... Artificial intelligence (“AI”) is rapidly changing the way lawyers work. Just as AI is overtaking the control of vehicles and the optimization of route selection to destination, AI is now applying judgment to datasets, attaching hierarchy to information based on the issues presented as well as predicting case outcomes based on the facts, legal precedent, jurisdiction, and presiding judge....
I. Introduction
Many practicing lawyers readily recall the days when their offices and hallways were occupied by bankers boxes filled with case documents. The boxes piled up when drafting briefs and preparing for depositions or trial. Each firm maintained large file rooms to warehouse the boxes and employed teams of paralegals to review, code, and index the documents.
The firm’s law library was a prized possession and the object of office tours given to clients and dignitaries. Firm attorneys spent hours in the library, reading case annotations, “Shephardizing” cases, and tracking down case reporters missing from the library shelves.
Information technology (“IT”) changed everything. The glass-walled libraries disappeared, freeing up space for conference rooms and attorney offices. The large file rooms were replaced by server rooms, housing vastly more information in radically smaller spaces. Databases replaced law books, and indexing software replaced paralegals.
Instantaneous communications and other technologies have enabled law firms to swell in size. Now, "global" firms with offices on multiple continents have become common place. Technologies, including those discussed below, have also empowered lawyers to practice effectively in smaller numbers and even solo practices to drastically reduce overhead expense and increase their personal profits. It is no longer necessary to be located in a specific geography to work with clients. And, it is no longer necessary to offer comprehensive services as practices can expand and contract through collaboration, as needed.
Despite the near-universal digitization of attorney workflows, IT’s contribution to the practice of law is arguably still in its infancy. By analogy, in the same way that mechanical and civil engineering mechanized travel but left the driver in control of the vehicle, IT empowered attorneys with efficiency enhancing tools to process and search data but left the analysis and use of the data up to attorney discretion and judgment. A problem with keyword searching, for example, is that it often retrieves large amounts irrelevant information that must be read and discarded.
Artificial intelligence (“AI”), however, is rapidly changing the way lawyers work. Just as AI is overtaking the control of vehicles and the optimization of route selection to destination, AI is now applying judgment to datasets, attaching hierarchy to information based on the issues presented as well as predicting case outcomes based on the facts, legal precedent, jurisdiction, and presiding judge. AI is also taking over the drafting of routine legal work product such as pleadings, discovery requests, and contracts, as well as the translation of foreign language documents in cross-border cases.
The legal profession is a fertile field for the continued growth and expansion of AI applications for a few keys reasons. AI utilizes “data hungry” processes to inform deep learning algorithms and analytical routines. The digitization of law libraries, statutes, regulations, corporate contracts, and broad U.S. discovery provide AI applications with a wealth of raw data for analysis. Moreover, datasets of briefing, legal opinions, legislative histories and other documents offer AI systems with extensive examples of the application of legal principles to varying factual scenarios, a type of cause and effect analysis that is conducive to AI learning. And, continued advancements in the core science of “natural language processing,” which enables software to scan, comprehend, and rationalize written text, guaranty that AI’s intrusion into the practice of law is here to stay. See, e.g., Zach Warren, Legal Tech’s Predictions for E-Discovery in 2021, Law.com (December 28, 2020).
II. Discovery
The application of IT to litigation spawned a new “e-discovery” industry. Keeping up with the rapid technological change and the expansive growth of electronic datasets was beyond the capacity of most law firms. As a result, the technical aspects of collecting, processing, and hosting data required by U.S. discovery rules were turned over to third-party vendors. These e-discovery vendors have grown in both number and sophistication, expanding not only their technological offerings but also their expertise in the minutiae of procedural rules and requirements of U.S. litigation.
A. Forensics, Collection, & Hosting
Discoverable information is no longer found only in filing cabinets. It resides in far-flung locations including cloud storage sites, company servers, personal computers, smartphones, and innumerable third-party hosted social and other media. The data is sometimes stored in legacy file formats or in partially corrupted media or files. The e-discovery vendors have developed extensive expertise and software tools for extracting, copying, and storing these disparate forms of data without disturbing their underlying “metadata” which records such things as the author and creation date of each record.
Once collected, the vendors process the data for hosting on their review platforms. The processing may include the removal of duplicate records, the conversion of imaged documents to text files, keyword indexing, and harvesting of categorical information to permit groupings by source, author, date, etc. The review platforms are multi-functional, offering “project management” features enabling teams of attorneys to coordinate their review and coding of the data as well as “hosting features” making the data available to attorney and client for keyword searching and general use.
B. Analytics
The e-discovery vendors are now introducing “predictive” features into their review platforms that endeavor to identify the documents, email strings, and other data that are most relevant to the case. In the same way that internet search engines attempt to prioritize search results based upon what is known about the user, prior third-party searches, and other information, these predictive review platforms prioritize data based on information fed into their learning protocols.
Working with their client attorneys, the vendors initially present the AI systems with exemplary “seed” sets of key documents and information about the case to initiate the AI learning process and data prioritization. Most systems further refine their knowledge of the case over time by employing “continuous learning” protocols that observe and learn from the attorneys’ key word searching and usage of the database. See, e.g., How to Make the e-Discovery Process More Efficient with Predictive Coding, Thomson Reuters.
After “learning” the key issues in the case, the platforms are then able to identify the major individuals and entities associated with those issues. The platforms provide visual mappings of communications between the key players, noting such things as concurrently discussed subject matter, frequency of the discussions, and their timing, e.g., after business hours or during the weekend. The platforms may also provide correlations between subject matter discussed among certain individuals and the creation or potential deletion of related documents and data.
These predictive analytics help ameliorate the problem of under-and-over inclusiveness of keyword searching and enable attorneys to more quickly find and study the key evidence and relationships residing in the data produced in discovery. Although not currently provided in any single platform, the work of litigation attorneys is aided by a host of other IT/AI enabled research tools that guide the legal and substantive development of the case.
III. Building the Case
A. Legal Research
The digitization of legal publications followed a development timeline similar to the progression of e-discovery services. The initial focus was upon the creation of keyword-searchable online databases of state and federal statutes, regulations, and reported cases. The next step was the refinement of search capabilities using AI to identify the most relevant or analogous legal authorities.
The publishers have added natural language search capabilities both as an alternative to keyword searching and as a means for identifying the most relevant legal authorities. Recently, some publishers have introduced AI directed features that return search results based on user-uploaded documents. For example, after the user uploads an adversary’s opening brief, the system returns potential authorities to cite in opposition, based on the review of the facts and law referenced in the uploaded brief.
B. Patent Litigation & Prior Art Searching
The defense of patent invalidity is almost always asserted in patent infringement litigation. The patent’s validity depends upon whether the claimed invention is both novel and non-obvious in view of pre-existing patents and literature, i.e., the “prior art.” With three million new patent applications filed globally and innumerable scholarly and scientific papers published each year, the task of searching this expanding body of prior art is daunting.
Adding to the sheer volume of the prior art is its complexity and ever-evolving vernacular as technological advancements give birth to new descriptive terms and jargon. Traditional keyword searching has significant limitations in this environment. Responding to these challenges, vendors and government patent offices around the globe are developing AI systems enabled with “semantic” learning and search capabilities.
Semantic learning looks at the prior art as a whole, categorizing the subject matter, noting the totality of terms and language used to describe concepts, and finding trends in the development of technology. Thus, instead of relying on specific keywords when conducting prior art searches, these systems review, for example, a set of patent claims and extract not only key words from the text but also an overall understanding of the claimed invention. That understanding is then compared with the system’s analyses of the prior art within the field of the invention to return publications and patents that are conceptually most aligned with the search target, even if the target’s claim language and vocabulary are not used in the references. See, e.g., Lea Helmers et al., Automating the Search for a Patent’s Prior Art with a Full Text Similarity Search, Plos One (March 4, 2019).
These AI systems have not yet replaced traditional keyword searching conducted by technically trained attorneys and technicians. However, the capabilities of the AI systems are rapidly improving, and the systems now are serving as helpful adjuncts and time saving tools for these professionals.
C. Transactional Analysis
Complex commercial litigation often entails tedious review and mapping of thousands of transactions in able to prove a particular scheme, course of conduct, or outcome. In the past, that work was done manually by attorneys working with economists or accountants.
The accounting profession, however, has been hard at work developing auditing software that deploys deep learning routines and data analytics to rapidly make sense of large sets of transactional data. In addition to reviewing for regulatory compliance, these systems function to identify anomalies, missing data, and potentially fraudulent transactions. The systems are assisting attorneys in more efficiently and accurately developing their positions in commercial litigation.
D. Case Strategy & Outcome Prediction
Third-party litigation funding is a growing industry in the United States. Some investors are now turning to AI to assist them in making investment decisions. These AI systems evaluate litigation funding opportunities by making outcome predictions based upon machine learning routines that are run on reported cases, news articles, and other data sets. The systems provide odds of success as well as potential damages scenarios in view of the facts known about each potential litigation investment opportunity. See, e.g., Alan Freeman, Intelligent Funding, National Magazine (July 15, 2019).
E. Jury Selection
Jury consultants have now turned to IT and AI to assist attorneys during jury selection. These systems are designed to quickly review and assess publicly available Big Data and social media relevant to potential jurors. The systems generate juror profiles, predicting personality types, world views, and likely sensitivities to the facts of the case. The databases utilized by these systems are expansive, including voter registration, criminal records, social media, campaign contributions, financial and real estate records, professional directories, online authorship and publications, as well as petition signatures. See, e.g., Voltaire Uses AI and Big Data to Help Pick Your Jury, Artificial Lawyer (April 26, 2017).
F. Judge, Opposing Counsel, & Expert Witness Analysis
The focus of AI platforms has also been trained on judges and expert witnesses. These systems, for example, not only provide information about how many times a particular judge has ruled on a particular issue, but also the odds of succeeding on the issue before the judge. Similar systems compile publicly available information about testifying expert witnesses, quickly summarizing professional and biographical information as well as testimony that has been excluded in the past. See, e.g., Caroline Hill, LexisNexis Launches Well-Trailed Judge and Expert Witness Analytics Solution Context, LegalTechnology.com (November 29, 2018).
G. Automated Pleading/Discovery Drafting
At least one legal publisher offers automated brief drafting services. The service provides a library of substantive outlines for various motions and briefs, e.g., motion to exclude expert testimony. After selecting an outline, the subscriber selects the pertinent subheadings and case citations from among the information provided in the outline. Based upon the subscriber’s selections, the system then drafts a legal analysis section in paragraph format complete with case citations that can be copied into a brief.
Other systems are offering more ambitious AI assisted drafting services. These systems, for example, prepare responsive pleadings, document requests, and interrogatories based on the system’s review of the adversary’s complaint. Similarly, the systems prepare responses and objections to document requests after reviewing the uploaded requests. The systems can also be “trained” on a broad sample of a law firm’s prior work product to enable the system to draft its written output in a customized fashion, mimicking the firm’s style and content. See, e.g., Nicole Black, These Document Assembly Tools Will Keep Your Law Firm on Track, ABA Journal (June 25, 2019).
IV. Litigation Management
A. Class Action & Bankruptcy
Class action litigation and bankruptcy matters often have large numbers of claimants and interested parties. IT and AI systems have been developed to assist litigation counsel in identifying potential claimants, managing service of required notices, automating online claim filing, sending informational updates, and paying out disbursements. These systems significantly lighten the administrative burdens of these cases, allowing the attorneys to focus on the substance of the case and introducing cost savings for the benefit of the claimants.
B. Project Management
Project management systems developed for other industries have been tailored for managing litigation teams and workflows. These systems provide calendaring tools to monitor deadlines and send reminders and notifications when work needs to commence on projects. The systems also provide budgeting tools that assist attorneys in developing budgets on a project-by-project basis using the firm’s experience in prior cases. The systems also have auditing features that show when and how the litigation billing is either on or off budget.
C. Billing/Invoice Review (In-house)
Mirroring the litigation management software used by outside counsel, in-house counsel are empowered with increasingly powerful software tools used to track and audit the performance of outside counsel. These software packages analyze firm billing information to identify potential inefficiencies in the performance of outside counsel. AI routines in the software analyze the billing and work flow trends of the company’s outside counsel and create ratings and comparisons of firm performance that in-house counsel use to guide their future hiring decisions and negotiations with outside counsel.