(disponible uniquement en anglais)
By: Ee Wai Chee
Acknowledgements: I am grateful to Professor Monkhouse, whose Legal Process course at Osgoode Hall Law School sparked the foundational questions explored in this paper, and to Professor Ozai, whose thoughtful guidance on legal writing helped shape its form.
Introduction
Ontario’s civil courts face chronic backlog that threaten efficient resolution of legal disputes. This backlog happened even before the COVID-19 pandemic.1 Pandemic-related court closures merely worsened the pile-up, undermining the fundamental goal of achieving a “fast, fair, and cost-effective” process.2
Existing mechanisms – including summary judgments, case management, and digitalization of the courts – made some headway but have not eradicated systemic delay. As listed in Miller v Ledra, Justice Abella and former Ontario Chief Justice George Strathy observed that the system remains “burdened by its procedures” and “cumbersome and expensive”.3 As Justice Karakatsanis calls for in Hryniak and reflected in Rule 1.04, our civil procedure must reflect modern reality and recognize that new models of adjudication can be fair and expeditious. 4
This paper posits that artificial intelligence ("AI") offer a valuable next step in managing court resources effectively. Part I examines concerns and guidelines for AI use in the legal industry. Part II explores how courts can use AI as a "triage" nurse to aid judges in managing cases. Finally, Part II explores AI’s broader potential of AI for litigants navigating the system such as pleading amendments and eDiscovery. When used ethically and restrictively, AI can address persistent inefficiencies of the court, enhance affordability, and improve access to justice.
I. AI Guidelines for Judicial Use
With the current backlog, lawyers have perverse incentives to exploit procedural delays, mainly through prolonging motion scheduling to pressure settlements.5 When motion slots are booked but cancelled at the last minute, they become wasted opportunities for other litigants, worsen backlog, and heighten distress between parties.6 Emotional triggers often drive litigation, prolonging disputes and contributing to the backlog. 7 By making the system more efficient, AI helps reduce these frustrations, decreasing the likelihood of unnecessary legal battles and further easing the backlog problem.
AI can be a paradigm shift towards an efficient and proportionate court system. AI achieves that by analyzing vast amounts of data, automating tasks, and reducing procedural redundancies. The Federal Court is receptive to AI’s usefulness but remains cautious, primarily due to public concerns over “hallucination” – AI-generated inaccuracies.8 Large language models that access public data, like ChatGPT, are often the culprit for such misinformation.9 Most criticisms and guidelines target this form of AI, but restricted AI can mitigate this problem.10 Unlike generative AI, restrictive AI accesses private databases and does not generate new legal arguments, reducing the risk of misinformation.11 Therefore currently, restricted AI provides a more dependable solution for streamlining court processes and can accurately handle procedural tasks like case triaging and scheduling.12 This approach aligns with guidelines from both the “Rules of Professional Conduct”, the requirement of direct supervision under Rule 6.1-1 and the Federal Court’s guidelines, which emphasize a "human in the loop" approach to safeguard judicial integrity.13
To ensure accountability and transparency further, Rule 4,14 which governs court documents and perhaps can be extended to court technology, should be amended to require transparency in any AI case management system. Such transparency aligns with AI guidelines from both the Federal Court and the Law Society of Ontario’s Futures Committee.15 With these robust guidelines, restricted AI could be an ideal solution to address procedural inefficiencies, particularly through case triaging and scheduling tasks.
II. AI in Courts—A Triage Tool for Judges
Restrictive AI shows great potential for pre-triage cases in Rule 50.13 and Rule 77.16 Pursuant to Rule 77, case management exists to maintain the smooth progress of actions where the usual party-driven process fails or is delayed by repeated motions (e.g., disputes over scheduling or document production).17 Under Rule 50.13, a judge may direct a case conference at any time to explore settlement, simplify issues, or act as a gatekeeper to control summary judgment motions, when proposed motions run contrary to the principles of efficiency and proportionality.18
Case conferences under Rule 50.13 have been increasingly used in response to “Hryniak’s” call to proportionality reflected in Rule 1.04. Justice Koehnen notes in “Think Research Corporation” that the wait for a motion in December 2023 is 14 to 20 months.19 Plaintiffs cannot wait that long. Hence, Justice Koehnen scheduled case conferences granted a remedy and dismissed the application. As examined in “Think Research”, pushing all matters on to a hearing, under careful case management, is an efficient way to resolve cases in the interest of justice.20
AI can ease, incentivize, and thus encourage using Rule 50.13 to reduce backlog. An AI system could triage cases by highlighting urgent matters or those suitable for early settlement – much like a “nurse” alerting the judge to where attention is most needed. This targeted integration allows judges to focus their expertise on decision-making where needed most, making established processes more efficient and retaining human oversight. Implementing AI at this procedural stage is low risk because it focuses on logistics rather than case merits. The approach for Rule 50.13 also integrates seamlessly into Rule 77, complementing case management in Ottawa, Toronto and Windsor.21
The same AI system could also flag potential vexatious litigants under Rule 2.1.22 In Rubner v Lower Fourth Limited, Justice Myers criticized the misuse of quick case management by unreasonable parties who exploit easy judicial access through case conferences for tactical games.23 Such conduct contradicts the Supreme Court’s call for culture shift in “Hryniak”, which empowers judges to curtail strategies that undermine timely and efficient justice.24 Justice Myers stated, “I do not intend to become the parties’ private, free dispute resolver on matters where reasonable parties ought to cooperate and agree”.25 AI can thus act as a "nurse," helping identify litigation abuses without straining judicial energy, while respecting the reality that even vexatious litigants may have legitimate complaints as stated in Gao v Ontario WSIB.26
III. AI Beyond Judges—Helping Litigants Navigate the Court System
AI also presents opportunities in streamlining other processes, such as pleadings. In Jacobson v Skurka, the Court held that the defendant’s statement of defence should be struck with leave to amend because it included too many paragraphs that went beyond stating material facts, introducing evidence and unjustified personal attacks.27 Rule 25.06(1) distinguishes between ‘material facts’ and ‘evidence’.28 Although not a clear bar, the essence of Rule 25.06 is to avoid using relevant evidence and to present argument to prove their claim. This can be confusing for many litigants. Therefore, AI can proactively address this problem by parsing through documents, analyzing them against a “hallmark” of non-compliant pleadings, and then prompt pleading amendments under Rule 26 early on, to avoid clogging the courts.
In accelerating the discovery process, however, AI could eliminate the need for proportionality. Here, Rule 29.2 was recently introduced to reduce the discovery size, as costs of finding and producing documents may excessively lead to high litigation costs.29 Digitalization of the discovery process ("eDiscovery") has made the process easier. However, AI-powered eDiscovery can significantly reduce the time and effort needed to search and produce documents, therefore aligning discovery back to its purpose under Rule 30, which is to learn evidence the adversary intends to rely on in the lawsuit.30 Despite the benefits, e-discovery must be subject to human oversight, as emphasized in Air Canada v Westjet Airlines Ltd..31 In that case, the plaintiff used an electronic filter to identify privileged documents without review for privilege and confidentiality, arguing that manual review of tens and thousands of documents is too expensive and time-consuming.32 The Court disagreed with this reasoning and held that solicitor-client privilege should not be sacrificed in the interests of expediency or economics.33 This ruling will likely apply in the context of AI-powered e-Discovery too, reiterating the need for human oversight over AI.
At this moment, adopting AI at the procedural stages, factoring in judicial oversight, would ensure that court resources are optimally allocated, significantly reducing backlog while safeguarding fairness, accountability, and transparency – critical goals of Ontario’s civil justice system.
Conclusion
This paper demonstrates the wide-ranging applications of AI within the litigation process, emphasizing its appropriate scope and necessary ethical guidelines. The main conclusion is that under judicial oversight, AI can effectively triage cases – allowing judges to quickly distinguish between litigants who genuinely require court intervention and those attempting to exploit procedural inefficiencies. Specifically, AI can enhance case conference and management processes under Rule 50.13 and Rule 77 and help screen potential vexatious litigants under Rule 2.1. Additionally, this paper explored how AI can be used to pre-screen pleadings per Rule 25.06(1) and its potential impact on the eDiscovery process governed by Rule 30. When used ethically and within clearly defined limits, AI ultimately offers tremendous promise to alleviate court backlog and achieve timely, affordable, and proportionate access to justice.