Column

Tips Tuesday: Getting the Most Out of Legal AI Research Tools

It’s pretty much impossible to get away from a discussion of how AI is going to affect legal practice these days. From AI tools that allow users to summarize documents to tools that create new precedents to tools that carry out legal research, it feels like there’s an AI tool for every part of the legal process. 

We’ve also seen the stories about legal research using AI gone bad: for example, this Federal Court case or this Ontario case. (There’s even a database of legal decisions involving AI hallucinations.) The problem in these situations is not that the lawyer used AI, but rather that the lawyer did not independently confirm that these citations were correct. 

So what are the most important considerations when carrying out legal research using AI?

  1. First and most importantly: a legal research AI tool should not be the only tool used for research. The results provided should not be considered a final answer, but rather as a starting point for research or as confirmation of what has already been found through other means. Legal research AI tools can have issues with answering complex legal questions, especially when the question is complicated or novel in nature.
  2. Is the AI appropriate for legal research? There are several AI tools that are specifically designed for legal research. The advantages of these tools are obvious: they are trained only on legal data and they are far less likely to hallucinate. 
  3. What can the tool do (or not do)? Legal research covers a large area and it would be unrealistic to expect legal research AI tools to be able to do absolutely everything a researcher can do. If you ask a tool to do something it is not designed to do, you’ll be wasting your time and getting unhelpful results.
  4. Is the tool appropriate for research in a specific jurisdiction? A research tool designed for US lawyers is unlikely to provide good results for Canadian law. 
  5. What sources does the legal research AI pull from? In addition to case law and legislation, a legal research AI may pull from secondary sources, A legal research AI that has access to many secondary sources in the area you’re researching is more likely to give an accurate answer than one that doesn’t.
  6. If appropriate, narrow your search by jurisdiction. Given that legislation differs from province to province, limiting your query to one specific province can give better results.
  7. Avoid unnecessary detail in your prompt since it can return irrelevant results. You may need to experiment to find the ideal middle ground between being too general and providing too much detail.
  8. If relying on any cases, legislation, or resources cited by an AI, make sure you read them to confirm that they exist and that they actually say what the AI claims they say. AI tools are excellent at presenting information in a very credible way, but this information should always be double-checked. 
  9. Avoid entering confidential client information. 
  10. Keep a record of what prompts you used and when you used them. If doing research for a third party, it is good practice to make them aware that you used AI.
  11. And finally, make sure you are aware of any guidance your law society has issued on using AI.  

Susannah Tredwell

Comments are closed.