Artificial Intelligence: Trends in The Legal Industry

Jul 21, 2022
  • Author(s) : Dinesh Pednekar
  • Introduction
    Several high-end technologies like AI-powered research tools, document review, conversational humanoids, self-drive cars, AI voice assistants, and sophisticated robots are a result of high-end machine learning or Artificial Narrow Intelligence (“ANI”). This is distinct from artificial general intelligence (“AGI”)[1] or super-intelligent machines. There exists ample discourse to suggest that reliable AGI might emerge only in and around 2050[2]. Even without AGI, some apprehensions surrounding artificial intelligence include misuse/incorrect use of technology, loss of jobs, intrusions to privacy, algorithmic fallacies, ethical breaches and so on. The present trends can help us re-check and recalibrate our position with respect to responding to such technologies and addressing the concerns around AI before incorporating the same in the legal framework.

    One subset of AI i.e Natural Language Processing (“NLP”) is quickly making inroads and some experts even believe that the field of law is the best place to test and develop NLP[3]. NLP is a technology which processes tangible human language/communication either in text or voice. At the cost of oversimplification, NLP is the processing of human language and harnessing the ability of machines to understand different components of a sentence, conduct sentiment analysis, understand relationship between words, the logic behind such sentences and finally connect all the dots together to act as a useful information repository with minimal human effort/interference.

    However, NLP is also far from perfect. That is because the Human language is complex. It is filled with sarcasm, local lingos, fluid grammar, context, overspill from other languages and so on. To make things even more difficult, there are many languages and dialects within those languages. Even English language is understood differently in different countries. In this context, the reason why legal industry and NLP are a best fit together is because in legal parlance, ‘words’ as computational entities often have one and determinate meaning. For instance, nobody would be sarcastic while drawing up a contract. For algorithmic designers, legal idiolect brings much more order to the chaotic human language. Holding this thread, we move forward to analyze the developments in this area of tech and explore convergences between law and NLP.

    Legal research and review tools: Understanding the Context
    There are several legal softwares today which assist in legal research (word & concept based) as well as in reviewing documents or summarizing data. However such softwares are only assistive tools which help in the process of legal research with the help of refined search engines, conceptual directions, extracting information and summarizing huge chunks of data.

    An exciting development in review and research arena was the release of Google BERT in 2019 which was one major breakthrough in rather rare tweak-able product of Google i.e its search engine. With BERT the search engine now has the ability to understand the context of words rather than just the sequence-searching of words in a query. The AI research tools we have today can configure searches based on concepts and not just words which is a step-up in manual research assistance. However, presently only a fraction of activities can be automated in legal research/review of documents as these require a good grasp on the factual matrix at hand which is far from a simple question-answer pattern.

    NLP in speech: Negotiations & Arguments
    Some examples of NLPs are voice assistants, translation softwares, text-to-speech and dictation tools. To perceive its relevance in law, we need to understand the limitations of what it can do. A relevant example would be IBM’s Project Debator which is the first AI system that can debate with humans on complex topics which it was never trained on. When given a topic, within a few minutes, it scans through millions of documents and short pieces of texts taken from newspaper, journals etc and formulates text segments for opening speech and arguments. It filters through redundant text, selects the strongest remaining claims and evidence, and arranges these by themes to create a narrative[4]. This covers most of the activities undertaken by a lawyer in conducting legal research, understanding procedural aspects, grasping facts, preparing pleadings and presenting the same before a court of law. Project Debater lost a debate competition against an experienced human debater however it must be recognized that such a technology exists and it is getting better by the day[5]. The scenario could be very different where structured language is involved within a rubric or framework of well-defined rules (laws). Machines perform better in standardized environments and legal industry provides that.

    AI judges: Judicial decision making and assistive technologies
    Many visualize “robot judges” or “AI judges” as assistive mechanism for judicial decision making. Realistically speaking, replacing judges altogether is a debate which does not belong to this decade at least. In some countries the algorithms are already helping judges to determine bail-related matters[6] or cases of highly technical nature. AI assistive technologies for dealing with administrative matters like determining service, case management, assigning hearing dates etc are relevant. However, there are many non-quantifiable values in judicial decision making like the human element of ‘good faith’ and ‘mercy[7] which are not comprehensible by machines, yet. The present day algorithms would give a more mechanical decision if left to it. Hence in highly technical environments, AI can be of huge assistance in deciding strictly legal questions (not being dependent on factual variables) by providing quick legal solutions or determining position of law. AI can be used to determine issues like limitation, jurisdiction, settled point of law etc.

    – Potential Threats
    However, such technologies come with added hassle of cyber-security. A simple error in feeding data could have a colossal impact. One single error in algorithmic design, implementation or processing could make or break the case. Though many may argue that human discretion is not perfect, the trust deficit against machines is much more compared to human decision making. Apart from this, law has always emphasized on ‘speaking orders’ and unless there are ancillary developments in explainable-AI (“XAI”), the idea of AI judges will remain illusionary.

    Litigation & AI: An all-rounder technology
    Courts in India are heavily burdened with pending cases and at 21.03 judges per million people, its judge to population ratio is scathingly awakening[8]. Apart from this the Supreme Court of India has always embraced the idea of Artificial Intelligence to address the pendency of cases while clarifying that AI is here to stay but it will not “spill over to decision making. Former Chief Justice of India, Justice S. A Bobde[9] has reportedly stated “As I said, it fully retains the autonomy and the discretion of the Judge in deciding the case, though at a much faster at which the readiness of information is made available by AI.

    There already exist so many technologies which could act as assistive tools for lawyers and judges in litigation. For instance, the Stanford Question Answering Dataset (“SQuAD”) can comprehend data and answer complex questions just like humans[10]. This could help in identifying specific components from voluminous briefs. There are AI text summarizers[11] which claim to understand context and resultantly summarize huge sets of information/data. This could indeed help in briefing colleagues and kick-starting a research. Similarly, ROSS is an NLP based technology which helps to chalk out the most appropriate case laws based on facts[12]. Predictive analysis softwares could assist in settlement of disputes. Routine tasks could be automated. Softwares like Public Safety Assessment (PSA) which is already in use by courts in the US identifies if an Accused is a flight risk and/or Accused is likely to commit a crime. In any case, there is not a single machine or one-stop-shop solution to all the activities of a lawyer and we are yet far from it.

    Conclusion
    Machines are increasingly becoming an integral part of the legal industry. The legal fraternity need to realize that more investments are needed in this sector. The fear of being obsolete will make lawyers integrate AI into their practices. Embracing technology may well be the key to ensuring quality outputs at short timelines, something which the profession often demands. As abovementioned, there are many technologies at disposal of lawyers for due diligence, document review, structuring contracts etc. With such technologies around, developments in NLP and ever-increasing standardization of law and policy on a daily basis, the future of law and tech is in and around NLP. Both the consequent industries have ample potential to work together and draw a framework for development of NLP which would benefit all. Evidently, there is a vacuum in the legal industry which can be plugged with high-end technologies so that lawyers can dedicate more time to substantial tasks and/or contribute to policy developments.

    Disclaimer: The information provided in this update is intended for informational purposes only and does not constitute legal opinion or advice. Readers are requested to seek formal legal advice prior to acting upon any of the information provided herein. This update is not intended to address the circumstances of any particular individual or corporate body. There can be no assurance that the judicial/quasi-judicial authorities may not take a position contrary to the views mentioned herein.

    [1] Bernd Carsten Stahl, Ethical Issues of AI, 8 March 2021, https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7968615/
    [2] Cem Dilmegani, When will singularity happen? 995 experts’ opinion, https://research.aimultiple.com/artificial-general-intelligence-singularity-timing/
    [3] https://www.forbes.com/sites/marksears1/2019/05/14/ai-challenges-and-why-legal-is-a-great-place-to-kick-start-great-nlp/?sh=663901894408
    [4] https://research.ibm.com/interactive/project-debater/about/
    [5] https://www.vox.com/future-perfect/2019/2/12/18222392/artificial-intelligence-debate-ibm-san-francisco
    [6] https://perma.cc/WZW9-3DF2
    [7] Richard M. Re & Alicia Solow-Niederman, 22 STAN. TECH. L. REV. 242 (2019), Developing Artificially Intelligent Justice
    [8] https://economictimes.indiatimes.com/news/india/judge-population-ratio-stood-at-21-03-judges-per-million-people-in-2020-law-minister/articleshow/85072861.cms
    [9] https://analyticsindiamag.com/behind-supace-the-ai-portal-of-the-supreme-court-of-india/
    [10] https://rajpurkar.github.io/SQuAD-explorer/
    [11] https://sassbook.com/ai-summarizer
    [12] https://blog.rossintelligence.com/post/enough