Artificial intelligence (AI) has demonstrated huge potential in improving legal systems to make them more accessible, efficient, and equitable. The use of AI-based solutions is reshaping the legal profession. However, many obstacles and ethical concerns remain regarding the application of AI in legal systems. This article delves into the role of AI in legal systems, specifically the opportunities and challenges.
Image Credit: witsarut sakorn/Shutterstock
AI in Legal Systems: An Overview
Technological revolutions often have a minimal impact on the legal systems due to the legal profession's entrenched resistance to change. Technology rarely offers significant support for legal solutions like the development of judicial judgments and litigation programs, which increases the challenges to replace legal practitioners.
However, AI technology is anticipated to change this scenario in the legal domain and transform the functioning of legal practitioners and the way individuals access the justice system. It can effectively handle tasks like implementing and organizing regulations and information gathering.
AI can also simulate the legal reasoning process by developing a theoretically comprehensive computational model that generates answers, along with the corresponding rationale, for the presented legal issues. AI-driven research tools rapidly analyze extensive legal databases and provide innovative insights to attorneys.
Additionally, virtual assistants and chatbots provide legal guidance to individuals, which improves legal information accessibility. Money and time are saved by AI when document automation tools are used to produce legal documents.
Opportunities of AI in Legal Systems
E-discovery, Legal Research, and Document Automation: Natural language processing (NLP)-based tools can be utilized to retrieve relevant regulations, statutes, and case law. For instance, NLP tools like ChatGPT can answer simple legal questions and can also be used as a legal search engine to quickly search case law and legal provisions.
Similarly, virtual assistants with NLP capabilities, such as Genie AI, can automate time-consuming and repetitive operations like generating legal documents, contract analysis, and document review.
Predictive Legal Analysis and Legal Review: Predictive analysis is one of the major areas of opportunity for generative AI in legal work. Examining past data, AI algorithms can predict prospective threats, litigation trends, and case outcomes. For instance, Lex Machina is employed for litigation strategy as it can gain insights into the potential outcomes of a judge's decision by leveraging the court docket database.
Similarly, Law Notion utilizes generative and extractive AI to analyze past success rates of similar cases automatically to evaluate potential case outcomes. It also uses analytics results to support evidence-based arguments. Moreover, AI solutions excel in legal reviews; finding pertinent data, discrepancies, and patterns in legal documents; and in reading and summarizing documents.
Case Management, Legal Advice and Expertise Automation, and Information and Marketing: AI tools excel as personal assistants for timekeeping and billing time, scheduling client meetings, and calendar management, and for routine work like filing and sorting out files for case management.
Smart virtual assistants can also provide basic legal information to clients, handle routine client inquiries, and provide client support and communication. Additionally, AI tools are suitable as marketing tools and for virtual research assistants.
Current State of Legal AI
Legal Judgment Prediction (LJP): LJP is one of the crucial tasks in legal AI, specifically in the civil law system. In this system, the judgment results are decided based on the statutory articles and the facts. A study recently performed several experiments on C-LJP, which is a large-scale Chinese criminal judgment prediction dataset, using multiple classical text classification models, including DPCNN, TextCNN, BERT, and LSTM, and models specifically designed for LJP, including Gating Network, TopJudge, and FactLaw.
Pretrained parameters on Chinese criminal cases were used for the BERT parameters. Results demonstrated that the TopJudge model had the best performance among all evaluated models.
Similar Case Matching (SCM): In countries with the Common Law system, like India, Canada, and the United States, judicial decisions are made based on representative and similar cases in the past. Thus, identifying the most similar case remains the primary concern in the Common Law system judgment.
Thus, SCM has become an important topic of legal AI to better predict the judgment results within the Common Law system. A study was performed on the CM dataset to thoroughly understand the current advances in legal information retrieval using four types of baselines, including term matching method TF-IDF; Siamese network with two parameter-shared encoders, including TextCNN, BiDAF, and BERT; semantic matching models in document level/SMASH-RNN and sentence level/ABCNN.
The dataset CM contains 8,964 triples, with each triple containing three legal documents. Results displayed that existing neural models capable of capturing semantic information outperformed TF-IDF. However, the performance was still not sufficient for SCM.
Legal Question-Answering (LQA): The objective of LQA is to answer questions in the legal domain. One of the important tasks of legal professionals is to provide high-quality and reliable legal consulting services to non-professionals. A study was conducted on the JEC-QA dataset using different representative question-answering models, including HAF, Co-matching, BERT, and BiDAF to determine the progress of LQA.
The JEC-QA was selected as it was the largest dataset obtained from the bar exam, which ensured its difficulty. This dataset contained 28,641 multiple-answer and multiple-choice questions, coupled with 79,433 relevant articles to help answer the questions. Additionally, it classifies questions into case-analysis questions (CA-Questions) and knowledge-driven questions (KD-Questions) and reports the humans' performances.
Results showed that models cannot effectively answer legal questions when compared with their promising results in open-domain question answering. Thus, a huge gap still exists between humans and current models in LQA.
Navigating Challenges
Despite the significant opportunities for using AI in legal systems, the implementation of AI in legal systems also poses several challenges, including the possibility of large-scale replacement of human labor, enormous expense during the implementation phase, technical drawbacks, and ethical concerns related to the legal system's integrity, protection of client's rights, and the duties of lawyers.
For instance, AI is challenged while making choices in legal problems where the law is non-existent or ambiguous, which is a big problem as AI typically learns from the existing data. Probabilistic or heuristic approaches can be utilized by AI systems, which can lead to ambiguous or erroneous conclusions.
This difficulty can be addressed by developing AI systems that create their data through the trial and error process. Additionally, human input like moral considerations or expert knowledge can be incorporated into the AI decision-making process to overcome this challenge.
Accuracy and accountability are the biggest challenges of AI in legal systems. AI systems providing incorrect information/making errors/generating legal conclusions or interpretations deviating from established legal norms can result in severe consequences in legal matters. This leads to concerns related to accountability of decisions when AI tools have been used in making those legal decisions.
Determining the responsibility in these situations can be challenging as it involves apportioning the liability between the defective AI solution developer and the law firm using that defective solution. Lawyers are duty-bound to provide clear information and competent representation to their clients.
An awareness of the risks and benefits of AI technology is essential for this duty of competent representation. Additionally, lawyers must communicate with their clients by promptly informing them of any circumstance or decision requiring the informed consent of clients.
Fairness and bias are the other big challenges of AI in legal systems. Biases in the training data can be inadvertently perpetuated by AI algorithms, which can result in biased outcomes. This will lead to unjust results or unequal treatment, violating equal protection under the law and principles of fairness.
AI typically requires access to confidential legal documents and data, which leads to concerns related to privacy and data protection. Several privacy protection laws, like the European Union General Data Protection Regulation, have mandated lawyers to maintain client information confidentiality. Lawyers are also not allowed to represent clients during conflict-of-interest situations.
Establishing a client's trust is extremely difficult in the absence of human empathy and insight and contextual understanding due to the dependence on AI systems. Finally, extensive AI adoption in legal practice could lead to large-scale job displacement. Junior lawyers will not be able to acquire the necessary legal skills when mundane legal work is outsourced to AI assistants.
Overall, while AI offers legal system improvements in accessibility and efficiency, concerns linger regarding accuracy, accountability, and potential biases.
References and Further Reading
Burt, J. A. (2021). The Revolutionary Impact of Artificial Intelligence on the Future of the Legal Profession. Kutafin Law Review, 8(3), 390-402. https://doi.org/10.17803/2313-5395.2021.3.17.390-402
Zhong, H., Xiao, C., Tu, C., Zhang, T., Liu, Z., Sun, M. (2020). How Does NLP Benefit Legal System: A Summary of Legal Artificial Intelligence. ArXiv. https://doi.org/10.48550/arXiv.2004.12158
Marwala, T. (2024). AI And The Law – Navigating The Future Together. [Online] Available at https://unu.edu/article/ai-and-law-navigating-future-together (Accessed on 13 May 2024)
Farrukh, T., Qureshi, F. N., Abbasi, S. (2024). Artificial Intelligence (AI) in Legal System. Journal of Independent Studies and Research Computing, 22(1), 25-32. https://doi.org/10.31645/JISRC.24.22.1.2
Pietropaoli, I. (2023) Use of Artificial Intelligence in Legal Practice. [Online] Available at https://www.biicl.org/blog/69/use-of-artificial-intelligence-in-legal-practice (Accessed on 13 May 2024)