AI in Litigation Insights
Andrew Judkins* The role of AI in litigation is receiving increasing focus. In this briefing, we take a quick look at some ways in which AI is already used in the conduct of litigation and how it will be used in the future.
How is AI already used in litigation?
In modern commercial litigation, for practical purposes, disclosure is entirely electronic. As volumes of electronic disclosure have continued to expand, AI-based tools have provided vital support in order to manage the resulting resourcing, time and costs requirements.
A type of AI tool that is used extensively is predictive coding. This is an application of machine learning technology that uses algorithms to identify potentially relevant documents in disclosure exercises. Typically, predictive coding works by human reviewers first identifying a ‘seed set’ of example documents.
The algorithms then analyse the features of the relevant documents (such as keywords, phrases and metadata) and the coding decisions of the human reviewers. It then identifies similar documents ranked by relevance.
Predictive coding emerged in the early 2010s. It was first endorsed by the English courts for use in High Court litigation in 2016.
The procedural rules which govern disclosure in the Business and Property Courts (introduced in 2019 in their current form) expressly contemplate the use of predictive coding. Clearly, the role of AI in disclosure (and the extent this is reflected in procedural rules) will only increase.
2. Legal research databases
A critical part of preparing litigation submissions and advice to clients is identification of relevant case authorities and legislative/regulatory provisions. Identifying case authorities through manual review alone is no longer practical in the modern era: many thousands of judgments are reported in England & Wales each year and judgments in commercial cases often run to hundreds or thousands of paragraphs.
As a result, searches and review are invariably facilitated by legal databases. These databases use natural language processing algorithms in a similar way to search engines in order to maximise the relevancy of results. These tools map the user’s searches to relevant words and phrases in documents in the database.
An example of a specific technique used in legal databases is ‘term frequency/inverse document frequency’ (TF-IDF). Term frequency refers to counting the presence of the user’s search terms in matching documents (on the basis that this correlates to relevancy).
Inverse document frequency refers to discounting the overall frequency of the word across all documents (to avoid a commonly used term dominating the search ranking).
3. ‘Robot lawyers’
At present, the drafting of pleadings, witness statements and other court documents are prepared almost entirely by manual inputs from human lawyers.
However, within the domain of relatively non-complex claims (such as certain types of consumer claims) AI tools have existed for several years which can prepare such claims. For example, in 2015, a ‘robot lawyer’ service was launched which could automatically generate documents which contest parking tickets based on input from a user. The service now offers other services such as hotel complaints and flight compensation.
In terms of more recent developments, while ChatGPT is not designed specifically for legal research (nor regulated for that purpose), it is capable of generating responses to questions on legal matters. In May 2023, there was a cautionary tale in the US District Court for the Southern District of New York. In a personal injury claim brought against an airline, the claimant’s lawyers used ChatGPT to draft legal submissions, which generated case authorities (along with quotes and citations) that were fictitious.
How might AI be used in litigation in the future?
4. Enhancing legal arguments and expert opinions
As noted above, AI is already used to prepare legal arguments in some relatively narrow senses, e.g. legal databases use AI tools in the search process.
In the future, it is likely that AI tools (specifically designed for litigation, unlike ChatGPT) will play a broader and more direct role in preparing legal arguments and related tasks such as cross-examination preparation. This is likely to include identifying relevant case authorities and legislative provisions, identifying inconsistencies in witness statements and identifying key disclosure documents on specific issues. At some point, it is likely that AI tools will be available that are capable of generating substantive drafting input for legal arguments.
AI also has many applications in relation to how expert evidence is prepared, including pattern analysis of data sets. For example, AI tools are already used in the construction industry to identify defects in buildings through assessment of photographs or videos.
In June 2023, the Court of Kings’ Bench in Manitoba, Canada became one of the first common law courts to issue procedural rules requiring litigants to identify whether and how AI has been used in submissions to the court.
5. Predicting the outcome of the case
As mentioned above, machine learning technology has extensive application in disclosure. It is also beginning to be used in relation to analysing judgments (e.g. data points such as the court, judge, parties, damages awarded etc). A potential application of this analysis would be a data driven approach to litigation strategy and even predicting the outcome of the case.
In 2016, researchers at University College London, the University of Sheffield and the University of Pennsylvania created a model using machine learning algorithms that was able to predict the outcome of European Court of Human Rights cases with 79% accuracy. A similar study has been performed in relation to US Supreme Court cases, which achieved 70% accuracy.
The applications used in these studies were retrospective (i.e. they analysed case reports in respect of cases that had already been decided). However, with the increased use of litigation data analytics, there will be applications that can take data inputs and produce a forward-looking prediction of the outcome of the case. This could have applications in relation to, for example, evaluating the merits of a case prior to commencing proceedings or in relation to settlement strategy.
6. Electronic courtroom
In recent years and particularly since the pandemic, hearing bundles have become largely electronic in commercial disputes. However, the selection of documents for inclusion in bundles is still a manual process. As the sophistication of the use of AI in disclosure continues to increase, it is likely that AI tools will be able to further automate the hearing bundle process, including the selection of documents.
In addition, other aspects of the court room may become entirely electronic. For example, applications already exist which use AI speech technology to transcribe human speech without the need for a human transcriber. As such tools become more sophisticated and reliable over time, they are likely to be adopted for use in courtrooms. Similarly, AI is likely to play a role in courtroom translations, e.g. in relation to foreign language witness evidence.
AI already plays an important role in litigation. It is inevitable that the role of AI in litigation will continue to expand. As with AI generally, the breadth and pace of change will be significant and reshape how litigation is conducted.
Andrew Judkins is a dispute resolution lawyer based in London with Norton Rose Fulbright. He has extensive experience of High Court litigation, international arbitration and investigations. He also frequently advises on disputes risk assessment and mitigation issues. He is recognised as a ‘Rising Star’ in Legal 500 2023 (Commercial Litigation: Premium) and ‘One to Watch’ in Best Lawyers 2023 (Commercial Dispute Resolution).