This article appeared in Cybersecurity Law & Strategy, an ALM publication for privacy and security professionals, Chief Information Security Officers, Chief Information Officers, Chief Technology Officers, Corporate Counsel, Internet and Tech Practitioners, In-House Counsel. Visit the website to learn more.
With the corpus of law data becoming ever-more complex and nuanced, the use of machine-assisted research and analysis is becoming more of a requirement, rather than an option, in the legal profession. Because of this, some have expressed fear that robot-lawyers will replace legal professionals.
However, considering the high bar that the legal profession sets for accuracy and precision, the complexity and nuance of legal matters, and a lack of understanding of what these technologies can or cannot do, these fears begin to sound more like science fiction than science fact. The truth is that AI is already having a positive impact on the practice and business of law, making lawyers more efficient and effective at their jobs.
AI solves real challenges and answers real questions that lawyers face every day, such as: How is a particular judge likely to rule on a particular motion in a particular kind of case? Which specific parts of this 30-page contract do I actually need to review? What parts of this massive document are relevant to the particular legal issue I am researching?
Until very recently, lawyers faced with such questions have been forced to undertake tedious, time-consuming, data-intensive tasks in order to come up with satisfactory answers or solutions. AI can accomplish or facilitate these tasks more quickly, accurately and efficiently than even the most capable human experts — with the goal of augmenting their skills rather than replacing them.
Defining AI: The Perception of Intelligence
In order to dispel the fear and hype surrounding AI, it’s important to understand that AI is not a single method, process or technology. In most contexts, AI is the perception of intelligence assigned by humans to a computer application. To achieve that perception of AI, it is necessary to combine outputs from multiple technologies, such as machine learning, deep learning, natural language processing (NLP), voice and image recognition, and others, to create features or products that form “opinions” or “conclusions.”
For example, an airline website’s reservation assistant, or “chatbot,” can ask simple questions and recommend alternative flights based on price, departure time, weather or other factors. This interaction may be perceived as “intelligent,” however, it is likely nothing more than a computer program following a master script, accessing various databases and specialized subroutines, to provide responses to text-based queries.
More advanced interfaces, such as voice, may intensify the perception that we are dealing with a form of intelligence. The same is true of software that identifies and analyzes otherwise “unknowable” patterns in data and suggests a response. At the end of the day, all of these advanced technologies require a large set of training data, subject matter experts and specialized computer programmers to maintain and develop the master script and subroutines.
Advanced Tools + Relevant Data + Domain Expertise
The true test of good AI is that it is indistinguishable from a human response. To be really useful, AI requires massive amounts of relevant data, called “training” data. It also requires deep human and organizational domain expertise to distinguish between helpful and unhelpful outputs, and iteratively “train” algorithms to improve results over time. That’s why it can be misleading to think of AI as simply a single advanced technology.
In the legal world, we already have massive data sets comprised of statutes, opinions, rules, dockets, dictionaries, law review articles, contracts and much more. In addition, the amount of data generated by corporations annually is staggering and continues to grow, out-pacing the ability of even the most expert law librarians and other specialists to keep up. As such, AI is needed now more than ever to augment the capabilities of legal professionals and tame these rapidly growing volumes of data. Just as the advent of online legal research represented a massive shift in the way law was practiced, AI is poised to usher in another shift that will be at least as transformative.
Here are a few prominent examples of emerging AI-enabled solutions categories in the legal domain, in order of increasing AI complexity and technical skill:
Personalized Legal Research
In the early days of online legal research, attorneys had to rely on law librarians or specialists to construct Boolean queries using domain- and source-specific keywords to execute complex searches. Today, AI is used to find the right answers to research questions much more quickly and reduce our dependence on specialists. NLP allows lawyers to do their own research using everyday language, without having to learn a complicated set of search rules and terms. In addition, AI can “learn” what results are most relevant to the user’s query based on user feedback and search behaviors. This allows it to recommend additional research paths or source material, similar to how Amazon recommends affiliated products based on aggregated browsing and purchasing behaviors.
Paying lawyers to read through every page of a lengthy contract to understand obligations and identify potential risks and omissions can be cost-prohibitive in today’s legal environment. AI is already helping lawyers identify and classify different kinds of contract clauses, with the capacity to “learn” from large stores of similar contracts, make recommendations for revisions to new contracts, and target specific sections or clauses requiring further review. AI can also be deployed to manage or automate entire contract portfolios, searching through large numbers of documents to identify and recommend contracts requiring immediate review, updating or renewal.
Similar applications promise big benefits in reducing the formidable expense of M&A due diligence, which is data-intensive, tedious, conducted under tight deadlines and especially prone to human oversight and error.
‘Intelligent’ Interactive Interfaces
We have already begun to see simple, rules-based chatbots in the legal domain, but they are not yet equipped to address a meaningful range of topics or issues. Their more intelligent cousins, “voicebots,” use AI to convert speech to text, but so far struggle to grasp the nuances of “legalese,” which is a mix of (often truncated) Latin, English and jargon. Two examples of cases that might trip up AI are Batman v. Commissioner (189 F.2d 107 (5th Cir. 1951), cert. denied 342 U.S. 877 (1951)) or Death v. Graves (CGC-06-451316 (San Francisco Super. Ct. filed Apr. 17, 2006). To improve machine comprehension of the complex legal language, organizations with vast legal databases are training cognitive APIs to effectively distinguish legalese from ordinary language and “understand” it in specific legal contexts. This effort is akin to teaching Alexa to “think and talk like a lawyer.”
In a few years you can expect to see intelligent agents that will not only identify and comprehend legal language (including its verbal shorthand), but also ask targeted questions to refine queries and produce the most relevant answers. When lawyers can interact with search interfaces more naturally and have intuitive, two-way conversations, they can get more accurate and precise answers for their clients, requiring less time and money to do so.
Advanced Legal Analytics
AI technologies such as machine learning and NLP are helping us expand, enrich and knit together enormous data sets so they are readily searchable, can extract meaningful content, and present it in a visually engaging way for immediate consumption. For example, legal analytics tools mine millions of pages of litigation data to reveal previously inaccessible insights about judges, lawyers, parties and legal matters to inform tactical and strategic litigation decisions. They also help review large data sets from a litigation hold to find the most relevant documents to review and custodians to depose.
In addition, these tools help law firms and in-house counsel make sound business and hiring decisions based on historical data about an individual lawyer’s or law firm’s performance, expertise and background. The technology is evolving rapidly, enabling applications that will soon move beyond descriptive analysis to predict rulings and case outcomes in specific scenarios, and even prescribe or automate certain kinds of decisions — all to increase the efficiency and productivity of legal professionals.
A basic AI process involves “reading” or scanning a large corpus of data and extracting concepts and topics. This is arguably the most challenging AI problem in text processing: the ability to summarize a document rather than just pull out representative portions. Currently, computer-generated text summaries extract and compile sentences from the source material but do not provide any insights and require attorneys to proofread and sometimes rewrite them. With the help of the latest in deep learning techniques, computers will soon be able to generate well-constructed, unbiased abstracts and summaries that will give time- and budget-constrained attorneys a high-level overview of a document’s contents and context.
AI is not always the best or most cost-effective way to solve a problem, but it is already helping lawyers and non-lawyers perform complex, data-intensive work much faster and more accurately. With AI, we are leveraging legal data to train machines to develop models that project the perception of cognitive activities like learning, hearing, understanding language and predicting outcomes.
This will bring enormous benefits to lawyers and their clients, ranging from cost-efficiency to error reduction to more intuitive workflows and better legal outcomes. It will make the law more accessible and understandable by all, which will ultimately help us build a more just world.
Jeff Reihl is executive vice president and chief technology officer for the global legal business of LexisNexis. Rick McFarland is the chief data officer for LexisNexis. He is a data evangelist with a broad background in publishing, digital media, financial services, and management consulting.