Artificial Intelligence And The Legal Profession

Artificial Intelligence And The Legal Profession Technologies that may be able to automate typical legal duties, such as performing case research or creating standard contracts, have existed now for quite some time but have not been fully integrated into practice as of yet. However, in roughly the last decade, pressure on lawyers and law firms to reduce fees has increased and this has led to a more favorable attitude toward legal technologies in a push to increase efficiency throughout the profession. Second, pressure to improve access to justice by reducing both financial and structural barriers affecting disadvantaged groups has led to the development of several online or otherwise accessible legal technologies. Third, as the capabilities of computers grow to include higher-level processes, the possibilities for their integration into the legal field grow too. While previous legal technologies threatened to replace mostly clerical work, artificial intelligence may threaten to replace lawyers themselves. To gain a comprehensive un derstanding of this topic, it is necessary to explore the current state of artificial intelligence in the legal profession, how it impacts the demand for lawyer labor, and how the profession’s regulatory structures relate to the trajectory of legal technology. The term â€Å"artificial intelligence† describes how computers can perform tasks that might generally be thought to require some level of human intelligence. These tasks can range from flagging outliers in a set of data to transcribing an audio tape and everything in between. Fundamentally, computers operate based on sets of defined rules. Any task to be performed by computers must be able to be articulated as a set of basic rules to be followed. Deductive rules are those laid out in a step-by-step process that is followed by the computer until the task is completed. An example of this as applied to automated legal work is the use of automated document assembly, such as the creation of a template for a will or other standard legal document. In a matter of seconds, a document assembly application can pull relevant information about a client and use this to create a personalized document. Similarly, a computer could provide a list of cases from a particular court citing a parti cular statute. In addition to such tasks whose processes can be modeled explicitly, some other tasks can be accomplished with the use of data-driven rules. The relationship between a set of input variables and the resulting outputs can be estimated by a process called â€Å"supervised machine learning†, so called because the estimation is bounded by the set of training data. For example, a team of researchers was able to develop a model to predict the behavior of the Supreme Court, based only on data from past decisions. They were able to achieve 70.2% prediction accuracy at the case outcome level and 71.9% prediction accuracy at the justice vote level. While these predictions cannot be expressed based on a combination of deductive rules as in previous examples, with enough input data a relatively consistent pattern can be recognized. Similarly, software for automated document review has been developed and proven successful at determining relevancy of documents based on the i nput of a â€Å"seed set† of documents designated relevant or not. As well as being able to potentially replace or improve the efficiency of routine legal tasks, predictive algorithms like these have possible applications to the legal field as a whole. For example, results of a race-neutral sentencing predictor algorithm could be compared to actual sentences to determine the influence of human bias in such processes. Overall, the success of data-driven algorithms is significant primarily in that it highlights the routineness of some tasks that would otherwise be viewed as more sophisticated and complex. The adoption of technology into the legal workplace will be influenced by the market in addition to the quality and capabilities of available technologies. Historically, the demand for technology in law firms has been low for several reasons. The billable hour system economically encouraged inefficiency, while the typical partnership structure meant the funds for new technologies would come directly out of the pockets of partners, unlike a traditional corporation structure in which the money would come from shareholders. To the first point, in recent years there has been a notable shift in supply of lawyers relative to demand for lawyer labor. This leads to increasing pressure to improve efficiency and reduce cost to clients. Additionally, a rise in the share of legal work performed by a company’s house legal department compared to that performed by outside law firms allows these technologies to be purchased with corporation funds, which is more favorable. In addition to the g rowing demand for legal technologies, the capabilities of such technologies are expanding rapidly as well. The theory of disruptive innovation explains how this will also contribute to the growing prevalence of legal technology. When the only legal tasks that could be automated were clerical and other low-level services, law firms were willing to buy into such software to improve efficiency and better serve their clients. However, developers have since been able to expand their technologies to handle more complex tasks, and firms now are practically required to adopt them as client pressures grow. This phenomenon, in addition to the growing market interest, will lead to rapid adoption of technology into the legal realm. As the use of technology in the legal profession grows, the impact of artificial intelligence on the demand for lawyer labor in some areas has been or will be relatively significant, while in others is unlikely to pose a significant threat. The distinction between tasks that can easily be automated versus those that cannot is in the extent to which their underlying structures can be defined. For example, while document drafting can be successfully automated as discussed above, more complex legal writing characterizing the state of the law or its application to particular factual circumstances presents a much more challenging situation. The conceptual creativity and flexibility demanded by this type of writing cannot be defined by either deductive or data-driven rules. Another example is the distinction between document review during discovery and document review during due diligence. While the former can be automated with the use of explicit rules, the latter requires critical judgme nts that a computer cannot make. A trained lawyer can note, for example, any unexpected information or violations of appropriate rules that a computer would not be able to recognize without being explicitly primed to look for such things. Some advanced applications of artificial intelligence to the legal profession have found ways to extend its reach despite these limitations. For example, IBM’s Debater System is able to analyze documents and other materials annotated first by humans. While this is clearly less efficient than purely automated processes since it requires time put in by an associate, it alleviates some of the major issues with automated lawyering; any glaring contradictions or relevant subtleties can be highlighted before the materials are analyzed by the software. Another way by which artificial intelligence can be employed to perform tasks that a lawyer is qualified to do is through online systems to resolve minor disputes ranging from parking violations to e -commerce complaints. These technologies aid lawyers in negotiating by analyzing overlap between stated preferences of the two parties and can typically reach a mutual solution without the involvement of a lawyer at all. While such systems may thus be able to replace lawyers and even judges entirely, they will likely have little impact on the overall demand for lawyer labor since it would likely not be feasible economically or otherwise to hire an attorney and litigate. In this way, the full automation of legal services comes at no cost to lawyer labor. In fact, a study that categorized legal tasks by the impact of automation on employment found that only around 4% of lawyers’ time was billed to tasks most acutely threatened by artificial intelligence. In summary, while even moderately complex legal tasks have been successfully automated, the legal profession is unlikely to find itself obsolete within a decade as some headlines predict. As new technologies continue to develop and make their way into the practice of law, there emerges a need for a better way to protect the integrity of the legal system while ensuring consumer protection and access to quality services for all members of the population. With regard to consumer protection, computers offer the advantage of eliminating human error and standardizing services in some cases, but certainly not all. For example, online services cannot effectively analyze highly complex scenarios, but instead of returning an error message often return products completed in a way that places the client liable. While consumer protection concerns are not necessarily graver with automated legal services, they deserve at least the same attention afforded to legal services provided by human lawyers. Current professionalism guidelines limit the performance of legal services to those trained and licensed to practice law, and the stated reason for this is â€Å"to protect the public fr om the consequences of receiving legal services from unqualified persons†. These regulations are then enforced through disciplinary sanctions imposed by bar committees. However, these guidelines have several weaknesses when it comes to the regulation of new technologies in the legal field. They fail to specifically outline what tasks require the expertise of a licensed practitioner, which makes them unhelpful in governing what tasks may be left to automated providers. Next, even though computers may not be skilled enough to perform some tasks normally handled by lawyers, they may be competent enough to assist trained professionals, something not addressed by the guidelines. Finally, there has never been a sophisticated investigation into exactly what tasks computers can perform at the same level lawyers can. Although the quality of legal services provided by automated programs may be lower than that provided by a trained and experienced lawyer, the low prices associated with t hese types of services justify their employment in certain instances. For example, someone who needs a simple will written has little need for a full-service lawyer. However, the consequences of trading low quality for low cost are magnified in more complex high-stakes issues like custody disputes or messy divorce negotiations. The â€Å"access to justice† problem discussed briefly above is a key aspect of the legal profession and should not be redefined as â€Å"access to some form of legal services whether quality or not†. While the introduction of low-cost automated services does increase accessibility to low-income persons, the result could be a two-tiered system that does not equally serve justice to the disadvantaged. For these reasons, it is important that the use of technology in the legal profession is regulated in the future to prevent it from being driven forward on the basis of outcome alone. When it comes to the recent discussion surrounding technology in the legal profession, there are two major sides taken. Some argue that incorporating more technology into the field will reduce costs to client with little expense in terms of quality and oppose strict regulations while others argue that there is no equal alternative to the work of a trained professional. This paper has shown the principles of artificial intelligence that govern which tasks can be automated successfully and discussed the impacts of such automation on the legal profession. While the media forecasts an end to the legal profession, in reality only a small portion of legal tasks are affected and the overall effect on the demand for lawyer labor is moderate at best. With that said, current regulations are insufficient at providing a consistent framework to guide the incorporation of artificial intelligence and other emerging technologies into the field, and the careful creation of such regulations will prove necessary as automated legal services develop further.