case outcomes as ground truth. However, legal outcomes are often shaped by human interventions that are not captured in most machine learning approaches. A final decision may result from a settlement, an appeal, or other procedural actions. This creates label indeterminacy: the outcome could have been different if the intervention had or had not taken place. We argue that legal machine learning applications need to account for label indeterminacy. Methods to impute indeterminate labels exist, but are all grounded in unverifiable assumptions. In the context of classifying cases from the European Court of Human Rights, we show that the way that labels are constructed during training can significantly affect model behaviour. We therefore position label indeterminacy as a relevant concern in AI & Law and demonstrate how it can shape model behaviour.

Generative large language models as tools in the legal domain have the potential to improve the justice system. However, the reasoning behavior of current generative models is brittle and poorly understood, hence cannot be responsibly applied in the domains of law and evidence. In this paper, we introduce an approach for creating benchmarks that can be used to evaluate the reasoning capabilities of generative language models. These benchmarks are dynamically varied, scalable in their complexity, and have formally unambiguous interpretations. In this study, we illustrate the approach on the basis of witness testimony, focusing on the underlying argument attack structure. We dynamically generate both linear and non-linear argument attack graphs of varying complexity and translate these into reasoning puzzles about witness testimony expressed in natural language. We show that state-of-the-art large language models often fail in these reasoning puzzles, already at low complexity. Obvious mistakes are made by the models, and their inconsistent performance indicates that their reasoning capabilities are brittle. Furthermore, at higher complexity, even state-of-the-art models specifically presented for reasoning capabilities make mistakes. We show the viability of using a parametrized benchmark with varying complexity to evaluate the reasoning capabilities of generative language models. As such, the findings contribute to a better understanding of the limitations of the reasoning capabilities of generative models, which is essential when designing responsible AI systems in the legal domain.

We summarize the first Workshop on the Design of Responsible Hybrid Intelligence (RHI2023), co-located with the 2st International Conference on Hybrid Human-Artificial Intelligence (HHAI 2022), held on June 27, 2022 in Munich, Germany.

Data-driven AI systems can make the right decisions for the wrong reasons, which can lead to irresponsible behavior. The rationale of such machine learning models can be evaluated and improved using a previously introduced hybrid method. This method, however, was tested using synthetic data under ideal circumstances, whereas labelled datasets in the legal domain are usually relatively small and often contain missing facts or inconsistencies. In this paper, we therefore investigate rationales under such imperfect conditions. We apply the hybrid method to machine learning models that are trained on court cases, generated from a structured representation of Article 6 of the ECHR, as designed by legal experts. We first evaluate the rationale of our models, and then improve it by creating tailored training datasets. We show that applying the rationale evaluation and improvement method can yield relevant improvements in terms of both performance and soundness of rationale, even under imperfect conditions.

Approaches to court case prediction using machine learning differ widely with varying levels of success and legal reason-ableness. In part this is due to some aspects of law, such as justification, being inherently difficult for machine learning approaches. Another aspect is the effect of design choices and the extent to which these are legally reasonable, which has not yet been extensively studied. We create four machine learning models tasked with predicting cases from the European Court of Human Rights and we perform experiments in order to measure the role of the following four design choices and effects: the choice of performance metric; the effect of including different parts of the legal case; the effect of a more or less specialized legal focus; and the temporal effects of the available past legal decisions. Through this research, we aim to study design decisions and their limitations and how they affect the performance of machine learning models.

This paper provides a formal description of two legal domains. In addition, we describe the generation of various artificial datasets from these domains and explain the use of these datasets in previous experiments aligning learning and reasoning. These resources are made available for the further investigation of connections between arguments, cases and rules. The datasets are publicly available at https://github.com/CorSteging/LegalResources.

In AI and law, systems that are designed for decision support should be explainable when pursuing justice. In order for these systems to be fair and responsible, they should make correct decisions and make them using a sound and transparent rationale. In this paper, we introduce a knowledge-driven method for model-agnostic rationale evaluation using dedicated test cases, similar to unit-testing in professional software development. We apply this new quantitative human-in-the-loop method in a machine learning experiment aimed at extracting known knowledge structures from artificial datasets from a real-life legal setting. We show that our method allows us to analyze the rationale of black box machine learning systems by assessing which rationale elements are learned or not. Furthermore, we show that the rationale can be adjusted using tailor-made training data based on the results of the rationale evaluation.

The justification of an algorithm’s outcomes is important in many domains, and in particular in the law. However, previous research has shown that machine learning systems can make the right decisions for the wrong reasons: despite high accuracies, not all of the conditions that define the domain of the training data are learned. In this study, we investigate what the system does learn, using state-of-the-art explainable AI techniques. With the use of SHAP and LIME, we are able to show which features impact the decision making process and how the impact changes with different distributions of the training data. However, our results also show that even high accuracy and good relevant feature detection are no guarantee for a sound rationale. Hence these state-of-the-art explainable AI techniques cannot be used to fully expose unsound rationales, further advocating the need for a separate method for rationale evaluation.

In AI and law, systems that are designed for decision support should be explainable when pursuing justice. In order for these systems to be fair and responsible, they should make correct decisions and make them using a sound and transparent rationale. In this paper, we introduce a knowledge-driven method for model-agnostic rationale evaluation using dedicated test cases, similar to unit-testing in professional software development. We apply this new method in a set of machine learning experiments aimed at extracting known knowledge structures from artificial datasets from fictional and non-fictional legal settings. We show that our method allows us to analyze the rationale of black box machine learning systems by assessing which rationale elements are learned or not. Furthermore, we show that the rationale can be adjusted using tailor-made training data based on the results of the rationale evaluation.

In AI and law, systems that are designed for decision support should be explainable when pursuing justice. In order for these systems to be fair and responsible, they should make correct decisions and make them using a sound and transparent rationale. In this paper, we introduce a knowledge-driven method for model-agnostic rationale evaluation using dedicated test cases, similar to unit-testing in professional software development. We apply this new quantitative human-in-the-loop method in a machine learning experiment aimed at extracting known knowledge structures from artificial datasets from a real-life legal setting. We show that our method allows us to analyze the rationale of black box machine learning systems by assessing which rationale elements are learned or not. Furthermore, we show that the rationale can be adjusted using tailor-made training data based on the results of the rationale evaluation.

Many of the successful modern machine learning approaches can be described as ``black box'' systems; these systems perform well, but are unable to explain the reasoning behind their decisions. The emerging sub-field of Explainable Artificial Intelligence (XAI) aims to create systems that are able to explain to their users why they made a particular decision. Using artificial datasets whose internal structure is known beforehand, this study shows that the reasoning of systems that perform well is not necessarily sound. Furthermore, when multiple combined conditions define a dataset, systems can preform well on the combined problem and not learn each of the individual conditions. Instead, it often learns a confounding structure within the data that allows it to make the correct decisions. With regards to the goal of creating explainable systems, however, unsound rationales could create irrational explanations which would be problematic for the XAI movement.

Modern connectionist machine learning approaches outperform classical rule-based systems in problems such as classification tasks. The major downside of the connectionist approach, however, is the lack of an explanation for the decisions that it makes, which is a quality that the rule-based systems are known for. The new sub-field of Explainable Artificial Intelligence (XAI) therefore aims to combine the performance of the connectionist approach with the understandability of knowledge systems. Examples of such hybrid systems extract rules from neural networks, which can explain how a network comes to a particular decision. However, this is built upon the premise that the reasoning of a neural network that performs well is sound. Previous research has suggested that this is not necessarily true; a high performance does not guarantee a sound rationale. This study, therefore, aims to investigate how well machine learning systems are able to correctly internalize the underlying structure or rules of a dataset. This is accomplished with the use of artificial datasets, whose rules and structure are known beforehand. The results confirm that a high classification accuracy can be obtained without having to have learned all of the rules that define the dataset. Furthermore, the systems are able to internalize certain rules better than others. When multiple rules define the dataset, interaction effects occur that cause the systems to be less proficient in internalizing the rules.