Illustration by Gary Neill

Dahan has teamed up with Xiaodan Zhu, assistant professor in the Ingenuity Labs Research Institute and the Department of Electrical and Computer Engineering at Queen’s, to develop an AI (artificial intelligence)-powered set of tools to help level the legal playing field for lower- and middle- income Canadians.

With support from the government of Canada’s New Frontiers in Research Fund (NFRF), Dahan and Zhu, with team members at the Smith School of Business, are building an intelligent dispute resolution system that evens the playing field for self-represented litigants in employment, trademark, and personal injury disputes by offering them data-driven predictions, based on case law trends and negotiation data from thousands of similar disputes.

“We aim to democratize predictive legal analytics, a technology that’s only been available to big law firms and their corporate clients. We’re developing AI tools for Canadians with average incomes, including those who may have lost employment during the COVID-19 crisis and are priced out of the justice system, to represent themselves more effectively in small-claims disputes and improve their access to justice,” says Dahan, who is working with a team of more than 25 law students and data scientists on several cutting-edge tools and applications.

Applications to help Canadians with employment rights

In the wake of COVID-19, many Canadians find themselves out of work or unsure when (or if) they will be going back to their jobs. These situations open workers up to exploitation, particularly if they do not understand their legal rights or have access to an employment lawyer.

In collaboration with Jonathan Touboul from Brandeis University, Dahan and team have developed an intelligent dispute resolution application, , to help recently laid off workers understand their rights and calculate the amount of termination severance pay they could expect to receive under Canadian employment law. This predictive approach is unique because it is based not only on court cases, but also data from negotiated agreements, which is critical since most employment disputes are resolved through settlements. “Basing predictions on case law only is wrong because it doesn’t represent the reality of how most disputes are settled,” says Dahan.

This AI-powered, self-help tool uses deep learning technology to analyze legal trends and data from thousands of cases and negotiated settlements to make informed predictions about litigation outcomes and optimal negotiated agreements. The tools provide Canadian workers with a greater understanding of their options in this difficult time and also works to connect workers with top employment lawyers for a pro bono consultation related to dismissal.

“In the first few weeks of COVID-19’s arrival in Canada, we saw almost a million terminations and layoffs by employers across many different sectors," says Dahan. "These Canadians may be offered less than what they are entitled to, but often have no way of knowing whether the offer is fair or how much they should expect.”

An accurate severance prediction can help the person to evaluate the offer and make an informed decision about whether to accept it or negotiate for an appropriate higher amount. Dahan hopes that Open Court, which was , will eventually be available through a legal public-interest body, such as the Ministry of Justice or Access to Justice associations.

GILBERT aims to advance new AI frontiers in legal prediction and reasoning

Illustration by Gary Neill

When Samuel Dahan was introduced to AI specialist Xiaodan Zhu in 2018, he got excited from their discussion on how deep learning technology could be applied in law to increase the predictive power of the online dispute resolution tools Dahan was developing. Deep learning is an advanced AI technology, which is inspired by and can mimic the workings of the human brain. It can learn very complicated patterns from vast data sets in order to make accurate predictions and smart decisions.

The two researchers, hired at Queen’s in 2017 and 2018, joined forces to win a New Frontiers in Research Fund grant in 2019 and build their deep learning tool, called GILBERT (Generalized Intelligence in Law and Bidirectional Encoder Representations from Transformers), for cutting-edge legal applications in resolving employment, personal injury, and trademark disputes. “GILBERT aims to obtain a model pre-trained on millions of legal cases and legal texts. It works by filling the gaps when faced with new legal facts, using the legal cases and texts that it has been trained on,” explains Zhu.

With each new application, GILBERT gets smarter by building on the knowledge gained through other data or applications. “When we further train this model on the annotated data of specific projects, such as the trademark analytics project, it can make better predictions for those tasks than a model which would analyze legal data only in the context of one project.”

Advancing AI reasoning

Zhu is excited about applying AI and deep learning in new ways to solve practical legal problems affecting millions of Canadians. “This is an opportunity to develop better algorithms for real-world applications and law is a field which may benefit significantly from these advances,” says Zhu, whose research develops deep learning algorithms to enable computers to understand meaning of human language. Before joining Queen’s, Zhu was a researcher at the National Research Council of Canada from 2010 to 2017.

For a computer science researcher, law also presents great challenges and opportunities to increase the capabilities of AI in understanding and performing reasoning. “Causality and reasoning are abundant in law. We hope to train GILBERT to understand the cause of a dispute, potential solutions and the resolution. We are using specific legal cases and applications to develop better algorithms for understanding natural language and reasoning. If we can make progress in solving these core problems, then the impact of AI will be much greater within and beyond the field of law,” he says.

An intelligent tool for detecting knockoffs

Illustration by Gary Neill

�ٲ��󲹲�’s has been appointed to the Dispute Resolution Board of the European Union Intellectual Property Office (EUIPO), the highest specialized court for trademark dispute in the EU, to develop an AI-based, trademark comparison tool. The goal is to help small- and medium-sized enterprises (SMEs) more easily and accurately determine whether the owner’s trademark has been infringed in risk of confusion case, in which owners must show a likelihood of confusion among consumers between brands. Dahan and Zhu are using deep learning technology to assess the degree of similarity in such trademark cases, based on analysis of text and images.

The assessment tool is being trained on published decisions from about 100,000 past cases brought before the Court of Justice of the European Union (CJEU) and the European Union Intellectual Property Office, in order to learn and replicate how decision makers assess the risk of confusion between trademarks. By applying text analytics to determine likelihood of confusion and computer vision to assess the degree of similarity between two brands, this novel tool will help users to better predict whether trademark infringement has occurred.

“Trademark law is one of the best applications for AI in the legal field. There are a lot of cases and images to review to accurately assess risk of confusion, which is a complex, time-consuming task for lawyers and judges. Advances in image processing, for example, allow thousands of brands to be compared in just a few seconds. This project is a great opportunity for us to improve comparison tools in the trademark field,” explains Dahan, who worked on competition and trademark cases at the General Court of the European Union before joining Queen’s Law.

This smart comparison tool could help potential litigants resolve more trademark disputes out of court, since both parties would have easy access to relevant data to predict the outcome of litigation proceedings. Another goal is to help identify inconsistent trends in case law and enable judges to uniformly assess the risk of confusion between brands in future cases. “The European Court of Justice and the European Union Intellectual Property Office are excited about this project because it will help them to encourage better consistency in case law. Our tool could also be applied in Canada because Canadian and European trademark laws are similar,” says Dahan.

Zone of possible agreement and personal injury disputes

Illustration by Gary Neill

The Conflict Analytics Lab is hoping to develop another AI application that could help people quickly and easily resolve motor vehicle injury disputes of up to $50,000. They are engaged in discussions with administrative tribunals in Canada, including Ontario Tribunals, to lay the groundwork for this project.

 “Our AI tool would use deep learning to identify the three most similar personal injury cases and give users an accurate idea about the amount of personal damages they could expect to receive. This would give tribunal participants a zone of possible agreement to help achieve a resolution quickly at minimal cost,” says Dahan, who hopes to integrate the tool into the existing tribunal process.

In collaboration with Maxime Cohen at McGill University, Dahan and Zhu are also developing deep learning applications to help resolve consumer and trademark disputes. They are working with companies in the hospitality, banking, and travel industries on a tool to track and analyze their history of handling customer service disputes, accessing information that has traditionally been kept secret. “This AI system will analyze a vast amount of information to provide guidance and identify best practices for resolving customer service disputes,” says Dahan, noting that consumers could benefit from a similar tool to empower them in resolving disputes with large corporations.

Although his kickboxing days are over, Dahan is clearly not done fighting yet. Only this time the fight is to empower all citizens with data-driven tools to help them resolve legal disputes fairly outside the ring: “In a democratic society, everyone should have access to a minimum level of legal services, just like a minimum wage and basic health services.”