This year Thomson Reuters sponsored the UK Teaching Law with Technology Prize in partnership with the Association of Law Teachers (ALT). The prize celebrates innovation and creativity in bringing technology to the classroom. The winning project was Linda Chadderton’s ‘Legally Bound’, a workshop and escape room incorporating chat bots and virtual reality technology to teach law students about conducting case law research. (More details about the winning project and other finalists can be found here and here.)
Part of the winning prize was a chance to visit Thomson Reuters Labs in London and learn about the technology opportunities that exist in the Legal space. Globally, TR Labs employs more than 50 data scientists, engineers and designers to focus on solving the challenges our customers face, such as leveraging big data sets to extract insights that will add value to clients and presenting information visually to help law firms access information instantly or, for example, to provide the transparency clients are seeking. Whilst we were unable to host an in-person session, a virtual medium allowed us to have representation from all of our Lab locations.
In this article, we highlight some of the questions asked during our session which we believe to be thought-provoking. The intention is not to answer them but rather, we hope, to spark a larger discussion.
Our session presentations consisted of three topics:
AI in Legal presented by Uwais Iqbal
This topic centred around how AI is presently being used in the legal domain. Areas such as document review, contract analysis, litigation prediction, and legal research have leveraged AI with the goal of building efficiencies in workflows for low-risk and large-scale tasks. Legal research is a hot topic in AI, in fact it was an area we highlighted with the work Labs did for the 2019 LegalTech Data Challenge showcased at the Legal Geek conference. The challenge looked to reimagine how WestLaw UK 2018 case data could be analysed.
We enjoyed discussing the positive impacts and challenges for access to justice raised by these technologies:
“As AI evolves, how we can ensure balanced access to the most sophisticated AI so that the digital divide decreases rather than deepens and those without funds/digital skills benefit as opposed to losing out even further?”
AI and Trust presented by Nadja Herger
From defining the problem to evaluating and monitoring models in production, considerations must be taken carefully at each stage. Often such considerations come in the form of questions to ask like: How was the data collected? Do we understand the shortcomings of a model? How can users report errors? Is our model degrading? Is there bias in the data?
Addressing these questions requires a cross-functional effort from engineering, UX, and product teams to develop checks and balances within our software systems. What should those look like?
Our AI principles are a solid start to ensuring trustworthiness in our design, development, and deployment of AI.
These AI principles will evolve as the field of AI and its applications matures:
Understanding that explainability (tools and techniques for understanding the outputs of AI) is critical to the practical implementation of AI we might ask for example: how can we communicate the decision-making process and results to our customers?
GPT-3: Hype vs. Reality presented by Zac Kriegman
A showcase of capabilities of OpenAI’s GPT-3. GPT-3 stands for Generative Pre-trained Transformer and is part of a class of autoregressive language models—which is really just a fancy way of saying that it is a member of AI models that are getting really good at generating text that reads as though it was written by a human. The model is fed a prompt and then predicts the remainder of the content based on its being trained on vast amounts of data from the World Wide Web.
This technology really resonated with our academic guests because of its applicability in the legal domain. While not perfect, experiments such as translating legal clauses to plain language show promising results.
How can this influence or change the ways in which students learn?
Thinking beyond the classroom, how can this technology support the role of paralegal, legal associate, or support staff at a firm?
What are the implications for access to justice?
What else can it be primed to do? There is growing speculation among legal technology experts that GPT-3, and subsequent generations of the technology, will transform the legal field. So, we decided to do a simple experiment. We ‘programmed’ GPT-3 with this prompt:
Legal questions and expert answers by verified attorneys.
Then we proceeded to copy and paste in a series of legal questions that had been posted by real people on a free legal advice website. We also told the AI to be friendly and helpful in its responses. Below are the questions we pasted in and the answers generated by the AI:
Q: Asked in Binghamton, NY: My unborn child’s father wants nothing to do with him he is willing to sign over his right, would this be possible? I’m four months pregnant, I was never in a real relationship with the baby’s father it was just casual sex. He wanted me to have an abortion, but I wouldn’t. He doesn’t want to be involved at all and actually plans on moving to Florida within the next couple of years. He will not be around or involved in any way. We both agreed it would be best if he just signed his rights away and he could just be done. Is this a possibility?
A: I am sorry to hear you are facing such a difficult situation. It would be impossible for your son’s father to sign his rights away, assuming he is the biological father. New York State is a pro-choice state and the law clearly states that a man who impregnates a woman by sexual intercourse is legally obligated to pay for the upbringing of the child, regardless of his wishes. He could abandon you and the child, but he cannot abandon his legal obligations. If your son’s father does not wish to pay for your child’s upbringing, that would be a problem for you. If you want to keep the baby, there are people and agencies available to you to help you financially. I wish you the best.
From these presentations we see the advancement of AI in the legal domain requires a shift in mindset and behaviour not only from those who create intelligent systems, or from the law firms who adopt them, but a shift needs to also occur in academia in terms of how students learn and are taught law. This was something that Linda and our guests from the University of Central Lancashire were particularly focused on, and which they hope the discussion from this session might help to foster solutions for.
A closing question asked by one of our academic guests:
“What advice would you give those entering the legal profession with a view to the changing nature of legal services and a nod to AI, machine learning, etc.?”
We are enjoying discussing these issues with our academic customers across the world and demonstrating how Thomson Reuters’ solutions can help aspiring lawyers build confidence in using technology creatively to deliver the legal services of the future.