Welcome and introducing the session

G'day, welcome to the National Virtual Expo. I'm your host, Rob Malicki, on Gadigal land in Sydney. Welcome to this session on studying with AI. Let's be honest: if you're a student watching this, we know you are using AI, because just about everybody is. You're part of what I call Generation GPT — the first generation to grow up using these tools as natives. And if you're a parent watching along, like it or not, chances are your student is using AI in some way in relation to their study. So this session is all about how we do that as well as we possibly can. To that end, I'm joined by Aaron Driver from the University of New England. Aaron, welcome to the National Virtual Expo.

Guest: Thanks for having me, Rob. I'm really looking forward to this session.

Rob: The work that you and the team at UNE are doing — actually putting these tools into the hands of students so they can do better at their studies — is pretty phenomenal.

Guest: I'm the Chief AI Officer at UNE and I run Luminix70, which is the university's applied AI lab. We've rolled out a fully fledged AI platform to all of our students. It's a platform that lets students choose models, choose tools, and actually be builders and makers with AI rather than passive consumers of it. We're already getting a huge amount of feedback from our students that the technology is super valuable for helping with study and planning study.

Other AI sessions at the National Virtual Expo

Rob: You and I are doing a number of sessions throughout this National Virtual Expo covering different elements of AI, because this is revolutionising the way work happens in just about every field. What are the other sessions we're covering?

Guest: We'll tackle AI fundamentals for parents — that was actually your idea, Rob, and it's a great one. We'll talk directly to parents about how AI works, the sorts of things you need to be aware of, and how you can support your children. We've also got an AI safety discussion coming up, which is an important one for both parents and students. Our students have to complete a mandatory AI safety course before they're given access to the platform, so I think that'll be a good chat where we really dig into the dangers of using these tools inappropriately. We want to make sure everyone stays safe. And then right at the end of the week we're going to talk about what AI means for jobs and choosing degrees.

Rob: That last one feels very distant right now, Aaron — it's like the end of the week! And folks, if you're watching live, we'd love your questions and comments. You can drop them in the chat right here on the Session Hub page, or if you're joining us on social media, jump in the comments and those will come through. Everybody gets a reply, whether it's something specific about AI or a broader question about uni. If you're watching this on the replay, you can ask your questions in the CYU Discord — details are below. And parents, we have a Facebook community just for you. Let's get into it.

Three ways AI can help you study

Guest: I've been thinking about how to best break this down, and maybe we can go over three main areas. First, how AI can help us study — actually learning concepts, testing yourself, and preparing for exams. Second, how it can help with assignments, when it comes time to do actual work and research. And third, how AI can help with planning and emotional support. We all get stressed, anxious, and overwhelmed — I certainly do — and AI can be really useful in that area. I know we've got a lot of Year 12 students registered for the expo with high-stakes exams coming up, so I think this will be really topical.

Rob: Let's jump in with that first one.

Guest: UNE was the first university in the country to offer a fully fledged AI platform for students. What I'll do is share my screen and demonstrate a few tips and techniques our students are using, so hopefully everyone watching at home can get some benefits from that.

Rob: And of course, you can do this on a smaller scale by jumping into one of the tools that are out there — ChatGPT, Claude, or Gemini. The advanced uni students have got them all in one place.

Guest: That's right. So here we are in the platform. Students are given a default assistant, and like Rob said, they can use Claude, ChatGPT, or Gemini. Our platform is a little more sophisticated in that students can choose or create multiple assistants and choose multiple models. For today, just understand that the model is the actual large language model — the AI brain — and the assistant or agent is what you build on top of that. The model is the brain, but the assistant is your actual digital tutor, your digital helper.

Using a study assistant to learn concepts without cognitive surrender

Guest: UNE students have access to an assistant called Study Mate, which is designed to help you study. Let's try it out. I might say something like: "Explain how supply and demand works and give me some examples." That's a pretty basic economics question, but let's just do something simple. So what we get back is: "Supply and demand is a core economics concept. Before diving in, what's your course or unit?" So Study Mate has asked the student more about what they're studying rather than just launching into an answer. I'm going to say "high school economics." And it comes back: "In your own words, what do you think supply means and what do you think demand means? Don't worry about getting it perfect, just give me your best shot." Straight away we can see the AI's value in cognitive engagement. It hasn't just given me an answer — it's asked me to think about it first.

A colleague of mine at UNE likes to say there are only three things in life you can't outsource to other people: eating, exercise, and learning. And I really like that. You can't outsource your learning to somebody else. So the first thing when using technology for your learning is: resist the temptation to just get answers. Get the AI to work with you. There's a term in AI circles called cognitive surrender, or cognitive offloading — it means you're letting the AI take over your thinking and do the work for you. I get the temptation, but if you do that, entire parts of your brain are going to atrophy and not develop.

So I'm going to give some basic answers here: "Supply is how much of something I have. Demand is how much people want it." It zooms in on price, because I deliberately left price out. Then it asks: "If the price of a burger goes up, do you think the number of burgers a producer wants to sell goes up or down, and why?" I said it goes down — actually it's the opposite, and it's a really common mix-up. So now I've learned something by being wrong.

Being wrong with the AI, when it's just you and the AI, is a great feeling, because you don't need to be embarrassed. No one else is seeing you be wrong. You can ask silly questions over and over again and learn in a way that, let's be honest, most people aren't comfortable doing in a classroom surrounded by all their mates. You can also say "Can you give me a quiz question for practice?" and it'll generate a multiple-choice question. And if you ask it to just give you the answer, Study Mate pushes back and says you'll get much more out of it if you have a go first. Again: always resist cognitive offloading, always resist cognitive surrender.

Rob: And just as you're demonstrating that, Aaron, what I'm thinking about is the other material you could feed into an agent. You could take the entire syllabus for Year 12 Biology and drop that in, so the assistant has background knowledge of everything you're supposed to know by the time you get to that exam at the end of the year.

Guest: That's a great example. It depends on how much content you have — whether you put in an entire syllabus or just a summary, and maybe week by week you put in detailed notes on what you're working on that week. If you put in tens of thousands of words, the AI will summarise the document and only keep an index of it. And depending on which service you're using, if it's a free service there may be size limits. So you just have to play around with what information you put in. But yes, put that syllabus in, put that content in.

Understanding context windows and when to start a new chat

Rob: That's a really important thing to flag. These models, when you're in a chat, have what's called context — a certain volume of information you can use during a session. But the more you use a particular thread, the more you use up that context window, and the quality of answers can actually degrade over time. As a rule of thumb, you want to set up an agent with all the background knowledge about what you're trying to achieve, and then create new threads relatively frequently as you're working. If you're doing a quiz this week and an assignment next week, have two separate threads so the context is fresh each time.

Guest: Spot on, Rob. You could describe the context as the AI's working memory. It's like RAM on your computer, or like your own working memory day to day. If you put too much in your head you can't keep track of it all, and you'll get a feeling for when the AI isn't keeping track anymore — it'll start going off the rails. When that happens, start a new chat like Rob says.

Building your own customised study assistant

Guest: So how do you build an assistant like this? Let me show you the system prompt behind Study Mate. There's quite a large set of instructions written by our team in the lab at UNE. If you were to create yourself a custom GPT in ChatGPT, or an assistant in Claude or Gemini, this would be your system prompt or system instructions.

Here are the key features I would suggest you include in any assistant you create to help you through your HSC. First, start by understanding the student — put their name in, list the subjects they're studying. Second, scaffold understanding — start by knowing what the student already knows. Third, encourage active learning — let the student do the thinking, never just give answers. Fourth, reinforce learning — after a concept is taught, prompt the student to restate or apply it. Fifth, keep it dynamic — mix it up and ask one question at a time. That last one is really important.

Also, create a personality that works for you. Our assistant has to work for all 22,000 of our students, so the personality is friendly, grounded, intelligent — like a dynamic, supportive tutor. But you might want someone who's dry and funny, or a stern tutor, or someone who makes cheesy jokes all the time. Whatever works for you. Have fun with it.

And why not create a custom-tailored assistant that knows you, knows the subjects you're studying, and knows how you like to learn?

Rob: That last point is absolutely critical, because some people are visual, some people like to listen. We're all different in how we learn. This lets you literally customise your tutor to your needs.

Using AI to research and prepare assignments

Guest: The second area is preparing for assignments. I've opened up a different assistant here — an academic research assistant I created. Its prompt tells it to always verify resources and use specific tools to find information. I've asked it to use our Scholar AI tool to conduct a mini literature review on how the academic literature on supply and demand has evolved in the last ten years, with an executive summary, the top eight papers, research gaps and limitations, and an annotated bibliography.

It's running several parallel searches — and based on that initial wave of results it's decided to run more targeted searches on certain subjects. It ran eight searches in total. Could you imagine how long it would take a human to do all eight of those searches? Now we have a mini literature review on the evolution of supply and demand: an executive summary, the top eight papers each with a one-sentence insight, gaps and limitations in the literature, and a bibliography. You can click on any of these and open up the paper in question. This is the kind of research our students do to write a university economics paper.

The point is, even if you're not doing full literature reviews and searching academic databases, you can use the same methodology to do research. Instead, it might just use Google. So AI can help you first to understand concepts, and then to actually do the work. Really useful in both cases.

Rob: This is a good moment to shout out the AI safety session we're doing later in the week. We're going to dive into how to use this technology safely — and by safely I mean how to make sure it doesn't lead you into completely false territory. Any result generated by AI can be wrong, so please come along to that session on AI safety, or find it in the catalogue if you're watching this on the recording.

Guest: Great point. Hallucination is a real risk, absolutely.

AI for study planning and managing exam stress

Guest: The third area is the planning and emotional support side of things. It's really easy to get overwhelmed with study and stressed and anxious, especially if you're like me and leave everything to the last minute. So you can use your AI not only to help you understand concepts and work on assignments, but to get some emotional support and practical planning help.

Here I've described my situation: I've got two assignments due this Sunday evening and an exam I need to prepare for next Wednesday and I'm freaking out. The AI comes back: "Take a breath. This is manageable. What are the two assignments? Roughly how big are they, and how far along are you with each one?" I say: "A 1,500-word essay and a presentation. Both due this Sunday. Haven't started." It acknowledges that's stressful but says it's still very doable, and asks which of the two assignments I think will take more effort, before building a plan. Then I say: "Help me get started today." Because getting started is often the hardest thing, right? There's that huge barrier to beginning. And it gives me baby steps: brainstorm, outline, find sources — just get started today.

Rob: Can I just give a personal example here? I organised this whole National Virtual Expo — 85 sessions — using exactly this technique over the last two to five months. I designed the schedule week by week, and then as I've been going I've been literally feeding updates back into my own agent and saying, "Here's where I'm up to with these different tasks — can you please recalibrate this for me?" Week by week, month by month, we've looked at what's outstanding, reorganised the schedule, deprioritised some things we decided not to do, and elevated others that became more important. Having this kind of agent sitting in your corner helping you stay organised — honestly, I don't think I would have been able to pull this whole thing off without that technology keeping me on track.

Guest: We're definitely capable of doing so much more when you have a computer superintelligence on your side. Who wouldn't want a mathematical superintelligence that's read the entire internet in your camp? And that's the point of today's talk: by concentrating on getting AI to help you with your study — not cognitively surrendering, but actually helping you learn, helping you prepare and complete assignments, and helping you with the planning and emotional side of things — you can go much, much further and do much better in your HSC or Year 12 than you otherwise would have.

The importance of doing the work yourself

Rob: I just want to get on my soapbox for a second. One of the things I hear a lot from students is a worry about AI stealing jobs. We actually have a whole session on AI and the future of work — that's going to be a really good one if you love filmmaking or writing and you're wondering where the jobs will be. But my personal view is: do the work. There's such a temptation with this technology to just boop, get an answer spit out, rewrite it a little, and feel like you've done something. That's not the work. If you actually do the work using these tools the way Aaron is showing — get it to prompt you, think about it, go away and do the writing — you will be fine. Not just for your exams, but in life.

Guest: The real danger is if you just sit back and let the AI do the work for you. So own it: sometimes we procrastinate, that's fine, but the rest of the time you need to do the work. You get out of something what you put into it. If you don't put anything in, you're not going to get anything out. And the AI can help you — you saw the baby steps it gave me. It just said: "This essay is about — write that down. Then write the next sentence." The whole thousand-mile journey starts with a single step. Just get started.

Rob: It's literally what's happening in your brain. Either you are strengthening those pathways between neurons — and that happens in the little steps, being guided and having to think — or those pathways are getting weaker. If you're cognitively surrendering you're not getting smarter. It's physiological. So you make your decision about which of those two you want. You're either going forward or you're going backwards.

First steps: getting started with AI tools today

Rob: We're running up to time. Aaron, for students watching this right now who are thinking "okay, I want to do this properly" — what's the first thing they should do?

Guest: If you don't already have an account with Claude, ChatGPT, or Gemini, just pick one and open an account. That's step one. Then do a quick Google search on how to create an assistant or agent in whichever tool you chose, and start building your agent. You can come back and refine the system prompt over time. And here's one last tip: you can ask the AI to write its own system prompt. Just tell it: "I'm trying to create an assistant that'll help me study in Year 12. I don't really know how to create a system prompt. Can you make some suggestions?" And boom — out it comes. You don't even have to write the whole system prompt yourself.

Rob: This is going to be so useful for people. You would have loved to have had this going through those exams. The inner Spider-Man in me wants to say: with great power comes great responsibility. It's a bit of a cliche, but it's true. These tools are so powerful, you could coast through school and then fail your exams, because if you're just coasting, that's where you'll end up. But if you take that great power and use it responsibly, you can be jumping from building to building. My guest today is Aaron Driver, Chief AI Officer at the University of New England. Thanks for joining me, Aaron. I look forward to seeing you a little bit later in the week.

Guest: Thanks, Rob. Can't wait.