Over the last few weeks and months, universities and media outlets have published a stream of material about AI and natural language processing systems. ChatGPT, in particular, appears to have attracted the lion’s share of attention. Universities and media have focused a lot of their intention on how ChatGPT might be used by students for assessment. But what about law professors? Could AI take over instruction? Could AI make law professors redundant?
Your convenor this semester is …
The CEO of ChatGPT’s creator, Sam Altman, recently said:
I would much rather have ChatGPT teach me something than go read a textbook.
For instructors, this might sound a little confronting. If you have had the opportunity to play with ChatGPT you’ll know that it has an open, relaxed style of expression. It’s a style of expression that is more approachable than a lot of the textbooks in my office. And it generally answers the question that you ask. But, does that mean ChatGPT could take a role as a professor? Yes, and no.
Teaching any discipline, especially law that relies on precedent, means that there will always be some looking backward. There is always some part of the discussion in class about, ‘Well, this is how we got here.’ That’s where AI does excel. It can gather huge amounts of existing data to summarise what has happened in the past. AI can also identify and summarise patterns in existing data and offer suggestions on trends and potential future developments.
This all starts to sound a lot like a well-crafted lecture
Maybe. But, there are some problems.
We already know that AI systems don’t always find the correct data. Google’s AI platform goofed in front of an international audience who could immediately identify the error because they knew the field. Some Reddit users found that ChatGPT fabricates judgment names and citations. Our students don’t know if what they are reading from an AI system is correct.
AI also relies on existing data. The law can change quickly. But, publications examining how change will affect issues can take a little while to appear. AI is far from cutting-edge in its own store of knowledge.
There’s evidence of inappropriate, biased, and plain offensive AI responses. That’s because AI systems cannot detect biases in the source data unless they are specifically trained to do so.
Lastly, just like existing database search systems, users’ questions affect the information AI provides. That means the venerable computing concept of ‘garbage in, garbage out’ (GIGO) still applies. AI is clever, but it’s not a mind reader.
What about creativity and problem-solving?
Perhaps the most significant current obstacle to AI making law professors redundant is that it is not designed to teach some of the basic skills expected of graduates. For example, the Threshold Learning Outcomes (TLOs) for Australian law students expect that students will be able to ‘engage in critical analysis’, ‘make a reasoned choice amongst alternatives’, and ‘think creatively’. The commentary to the TLOs emphasises that what’s expected is that students will be able to weigh up different solutions to context-specific problems and recommend what’s ‘appropriate’.
The plain text of the law can’t solve every problem. But relies on applying the plain text to answer a question. That means that perhaps AI might be able to provide greater access to justice through simple advice on simple issues. It won’t help with messy, complex or emotional problems.
Could AI make law professors redundant?
Right now? No. In fact, how AI systems work means students need more support than ever. Putting an essay question into an AI is seductive. But, from the point of view of accuracy, critical thinking and creativity, it’s a step backwards, not a step forward.
That’s not to suggest that there isn’t value in working with AI. It excels at simple, basic information retrieval that can help students and law professors with simple ‘backgrounding’. It produces clear and friendly summaries that are often easier to understand than traditional textbooks. Subject to some careful checking of what it produces, AI could make teaching even better. But it still has a long way to go before it makes professors redundant.
Image: Mike MacKenzie. Image via www.vpnsrus.com
