Artificial Intelligence and Law最新文献

How can the forensic scientist rationally justify performing a sequence of tests and analyses in a particular case? When is it worth performing a test or analysis on an item? Currently, there is a large void in logical frameworks for making rational decisions in forensic science. The aim of this paper is to fill this void by presenting a step-by-step guide on how to apply Bayesian decision theory to routine decision problems encountered by forensic scientists on performing or not performing a particular laboratory test or analysis. A decision-theoretic framework, composed of actions, states of nature, and utilities, models this problem, and an influence diagram translates its notions into a probabilistic graphical network. Within this framework, the expected value of information (EVOI) for the submission of an item to a particular test or analysis addresses the above questions. The development of a classical case example on whether to perform presumptive tests for blood before submitting the item for a DNA analysis illustrates the use of this model for source level questions in forensic biology (i.e., questions that ask whether a crime stain consisting of a particular body fluid comes from a particular person). We show how to construct an influence diagram for this example, and how sensitivity analyses lead to an optimal analytical sequence. The key idea is to show that such a Bayesian decisional approach provides a coherent framework for justifying the optimal analytical sequence for a particular case in forensic science.

This article discusses the desirability and feasibility of modeling precedents with multiple interpretations within factor-based models of precedential constraint. The main idea is that allowing multiple reasonable interpretations of cases and modeling precedential constraint as a function of what all reasonable interpretations compel may be advantageous. The article explains the potential benefits of extending the models in this way with a focus on incorporating a theory of vertical precedent in U.S. federal appellate courts. It also considers the costs of extending the models in this way, such as the significant increase in the functional size of the case base and the need to provide some kind of ordering on interpretations to select a “best” interpretation. Finally, the article suggests partially incorporating multiple interpretations of dimensions as a realistic starting point for incorporating interpretations generally, and shows how doing so can help address difficulties with dimensions. The conclusion remarks on the use of interpretations to deal with inconsistent precedents.

The structures of discourse used by legal and ordinary languages share differences that foster technical issues when applying or fine-tuning general-purpose language models for open-domain question answering on legal resources. For example, longer sentences may be preferred in European laws (i.e., Brussels I bis Regulation EU 1215/2012) to reduce potential ambiguities and improve comprehensibility, distracting a language model trained on ordinary English. In this article, we investigate some mechanisms to isolate and capture the discursive patterns of legalese in order to perform zero-shot question answering, i.e., without training on legal documents. Specifically, we use pre-trained open-domain answer retrieval systems and study what happens when changing the type of information to consider for retrieval. Indeed, by selecting only the important parts of discourse (e.g., elementary units of discourse, EDU for short, or abstract representations of meaning, AMR for short), we should be able to help the answer retriever identify the elements of interest. Hence, with this paper, we publish Q4EU, a new evaluation dataset that includes more than 70 questions and 200 answers on 6 different European norms, and study what happens to a baseline system when only EDUs or AMRs are used during information retrieval. Our results show that the versions using EDUs are overall the best, leading to state-of-the-art F1, precision, NDCG and MRR scores.

Court rulings are among the most important documents in all legal systems. This article describes a study in which natural language processing is used for the automatic characterization of Spanish judgments that deal with the physical custody (joint or individual) of minors. The model was trained to identify a set of elements: the type of custody requested by the plaintiff, the type of custody decided on by the court, and eight of the most commonly used arguments in this type of judgment. Two jurists independently annotated more than 3000 judgments, which were used to train a model based on transformers. The main difficulties encountered in this task were the complexity of the judicial language and the need to work with appellate court rulings that have a more complicated structure than decisions at first instance. For the complete court rulings, the F1 score of the inter-annotator agreement ranged from 0.60 to 0.86 and the Kappa index from 0.33 to 0.73. The F1 score of the agreement between the model and the annotators ranged from 0.66 to 0.93 and the Kappa index from 0.57 to 0.80. These results in which the model performance exceeds even the inter-annotator agreement show the high ability of transformers to identify abstract entities in legal texts.

Automated classification of legal documents has been the subject of extensive research in recent years. However, this is still a challenging task for long documents, since it is difficult for a model to identify the most relevant information for classification. In this paper, we propose a two-stage supervised learning approach for the classification of petitions, a type of legal document that requests a court order. The proposed approach is based on a word-level encoder–decoder Seq2Seq deep neural network, such as a Bidirectional Long Short-Term Memory (BiLSTM) or a Bidirectional Encoder Representations from Transformers (BERT) model, and a document-level Support Vector Machine classifier. To address the challenges posed by the lengthy legal documents, the approach introduces a human-in-the-loop approach, whose task is to localize and tag relevant segments of text in the word-level training part, which dramatically reduces the dimension of the document classifier input vector. We performed experiments to validate our approach using a real-world dataset comprised of 270 intermediate petitions, which were carefully annotated by specialists from the 15th civil unit of the State of Alagoas, Brazil. Our results revealed that both BiLSTM and BERT-Convolutional Neural Networks variants achieved an accuracy of up to 95.49%, and also outperformed baseline classifiers based on the Term Frequency–Inverse Document Frequency test vectorizer. The proposed approach is currently being utilized to automate the aforementioned justice unit, thereby increasing its efficiency in handling repetitive tasks.

Computational models of legal precedent-based reasoning developed in AI and Law are typically based on the simplifying assumption that the background set of precedent cases is consistent. Besides being unrealistic in the legal domain, this assumption is problematic for recent promising applications of these models to the development of explainable AI methods. In this paper I explore a model of legal precedent-based reasoning that, unlike existing models, does not rely on the assumption that the background set of precedent cases is consistent. The model is a generalization of the reason model of precedential constraint. I first show that the model supports an interesting deontic logic, where consistent obligations can be derived from inconsistent case bases. I then provide an explanation of this surprising result by proposing a reformulation of the model in terms of cases that support a new potential decision and cases that conflict with it. Finally, I show that the reformulation of the model allows us to verify that inconsistent case bases do not make verification that a decision is permissible substantially more complex than consistent case bases and to introduce intuitive criteria to compare different permissible decisions.

Freedom of information laws promote transparency by permitting individuals and organizations to obtain government documents. However, exemptions from disclosure are necessary to protect privacy and to permit government officials to deliberate freely. Deliberative language is often the most challenging and burdensome exemption to detect, leading to high processing costs and delays in responding to open-records requests. This paper describes a novel deliberative-language detection model trained on a new annotated training set. The deliberative-language detection model is a component of a decision-support system for open-records requests under the US Freedom of Information Act, the FOIA Assistant, that ingests documents responsive to an open-records requests, suggests passages likely to be subject to deliberative language, privacy, or other exemptions, and assists analysts in rapidly redacting suggested passages. The tool’s interface is based on extensive human-factors and usability studies with analysts and is currently in operational testing by multiple US federal agencies.

Legal Judgment Summarization (LJS) can highly summarize legal judgment documents, improving judicial work efficiency in case retrieval and other occasions. Legal judgment documents are usually lengthy; however, most existing LJS methods are directly based on general text summarization models, which cannot handle long texts effectively. Additionally, due to the complex structural characteristics of legal judgment documents, some information may be lost by applying only one single kind of summarization model. To address these issues, we propose an integrated summarization method which leverages both semantic and structural information to improve the quality of LJS. Specifically, legal judgment documents are firstly segmented into three relatively short parts according to their specific structure. We propose an extractive summarization model named BSLT and an abstractive summarization model named LPGN by adopting Lawformer as the encoder. Lawformer is a new pre-trained language model for long legal documents, which specializes in capturing long-distance dependency and modeling legal semantic features. Then, we adopt different models to summarize the corresponding part regarding its structural characteristics. Finally, the obtained summaries are integrated to generate a high-quality summary involving semantic and structural information. We conduct comparative experiments to evaluate the performance of our model. The results show that our model outperforms the baseline model LEAD-3 by 14.78% on the mean ROUGE score, which demonstrates our method is effective in LJS and is prospected to be applied to assist other tasks in legal artificial intelligence.

Automated legal reasoning and its application in smart contracts and automated decisions are increasingly attracting interest. In this context, ethical and legal concerns make it necessary for automated reasoners to justify in human-understandable terms the advice given. Logic Programming, specially Answer Set Programming, has a rich semantics and has been used to very concisely express complex knowledge. However, modelling discretionality to act and other vague concepts such as ambiguity cannot be expressed in top-down execution models based on Prolog, and in bottom-up execution models based on ASP the justifications are incomplete and/or not scalable. We propose to use s(CASP), a top-down execution model for predicate ASP, to model vague concepts following a set of patterns. We have implemented a framework, called s(LAW), to model, reason, and justify the applicable legislation and validate it by translating (and benchmarking) a representative use case, the criteria for the admission of students in the “Comunidad de Madrid”.