AI in legal is the new normal. AI and the legal profession are inextricably intertwined. Many of the applications disrupting the legal industry run on artificial intelligence. But what exactly is artificial intelligence? Before we go into a deep dive of how AI is affecting the practice of law, let’s go into a quick primer of artificial intelligence for the uninitiated.
What Is Artificial Intelligence?
The first thing you need to understand is that artificial intelligence has no universally agreed-upon meaning. Its definition depends on whom you ask. One of the most useful working definitions of artificial intelligence in the context of AI and legal profession is the use of technology to solve problems, automate decisions or predictions for tasks that typically require intelligence when humans do them. Good examples include playing chess, translating languages, driving cars, or using AI in legal to predict a judge’s decisions. These tasks all require high cognitive processes such as spatial reasoning, thinking, and abstract conceptualization. So, if a computer can solve tasks requiring such cognitive abilities, it is said to possess artificial intelligence. The crucial point is that computers solve these tasks using a very different approach from humans. This is important because when laypeople think of artificial intelligence, they think of computers that can think like humans. Hollywood takes much of the blame for that, but this kind of “strong AI” is still far on the horizon from becoming a reality. Today, the most advanced AI algorithms cannot think or replicate human high-order cognitive skills such as abstract reasoning. A 3-year-old human has much more advanced cognitive skills than the most sophisticated AI application in use today. So, words used to market applications describing them as “super-intelligent,” “cognitive AI,” etc. are an exaggeration. Therefore, the AI currently in use is a bit of a misnomer because there is no development remotely close to coming to the level of natural human intelligence.
So, if AI is not what you thought it to be, then what is it?
AI in legal is principally pattern-based artificial intelligence. Applications make use of algorithms to scan large amounts of data for patterns. This is a powerful tool that is used to automate tasks such as driving cars and language translation. There also exists rule-based AI though it isn’t as dominant as pattern-based AI.
There are three primary artificial intelligence techniques:
- Rules, Logic, and Knowledge Representation Approach
This is part of a larger category known as Knowledge Representation. The goal is to model real-world processes and systems in a format a computer can use to process data. Essentially, software engineers and subject matter experts observe tasks in the real world and then write code that breaks down those tasks into definite steps and rules that a computer can follow to accomplish a task. A good example of this kind of AI in the legal profession is tax software such as Turbotax that has converted the income tax code into logical computer rules to compute your tax liability.
This approach allows computers to use deductive reasoning to perform tasks much faster and with greater accuracy than human beings. The early history of AI involved knowledge representation where expert knowledge is codified into applications.
- Machine Learning Approach
This approach gained popularity at the turn of the century and is credited with legal tech AI’s takeoff as we know it today. In machine learning (ML), algorithms examine large data sets, detect patterns, and then harness these patterns to accomplish many useful things. For example, self-driving cars mostly use machine learning, as do Amazon recommendations, email spam notifications, and document analysis in the legal profession. Do not think of machine learning as one thing but a series of approaches and techniques such as neural networks, deep learning, predictive analytics, logistics regression, and more. The following are the main characteristics to consider when determining if an application uses machine learning:
- Learning: ML algorithms are designed to learn and improve their performance over time. This is achieved by “showing” the application good and bad examples. Therefore, the reality is that learning is not the kind expected from a human but rather a metaphor to describe the application getting better at a task over time.
- Pattern detection: At a basic level, ML is sophisticated pattern detection capabilities. These patterns are encoded into a computer model.
- Data: Machine learning requires data and often a large amount to detect patterns that then automate decisions. This often presents a problem for artificial intelligence in the legal sector because data isn’t always available in a highly accessible format.
- Self-Programming: In ML, the application takes good and bad examples and figures out what activities lead to desired behaviors, and then writes a code to follow in the future.
- Hybrid Approach
Many successful artificial intelligence legal profession systems use both approaches. For example, self-driving cars use a mix of these approaches. Hybrid systems also include systems that work hand in hand with human intelligence. Such applications support humans to make better decisions and are common examples of AI in legal sectors. In reality, there are few fully autonomous AI applications. Most applications have humans in the loop.
History of AI in Legal
Early Thinkers (17th and 18th Century)
AI in legal goes as far back as Gottfried Wilhelm Leibniz, who made significant contributions in mathematics and jurisprudence in the 17th and 18th centuries. He expressed the idea that society could use math or algorithms to make the law more predictable.
At the heart of AI in legal is utilizing computers and mathematical formulas to make the law more understandable, accessible, useful, predictable, and manageable.
Lawyers have made critical mathematical contributions. For example, linear algebra, which is the mathematical backbone of most machine learning, originated partly from lawyers. Two of the founding fathers of linear algebra, Arthur Cayley and James Joseph Sylvester, were both lawyers. Many of the most famous mathematicians were lawyers as well. Leibniz, Cayley, and Sylvester are just a few examples. There is a long tradition of lawyers influencing math in the interconnectivity between law and technology.
Pre-AI Modern Era (the 1940s)
In the 1940s, Lee Loevinger, an American jurist and lawyer, came up with the term jurimetrics, which he described as using probability and statistics to answer legal questions in a 1949 Minnesota Law Review article.
The Era of Knowledge Representation (the 1960s – 1980s)
There were many early pioneers from the 1960s to the 1980s. At the time, it was a rules-based and knowledge representation-based system. Many of these early computer scientists started applying these ideas of representing law rules. In this era, there were many attempts to model legal arguments in ways that computers could understand formally. They attempted to model legislation or regulation in ways that computers can understand. Several conferences dedicated to AI in legal, such as IAAIL and Jurix, were conceptualized and formed.
Interestingly, most of the AI in legal research occurred in Europe and not in the United States. Italy, The Netherlands, England, and Germany have historically been trailblazers and remain important centers even today.
Legal Tech and Machine Learning Era (2000 to the present)
Major AI in legal developments in this era include Stanford Codex’s founding in 2006 and the recent proliferation of legal tech startups. Since 2013, there has also been much interest by venture capitalists in the legal technology industry. Many startups have successfully raised millions of dollars in seed money after demonstrating proof-of-concept.
AI in Legal Today
For some years now, many lawyers and law firms have known about AI and are preparing for artificial intelligence in the legal profession. There are three categories of artificial legal intelligence in use today.
- Law Administrators
These are lawmakers and enforcers. They include legislators, judges, police, other law enforcement, and government regulators and officials. Let’s consider some examples of artificial intelligence and the legal profession in the instance of law administrators.
- Judges: Artificial intelligence is now being used in bail and sentencing. Judges in some jurisdictions get AI reports that predict the possibility of a criminal defendant absconding. These reports do not bind them, but they still have a persuasive effect.
- Government officials: Officials in various government departments use AI to decide on how to distribute limited government benefits.
- Police: The police are now using predictive analytics to predict crime upsurges in particular neighborhoods. Facial recognition technology has also been deployed in many cities worldwide and has been instrumental in mitigating terrorist attacks.
- Law Practitioners
This category includes lawyers and in-house counsel. Artificial intelligence and the legal profession are joined at the hip. Lawyers today use artificial intelligence in legal operations, including:
- Artificial intelligence and legal analytics and venue analysis.
- Automated document analysis and diligence.
- Case management, docketing, and workflow systems.
- Case outcome prediction.
- Document creation and assembly.
- Full practice management.
- Legal research.
- Litigation eDiscovery and technology-assisted review.
- Natural language processing of legal documents.
- Smart contracts.
- Legal Service Consumers
These are ordinary individuals and businesses that consume legal services and comply with the law. Examples of artificial intelligence by legal service consumers are as follows:
- Automated dispute resolution services.
- Automated legal document analysis.
- Automated legal document assembly.
- Legal chatbots.
- Legal expert systems such as tax software.
- Regulatory compliance engines for businesses.
Legal Technology Artificial Intelligence and the Future of Law Practice
While artificial intelligence and legal profession applications have come a long way, limits on artificial intelligence that are not likely to be surmounted any time soon remain. Some of these limitations include:
- Artificial intelligence systems need data. They can’t think on their own and must be fed with expert data. Gathering this data can be very expensive.
- They also require patterns or rules. If a data set lacks patterns or the system isn’t given a set of rules to guide it, it’s useless.
- We are yet to approach 100% accuracy. Just look at the translation and transcription services provided by Google, Amazon, IBM, and other services. While pretty good, they are only 90-95% accurate. If you had a billion-dollar merger agreement that you needed to be translated into a foreign language, you probably wouldn’t use any of these services because a slight error could end up very costly.
To understand the future, one must consider artificial intelligence in legal sector policy issues. Key among them are:
- Automation of legal jobs: Many tasks traditionally done by lawyers are being automated by artificial intelligence. Certain types of legal jobs that focus on routine and mechanical tasks will be automated in the future. However, lawyers do a lot of stuff that AI cannot do. There isn’t likely to be any AI for abstract and conceptual tasks such as; legal problem solving, advocacy, client counseling, human emotional intelligence, policy analysis, big picture strategy, and creative thinking.
- Creation of new legal jobs: While new technology automates existing jobs, it also creates new jobs. For example, the invention of the personal computer led to the creation of a whole new category of computing careers. Similarly, AI in legal leads to new jobs such as legal data analysts, software engineers, and product managers.
- Ethics: AI in legal is also leading to ethical policy issues. For example, France recently banned the publishing of predictive analytics, showing how a judge is likely to rule based on when the facts are similar. The authorities there cited the need to prevent litigants from influencing which courts their matters are heard in. There is also the issue of biased crime data sets that can wrongly influence judicial or law enforcement officers. This illusion of mechanical objectivity can be very dangerous when AI systems begin to be regarded as infallible.
- Privacy: AI in legal also has implications for data privacy. Private data is information that a user has chosen not to reveal publicly. But AI can infer non-disclosed personal data from public data. For example, an AI system can look at a set of public data (tweets, credit score, location, friend networks) about a person and accurately infer their social status, political orientation, sexual orientation, religion, and many other facts about individuals that they haven’t publicly disclosed.
AI in Legal Conclusion
It is important to have a realistic view of AI in legal. AI is neither the magic bullet that will improve access to justice nor the villain that will render lawyers jobless. As with all technology, there will be some winners and losers, but society as a whole will enjoy a net gain.