Oxford Conference: Unlocking the Potential of Artificial Intelligence for English Law
AI is taking over!!
Well, it is not taking over the world, at least yet, but it sure is taking over the news and discussions in everyday life, on the Internet, and in academia. The disruptiveness of technology and its number of possible applications is not something to be ignored. We are likely to see drastic changes in areas reliant on human knowledge and many anticipate a transformation on a global economic scale. Among the affected industries, the law system is of particular interest since it holds a dual societal role, acting as a governing economic order, but also as the backbone of social order. Considering the consequences of a potential mishandling of AI implementation, extra care should be taken when contemplating about how to proceed in the future.
That was approximately the introduction of the seminar we have been attending for the past four weeks. Of course, with an introduction like this one, one can immediately prepare themselves for a deep dive into some complex and technical issues. And let me assure you, the seminar completely stood up to the expectations. From 2019 to 2021, the University of Oxford, in a joint effort across the departments and faculties of Law, Economics, Computer Science, Education and Said Business School, undertook the demanding task of researching, in a multi-angle and holistic perspective, the questions, challenges and issues that the rise of AI technology poses on the legal sector. Funded by the Industrial Strategy Challenge Funds (ISCF), Next Generation Services Research Programme and UK Research and Innovation (UKRI), the project also focuses on the changes, transformations and adaptations that need to be considered and implemented in order ensure a smooth and successful transition. They explored these issues in a set of four seminars, spanning on a weekly basis through September of 2021, segmented into six work packages (WP), extending from understanding the new formed business models and their introduction in the industry, all the way to specific applications of AI and discussions such as the future of legal education.
Across the webinars, several interesting research findings were presented, and many fascinating discussions took place among the panellists on a variety of topics. We are not going to refer to all of them in this blog post, as all the seminars are available through their website, and we strongly advise you to look into them if you are interested to know more. However, we will provide a general overview of some main points and thought-provoking findings.
Let’s begin with some news which may alleviate the fear many of we lawyers experience at this stage, the fear of substitution by AI. As Prof. John Armour, Professor of Law and Finance at Oxford University, explained, the impact of AI technologies on lawyers will be threefold, with some applications seeking to remove the legal professional from the equation. Don’t worry, it gets better! Because those would mostly be relevant to the menial and routine tasks which can and should be automated. Others will augment the role of legal professionals, as they will complement their work and provide them with tools to work better and faster. Finally, some applications will lead to the creation of new roles for legal experts, such as the legal engineer and legal designer that we already see today. This shift will of course further impact the business model of law firms internationally, possibly by fragmenting legal services into advisory, legal operations and LegalTech. The business model shift is supported by research data from the Said Business School, by Prof. Mari Sako, of Management Studies in Oxford University, and Matthias Qian, Research fellow in Finance and AI, who, after comparing data from London, New York and San Francisco on the LegalTech and FinTech industries, discovered that knowledge diversity plays an important role in the growth rates of start-ups in LegalTech, with start-ups employing professionals of a diverse knowledge background growing up to three times more than the ones that operate on a lawyer only basis.
On an even more technical angle, Prof. John Armour along with Prof. Thomas Lukasiewicz of Computer Science and PhD student, Alina Petrova, explored the potential of AI in predicting legal case outcomes in a case study involving ECtHR cases and being able to achieve around 70-80% accuracy in their predictions. Although an enormous success, there are still many questions and issues to be addressed on the subject of outcome prediction. These include the explainability of the predictions and the AI reasoning methods, as well as implementation issues in common law jurisdictions where law is based on precedents, which are not organised in a Machine Learning-friendly way. Furthermore, another huge challenge for the implementation of such algorithms is the lack of available data, since not only is access restricted, but the data itself is fragmented (not all cases go to court) and unannotated. If some of you are now asking questions like “how do we evaluate legal services in order to train the AI to do so?” and “should we just be applying deep learning algorithms used in other industries in law, or should we brainstorm bespoke ways to deal with specific legal issues?”, you should definitely go on and watch the seminars online, as those where the same questions raised by the panellists.
Finally, a significant amount of research was focused, and rightfully so, on the regulatory and educational reforms needed, the interaction of AI with already existing regulations such as Data Protection laws, and how automated decision making could be a part of administrative processes. In a series of presentations, these issues are explored in detail, along with many others; however, as a small taste to whet your appetite, let us restate some intriguing questions resulting from respective research on the field.
Can judgements provide robust evidence on claimants and cases? How compatible is the use of justice data with privacy law and GDPR in legal tech and what would be the legal basis in processing textual justice data? Can decision-making algorithms be used in public law, what risks are involved, is algorithmic bias dangerous, and if so, is it more dangerous than human bias? Are ADM systems transparent? Should the core legal knowledge change, or should institutions provide students with alternative skills to prepare them for future market demands?
If you are interested in hearing more about these issues and would like to watch the recorded seminars yourself, they have all been recorded and can be accessed on the Oxford University’s Website here:
https://www.law.ox.ac.uk/unlocking-potential-artificial-intelligence-english-law