There is a global “digitisation” race between countries to modernise their courts.
A key trend in international litigation is using artificial intelligence (AI) technology for case analysis and strategic decision-making based on analysis of the prospects of the case succeeding. Earlier this month, on 4 March 2019, the Lord Chief Justice of England and Wales announced the establishment of an AI advisory group. The group will be chaired by Professor Richard Susskind and will advise the judiciary on AI developments, its likely impact on the judiciary and court system and the most pressing effects of AI.
In light of these developments, this post considers why a national digital case law database and predictive analytics are important for the UK litigation market to consider in order for the UK to maintain its reputation as premier centre for dispute resolution.
Why the UK needs a national digitised case law database
Consider the following common scenario for court users: you need to assess how UK law impacts the prospects of your or your client’s claim succeeding at trial, where do you look? If you are prudent, you are most likely to search across several different sources.
The problem is that there is currently no official centralised database of UK case law, legislation and criminal records. According to a Council of Europe survey in 2016, of the 44 member states who participated, only England and Wales, Scotland, Denmark, Serbia and the Russian Federation did not have a centralised national case law database.
UK case law can be accessed through a myriad of sources including public, academic and commercially run online databases and some UK courts publish their decisions online.
There are, however, limitations to not having a single centralised digital database of UK case law. A significant limitation, from an innovation perspective, is that most UK judgments and case decisions are published in varying unstructured formats which can make them difficult for AI to generate valuable insights from the cases.
What is a digitised case law database and why is it important for using AI?
Digitising case law refers to the process of converting records of legal cases into a digital format which can be processed by computers. For example, last year, in order to facilitate innovation in legal services, the Harvard Law School Library Caselaw Access Project digitised and published online over 6.4 million US cases spanning 360 years.
AI refers to the ability of computers to replicate the cognitive abilities of humans. AI works best on structured data. Structured data is digital data organised into clearly-defined and easily-searchable data types, formats and fields which are in a database management structure or system.
The ability for AI to produce useful outputs from datasets also depends on the quality of the data, although there is huge variation in how UK court cases are reported; they are usually published as unstructured data.
What is predictive analytics and how is it used in the litigation industry?
Predictive analytics generally refers to a range of techniques to extract information from existing datasets to identify patterns and predict trends and outcomes. Predictive analytics techniques, such as machine-learning (humans training computer algorithms to code data and the algorithms then autonomously apply what they have learnt to new datasets) are commonly used in AI technologies.
AI systems which apply predictive analytics (predictive AI) combine several technologies and techniques. They typically involve the application of machine-learning to analyse a known, defined set of data to study the patterns and relationships within a dataset. This learning is then extrapolated and applied to new data in order to provide predictions which can be used in human decision-making.
Currently, the main use of AI in the litigation industry is predictive coding to facilitate the review and production of large datasets for the disclosure process in civil litigation. However, “litigation prediction AI” or “case outcome prediction AI” is also increasingly being used in a range of contexts. Litigation prediction AI uses predictive analytics and other techniques to analyse large case law datasets at superhuman speeds in order to forecast case outcomes using various parameters, such as judicial decision history, case success rates of legal counsel and the type of court application. The predictions generated by the system can then guide decisions, such as whether to settle a case or to pursue it in the courts.
Litigation prediction AI is widely used in the USA. Key industries using the technology include litigation firms using the data to inform the advice they provide to clients on case prospects and litigation strategy, and insurance companies and litigation funders using predictive analytics to guide how they decide which litigation to fund or insure. In comparison, litigation prediction technology is not as prevalent in the UK legal sector and has only recently started to become available for UK cases.
Is predictive analytics transparent enough to use in the litigation sector?
UK judgments are published as unstructured data and need to be converted into structured data before predictive AI algorithms can be applied. The conversion process is usually carried out by humans. From the perspective of the users of AI systems, there is very little transparency in this process. This makes it difficult to verify if mistakes have been made. This, in turn, could have an impact on the accuracy and reliability of the predictions made by the predictive AI system which is applied to the data. It is also important to understand what data has been used to train the system in order to assess whether there is any unconscious bias in the algorithms.
In addition, there is a tension between, on the one hand, the desire to protect the intellectual property rights in the source code and technical documentation for AI systems and, on the other, the need for sufficient transparency to audit the accuracy and intellectual integrity of AI. For example, it cannot be fully explained how HART, an AI system predicting the risk of suspects committing further crimes which is used to help UK police with custody decisions, works. The HART model contains over 4.2 million decision points (points at which decisions which are made about the data), which are interdependent on other decisions which precede them, and the police force testing the system has refused to publish the underlying code.
A national digital UK case law database and predictive analytics could facilitate deeper game-changing case analysis and litigation strategy. Nevertheless, the litigation industry ought to be cautious and measured about using these technologies.
Both the challenges and opportunities should be examined to ensure that there is transparency and adequate human supervision of how predictive analytics are used. As Nobel Peace Prize-winning Internationalist, Christian Lous Lange, cautioned: “Technology is a useful servant but a dangerous master.”
Natalie Osafo is President of the Junior London Solicitors Litigation Association.