Similarly, law firms cannot understand NLP`s limitations. Relying entirely on these flawed tools can lead to critical omissions or errors. In the legal field, these errors can result in fines or imprisonment. Research shows that even deep learning models struggle to classify legal texts correctly, and the ethical implications of these tools are significant. These barriers are hard to ignore given the high stakes in the legal industry. NLP can do simpler legal work, but it is not enough to conduct extensive research and legal interpretation. Legal advisors are interactive systems that create advice tailored to the user`s situation and needs based on a series of questions posed by the system. In many cases, it is a legal document, so legal advice often boils down to document automation. DoNotPay was developed by Stanford student Joshua Browder in response to his own experience with parking tickets. But law firms are also interested in offering legal advice systems. Automation has obvious advantages here, as it provides legal services to those who otherwise could not afford it or would be willing to pay for it.
A study by the National Legal Research Group, Inc. found that AI tools allowed experienced legal researchers to complete their searches 24.5% faster than traditional legal research, saving between 132 and 210 hours per year. These extra hours can be spent designing, reviewing, and managing cases at a higher level, tasks that computers can`t do. Project Recon, a collaboration between computer scientists and lawyers at Stanford and the University of Oregon, aims to pilot a machine learning system that reads records for California`s probation hearing system. We received a dataset of over 35,000 transcripts from probation hearings from the State of California. Last year, we won a lawsuit against the California Department of Corrections and Rehabilitation (CDCR) to obtain data on race and representation of attorneys for the prisoners mentioned at the hearings. We look forward to the many technical challenges that lie ahead. Ambiguity is an important issue in legal documents.
This can limit the protective measures provided, change the conditions and create confusion for all concerned. Often, lawyers may overlook these issues because they do not recognize ambiguities in the wording of the contract they are drafting or reviewing. Just as people may assume that “I haven`t slept in three days” means that someone suffers from insomnia, a lawyer may believe that the meaning of his clause or amendment is clear to the reader and can only have a reasonable meaning. This is especially true if the lawyer has to review a contract that spans hundreds of pages and contains complex terms and agreements. NLP can contribute to this review by identifying ambiguities and suggesting revisions for improvement. Legal work is rarely easy, which can be frustrating for lawyers and their clients. AI tools such as natural language processing can help streamline and improve legal work and achieve better outcomes for everyone involved. We call this the “recognition approach” and believe that there are demands that go far beyond probation. For example, the approach could be adapted to be applied in social security administration, where administrative judges must decide whether an unemployment claim is valid. It could also come into play in immigration procedures, where an individual official must decide whether or not to grant asylum. In human-led legal decision-making, machine learning can play the role of making mountains of records visible – documents that would otherwise sit on shelves in dusty archives.
We describe this role in an upcoming article in the Berkeley Technology Law Journal entitled “The Recon Approach: A New Direction for Machine Learning in Criminal Law” (published in Berkeley Technology Law Journal). Our team includes Nick McKeown, Professor of Computer Science and Electrical Engineering at Stanford, Kristen Bell, Professor of Law at the University of Oregon, Christopher Manning, Professor of Computer Science and Linguistics at Stanford (Associate Director of Stanford HAI), and Jenny Hong and Catalin Voss, PhD students at Stanford, supported by Stanford HAI. As in many other fields, the nature of work in the legal profession is threatened by NLP and AI in general. In early 2016, Deloitte found that 39% of legal jobs are expected to be automated over the next decade. Recently, McKinsey estimated that 22% of a lawyer`s work and 35% of an articling student`s work could be automated. And as is often the case in other fields, you`ll often see a positive spin, with the usual claims that “technology frees workers to do more interesting things.” .