Written by Samuel Dahan, Zach Berg, David Liang, Queens University
I. Introduction: Canada’s Access-to-Justice Paradox and the Self-Medication Problem
When a right is violated or damage is caused, the right to effectively access the courts is of fundamental importance for the injured individual and for the “enforcement of every legal right […] which exists under the rule of law”. Yet, while there is a strong legal argument to be made in support of the existence a fundamental right to access to justice, it is not an absolute right. In fact, the Charter does not provide a broad “right to counsel”. An individual may be self-represented if they cannot afford counsel, do not qualify for legal aid, and cannot convince the court that state-funded counsel is necessary for a fair trial.
This is problematic for two reasons. First, access to courts without actionable legal help – that is, without knowing how to enforce legal rights – renders the concept of access to justice devoid of substance.Access to courts without representation is like access to a healthcare system in which only the most severely ill can get medical help, and everyone else must self-medicate because doctors are unaffordable.
Second, the number of people that do not qualify for legal help includes most litigants. As noted by Julie McFarlane, 86% of self-represented litigants seeklegal assistance, but ultimately chose not to hire a lawyer. This means that most people, both poor and middle-class, are summarily excluded from the system. Why is that? This brings us to what we call the access-to-justice paradox: Most cases are too simple for hiring a lawyer, but too complicated to resolve without one. While the legal issue involved may not be complex, the specifics of each case require significant legal research, making legal representation expensive. For instance, while most employment lawyers negotiate a contingency agreement, many still charge an hourly rate of $200-$400.
We believe that data science research can make a difference. While algorithms can’t replace good legal and negotiation skills, data science can help all Canadians determine whether they have a case and obtain actionable legal help.
II. What Can Data Science Do for Access to Justice?
In theory, anyone can access legal information online. What is more difficult and expensive is to obtain relevant information such as knowing the odds of winning a case, determining whether a right has been violated, or calculating damages. In addition, once information is obtained, another challenge is to determine what to do with that information, that is, how to enforce a right, or make an employer pay notice damages.
What can be done about it? Canadian researchers have been devoting significant attention to finding solutions. One promising solution has come out of Queen’s and McGill Universities, where researchers have leveraged artificial intelligence to put relevant legal information into the hands of people who need it, not just people who can afford it.
The principle of legal precedent dictates that when dealing with a legal issue, previous determinations of that issue by the court should be binding. In practice, this means that the answer to a legal question can be determined by finding cases involving the same legal issue and similar facts, and then accounting for slight differences in the facts. This task requires significant legal skills and can only be done effectively by someone with legal training.
But what if this weren’t the only way to solve a legal question? The emergence of AI has made this a real possibility. AI algorithms are now capable of understanding precedents and their subsequent treatments in the case law, as well as suggesting possible outcomes of a hypothetical situation. This is because a computer can conduct complex statistical analysis of how a set of factors fit together based on exactly how those factors have been treated by judges in the past.
Using AI to predict legal outcomes is a relatively new idea and has demonstrated promise. AI will not replace lawyers, but will greatly increase their efficiency. If a lawyer does not have to spend all day doing research because an algorithm has found the most relevant cases in seconds, it greatly reduces the amount of time they might have to spend on a file.
AI algorithms have also shown promise in providing non-lawyers with relevant legal information in relatively simple cases. In some situations, an algorithm can provide a non-lawyer with information about their situation and what the likely outcome would be if they were to proceed with their legal action. Empowered with this information, a person can make a decision that is best for them, answering questions such as, Does it make financial sense to hire a lawyer? Is this a good settlement offer? What are my rights?
III. AI Legal Aid in Practice: MyOpenCourt
The Conflict Analytics Lab, a global consortium concerned with the application of data science to dispute resolution, recently launched MyOpenCourt, an open-access platform designed to help the two millions of Canadians who have just lost their jobs due to the Covid-19 pandemic. MyOpenCourt uses AI to help users, both employees and employers, determine whether a worker is an employee or an independent contractor and the reasonable notice period they are entitled to upon termination.
The tools ask questions related to the worker’s specific circumstance, which are derived from the traditional Bardal Factors and the Sagaz Test. Once the user inputs the relevant information, the algorithm generates a prediction along with a summary of relevant case law and the main reasons leading to the prediction. In the case of the “Termination Compensation Calculator”, the prediction takes the form of a final award amount likely to be stipulated by a court, which also factors in the duty to mitigate by the dismissed employee. As for the classifier, once the user inputs the relevant information, such as ownership of tools or the ability to subcontract work, the algorithm determine whether the worker was misclassified and what can be done about it. Should the system determine the user has a case, they are offered the option of scheduling a free consultation with a lawyer from among MyOpenCourt’s partner firms. If accepted, the lawyer receives the user’s answers, the tool’s predicted results, and a pre-drafted demand letter with the relevant case law to assist with the lawyer’s preliminary legal research.
MyOpenCourt is the result of several years of work by researchers at Queen’s, McGill and Brandeis Universities as well as Google AI and the Scotiabank Centre for Customer Analytics. The undertaking was originally a theoretical project aimed at testing advanced algorithms used in technological legal solutions that have proved successful (such as BlueJ or Lex Machina). However, the project took a different turn when the researchers realized that the democratization of the research could potentially have a great impact on workers, small- to medium enterprises (SMEs) and smaller firms. SMEs can use the system to determine their litigation risk. As for firms, the system can operate as a research tool as well as a screening system that helps to more efficiently determine the merits of a case, facilitating out-of-court settlements by providing a middle ground to begin negotiations.
IV. The Future of Access to Justice – Shedding New Light on Dispute Settlement
While it is difficult to predict the future of access to justice, we hope that analytics and legal aid will play a role. In the case of MyOpenCourt, it has already ventured into new areas such as trademark likelihood of confusion, investment complaints, and personal injury. As a further step, the team is now exploring how to retrain new models based not only on case law but also on past negotiations without violating the confidentiality of the negotiation agreements. Considering that most employment disputes are resolved through negotiation, this research should shed new light on dispute settlement in the employment context. It will be interesting to investigate negotiation biases and whether negotiation outcomes are aligned with the case law. Finally, while this is very much a work in progress, it would be a significant step forward if our legal partners had access to predictions based on both court decisions and negotiation outcomes.
Zach Berg, Samuel Dahan, David Liang, “Can AI Improve Access to Justice? An Employment Law Example” Canadian Law of Work Forum (June 25 2020): http://lawofwork.ca/?p=12762
BCGEU v British Columbia (Attorney General), 2 SCR 214 at para 15; Access to Justice as a Human Right(Oxford University Press).
Christie v. British Columbia (Attorney General),2007 SCC 21
As explained in A. (J.) v. Winnipeg Child & Family Services, 2003 MBCA 154 (Man. C.A.) [para. 34].
Bardal v. Globe and Mail Ltd. (1960).
671122 Ontario Ltd. v. Sagaz Industries(2001).