So asks a recent study by two academics from Stanford Law School, David Freeman Engstrom and Nora Freeman Engstrom, on the potential impact of AI on the civil litigation landscape in the US.

The tilting scales of justice

While the US has long been famous for its ‘sue’ culture, recently available data demonstrates a significant fall in the plaintiff win rate in civil cases litigated to judgment in US federal court, from 70% in 1985 to below 40% in recent decades; a fall in median jury awards in state court civil cases from $72,000 in 1992 to only $43,000 in 2005; and a skewing in the composition of the country’s civil dockets, with a rise in debt collection claims and a fall in tort cases.

The study says these trends suggest a widening gulf between the litigation “haves” and “have-nots”. The “haves” are well-resourced litigants, typically repeat players who pursue debt claims or defend tort claims and can hire top tier, well-resourced law firms, and who can often “play for rules, not just wins”. The “have nots” are “one shot” litigants, typically individual plaintiffs in tort actions, and defendants in debt claims, whose means are limited to hiring smaller firms and sole practitioners.

According to the study, not only have the “have nots” not had the means fairly to compete against their better resourced opponents, but it appears the “tilt” of the civil justice system in the US may be getting steeper, with recent reforms to the law, and the way judges manage their caseloads, having rendered the US, in some ways, a less plaintiff-friendly jurisdiction than it once was.

It is against this landscape, the study observes, that champions of legal tech have suggested that there is an opportunity for legal tech to “democratise” litigation and put litigation’s “haves” and “have nots” on a more equal footing, by arming smaller firms and sole practitioners with the tools necessary to do battle against their better resourced opponents, and cutting the cost of legal services, putting lawyers within reach of a wider swathe of people.

But is this a real opportunity, and will AI be key to its realisation?

The (incrementally) shifting litigation landscape

The natural place to start any analysis of the likely impact of AI on the litigation landscape is e-discovery, an area in which AI, in one form or another, has been in use for over a decade. Technology assisted review (TAR) in e-discovery has long been embraced by the courts and legal profession as a tool which can help to narrow, and thereby make more efficient and cost effective, significant discovery review exercises.

While TAR has been around and in use for some time, it has seen improvements over time, albeit gradual ones. The more sophisticated TAR tools, prior to the advent of ChatGPT, work to prioritise documents for review based on machine learning, apply predictive coding, and assign in percentage terms predictions as to whether a cohort of documents are likely to be discoverable in respect of a particular issue.

The capacity for today’s more sophisticated engines to improve on this capability is obvious and is already being harnessed. What is less obvious is the extent to which these more sophisticated engines will come to be accepted by litigants, the legal profession, and the courts as a capable substitute for human review.

Beyond discovery

But AI is having other impacts on the litigation landscape beyond e-discovery. A recent proliferation of legal automation programs including self-help chatbots may have helped to remove some structural barriers to the justice system, for example:

  • providing individuals with a ‘base level’ assessment of their potential rights in respect of common legal problems: the US-based online legal service and chat-bot “Do Not Pay” has helped to overturn more than 100,000 traffic offences since its inception; and
  • the document assembly tool “RocketLawyer” assists individual litigants to create legal documents.

When it comes to legal research, the integration of AI into platforms including Lexis-Nexis and Westlaw will also bring improvements and create opportunities for greater efficiency, reduced costs and increased responsiveness. While it is difficult to envisage AI generating a reliable brief to counsel or a submission to a court or tribunal that does not need at least some input from or oversight by a lawyer, there is clear potential for this software to generate at least a “first draft” (notwithstanding the risks of ‘hallucinating’ case law).

Clearly then, there is scope for legal technology to improve upon some aspects of access to justice, although much will depend on how affordable technologies become. But there are some powerful tools arriving on the scene which, some scholars argue, could see the pendulum swing further towards the ‘haves’.

The potential for AI to widen the gap

Accessing the highest quality and most recent legal technology is likely to be expensive and geared towards those repeat litigants and large law firms willing to pay top dollar. While in a vacuum, any legal technology which assists smaller firms or pro se litigants is beneficial, if the AI available to the “haves” far outpaces that of the “have nots”, the asymmetry between these parties may increase.

One area which scholars have expressed particular concern with is the datafication of the court process. As court’s have increasingly accepted the digitalisation of the legal system through practices such as e-filing and online hearings, the amount and quality of data from relevant legal proceedings has grown. Were litigants granted access to this data, well-resourced law firms and repeat litigants could embark upon a ‘mining’ expedition, developing predictive AI models to leverage against their counterparts and further increase the power imbalance.

For example, companies such as Walmart are already utilising software dubbed by the authors of the abovementioned study as ‘the Walmart Suite’, which seeks to ‘rationalise recurrent areas of litigation’ such as slip-and-falls and employment disputes by evaluating key case factors before offering a prediction about a case’s outcome and likely expense to Walmart. Although it is not yet clear whether the utility of these predictive models may extend beyond frequent litigation matters to more complex, nuanced and longer-tail issues, such technologies have the potential to further widen the gap.

Whether courts will allow access to this information for the purpose of data mining remains to be seen, although it seems unlikely. Still, as two US scholars have argued, this datafication of court processes will mean the legal system’s ability to deliver justice will increasingly depend upon the health and management of its data ecosystem.

Will someone think about the lawyers?

With all the talk of predictive modelling and ‘robot-lawyers’ it is no wonder AI has both heightened expectations and generated trepidation about the future effect on legal practice and the legal landscape more broadly. That trepidation, rightly or wrongly, has likely been felt by more junior lawyers who have probably asked themselves from time to time “am I replaceable?” The reference here to more junior lawyers is not intended to suggest that more senior lawyers and managers have not asked themselves the same question. It is just that the more junior lawyers who are at the coalface of legal practice and who might traditionally be expected to be involved in high volume repeatable workstreams might genuinely ask that question more often than someone who is more progressed in their legal career.

To be fair, the question is a legitimate one. When consideration is had to the number of routine domestic and professional tasks that AI has impacted to date, one could be forgiven for contemplating the issue of replaceability more frequently including in a professional context.

It is reasonable to assume that as AI develops, and new applications are discovered so too will the practise of law. But the real question is not will AI continue to develop and change the way we work. That is a given. The question is ultimately to what extent will AI change the way we work?

Despite the rise in online hearings during the COVID-19 period, courts have largely returned to in-person hearings. Although some form of online dispute resolution remains available, when it comes to large matters, judges expect to see representatives in court. While large scale litigation exists, the role of paralegals and more junior lawyers in the physical logistics of litigation seems unlikely to be overcome by AI. Nor can technology replace the presence of counsel or solicitors within the courtroom. How could a Court reasonably hear a submission from a computer? Equally, it is difficult to fathom a circumstance in which the law will accept administrative decision making using artificial intelligence without human involvement, at least at this point in time. So much can and has gone wrong with automated decision making in the past, a prominent example being the Robodebt Scheme in Australia.

Where does that leave all litigators?

As the growth of legal technology goes from strength to strength, the idea that technology is the “be end all” can be alluring. However, while AI may reduce the justice gap between the “haves” and “have-nots” of litigation, it could also exacerbate existing inequalities.

That AI may evolve to such a point that it can attend accurately and comprehensively to specialised legal tasks such as legal research and e-discovery cannot be put beyond the realms of possibility. While the use of predictive outcome models could certainly be an interesting element in a litigator’s decision-making process, whether it will have any real impact on disputes beyond those of a highly repetitive nature with more straightforward legal reasoning is a question which is yet to be resolved and one that is unlikely to be resolved soon. Similarly, unless another pandemic arises, it seems unlikely we will witness any significant instances of online hearings – creating a consistent need for legal professionals in the courtroom.

But a new technology could arise which shifts the conversation entirely. So, the best thing to do right now is understand the applications that are here and now and how they may be extended into the future.

Read more: Large Legal Fictions: Profiling Legal Hallucinations in Large Language Models