The honest starting point for this approach was that students were going to use AI whether or not I designed for it. So, the question wasn't “how do I stop this?”, it was “what does good use actually look like in law and how do I teach it?”. I worked to address this through three areas:
Building critical judgement
A legal argument is only as good as the source it rests on. However, while AI is very good at producing text that sounds like a legal argument it is often wrong about the authority underneath it. It invents cases, misstates principles, or cites incorrect sources. If I do not support students to tell the difference between something that sounds like the law and something that is the law, an AI tool is not going to help them.
Supporting equity
In my unit many students are working in their second or third language. For these students AI is genuinely useful. It can help them get a first draft down or check whether sentences read naturally and that lowers a real barrier. However, if I leave AI use unstructured, the students who already know the discipline get the most out of it, and the students still learning the discipline get the least because they can't evaluate the AIs response. An unmanaged AI environment doesn't level the playing field, it continues to reinforce the differences.
Creating consistency
As a chief examiner of a unit that has grown substantially, marking consistency at scale has become a challenge. A rubric that's slightly ambiguous will be read by 15 people, slightly different ways and that's a fairness issue for students. As part of scaffolding student’s use of AI I have also worked to ensure that assessment instructions and rubrics are clear to ensure consistency and that these support how I allow AI to be used.