Dr Leo Ng on Design Thinking and AI in Research, Teaching & Health
Physio, award-winning educator and tech innovator Dr Leo Ng is using design thinking and AI to reshape teaching, research and healthcare.
Dr Leo Ng is never short of ideas on how we can better teach, learn and care in a world transformed by technology. He’s a physio, award-winning university lecturer, researcher and AI innovator, among other things. The big question is: how does he fit it all in?
From building AI tools that save researchers hundreds of hours to designing digital teaching models that genuinely connect, Leo brings empathy, insight and creative thinking to everything he does.
This month, he joins AI & Us to share his thoughts on the promise of technology and the ways AI can support – rather than replace – teachers and healthcare professionals. He also reflects on the personal motivation that underlies his work building advanced AI tools to accelerate medical research and offer hope to those facing uncertain futures.
Shauna: So Leo, we first got chatting because you’d read an earlier AI & Us interview with Tony Flanagan, a physio of 40 years about to leave the industry after an extraordinary career. Reading about your work, I was struck by how different things look for the next generation of physios. They’re training in a world where robot clinics already exist and digital tools are standard. Is there any continuity in how physio is taught and practised, or are students entering a completely new profession?
Leo: That’s a great question. I’m still trying to separate whether students will practise differently or whether I hope they will. Even with so many new tech tools available like smartphone apps and wearable sensors, many students are still being taught with the same equipment their mentors used.
The goniometer is a good example. It’s like a big plastic protractor we use to measure angles – for example, to see how far you can bend your elbow. I’ve taught it for 17 years, but now there are all these digital alternatives, including Halo, which was developed by a physio student quite a few years back. And yet people are still using the big plastic goniometer.
We've been saying for the last decade that physios should be carrying different tools in their tool belts. But somehow, it still gets used quite a bit.
There are early adopters using different tools, of course, but I think there are generally three things that limit uptake.
One is the perceived financial cost. Two is the cost of learning a new tool. And three is that knowledge is transferred through generations.
We require and rely on mentors who’ve been practising for years, which is great in that they've seen so many patients and developed their own patterns in their heads. And clinical reasoning for healthcare professionals is about pattern recognition, so that’s experience we don't want to lose. But at the same time they're used to using traditional tools that get passed on with the expectation that our students will use the same thing, because that's how they did it.
That said, the shift is happening. Is it changing at the same speed as other professions? Maybe, maybe not, but then there are reasons to keep the best of the past and present as well.
There's a lot of potential for change that will supplement but not replace what we do. And that's probably another thing. I think there are lots of healthcare professionals who worry that if we use this technology they’ll lose their jobs. But I don’t think that is the case.
If you look at the technology that's being picked up now, it’s really helping us. AI transcription of notes doesn't stop us from seeing patients. It just saves time and reduces admin. But there's still that fear.
Shauna: So how do you prepare your students for this brave new world when so much of it is yet to take shape?
Leo: Good question! What does someone like me with a PhD in low back pain know about teaching? Well honestly looking back I realise that when I started out in higher education I didn’t actually know how to teach that well.
Like many academics, I spent the first ten years teaching the way I was taught, and my teachers had done the same. I used to think “wow, I'm a great lecturer because I gave a two hour lecture, just sharing my experiences!”.
It wasn’t until COVID hit that I decided to formally study education and completed a graduate certificate in higher education. That totally changed how I think about teaching. I realised that just giving a good two-hour lecture doesn’t mean students are learning. I now focus on active learning and engagement and really taking an interest in my students.
As a Vice-Chancellor once told me: five years after graduation, students might not remember what you taught exactly, but they’ll remember how much you cared about them. And that’s so true.
When I catch up with my alumni to see how they are, they really appreciate it, they’ll say, ‘Leo, I became a physio because of you’. Not because of what I taught in terms of the content, but because of the connection and commitment to their learning. That’s what sticks.
Shauna: And where does AI fit into that now that we all have ChatGPT to research and write our assignments?
Leo: When generative AI first emerged, many universities banned it outright. But that doesn’t work. If you ban it, students will still use it secretly. So instead, we focus on teaching them how to use it well and ethically.
I think in five years, AI literacy will be taught in high school or perhaps even earlier. But for now, it’s on us as educators to guide students who missed that foundational training.
For my students I have a huge focus on communication skills. I can be really technical when it comes to building AI tools, but as a teacher I concentrate on ‘future-proofing’ my students by equipping them with communication and clinical reasoning skills as well as technological know-how. That’s because one of the biggest strengths of humans versus AI is our ability to develop therapeutic alliance or rapport.
So for example I might say look at this beautiful system that can write out the exercise program for your patient. I would very much be saying that the delivery and the motivation should be developed by humans, but you can use the digital tool to help the patient remember what the program is.
Shauna: What about the technical side of your work outside the classroom? You’ve gone as far as co-creating an AI tool of your own called Research Screener. Can you tell me a bit about that?
Leo: Research Screener is an AI tool that helps researchers with one of the most painful and time-consuming parts of doing a systematic review. Systematic reviews themselves involve carefully collecting and analysing all the available studies on a specific topic to answer a clear question. That usually means screening thousands of papers to find the ones that are actually relevant – definitely like looking for a needle in a haystack!
Our tool uses AI to make that initial screening step much more efficient. It reads the abstracts rather than the whole papers to screen for relevant abstracts to answer your research question, but it can reduce your screening time from 100 hours to just 5, by prioritising what’s most likely to be useful.
Here, Leo chats about winning the Physio Pitchfest People’s Choice Award for Research Screener, and shares his top tips on developing new tech tools for physios.
Shauna: So much time (and angst) saved! How and why did you get involved in cutting-edge AI development in between all your teaching and research work?
Leo: It started during a commercialisation incubator. At first, I wanted to develop a virtual reality rehab game, but I quickly learned that would cost $200,000 to build and likely no one would pay for it. So I shifted focus. I thought, what’s a real problem researchers face? And systematic reviews came straight to mind.
They’re incredibly time-consuming. Researchers can spend hundreds of hours combing through thousands of papers just to find the relevant ones and that initial screening is the most painful part.
So I joined forces with computer scientist Dr. Kevin Chai and developed Research Screener to help with just that step. The goal was to make the process more efficient, without replacing the deeper work that follows.
It’s now been used on topics from domestic violence to cancer, and we’ve even won awards for our efforts.
As for the why? My motivation is really a personal story. I lost my dad to lung cancer 11 years ago. He was originally given just three months to live, but he ended up surviving six more years thanks to a trial drug called immunotherapy, which not many people had heard of back then. That gave him time to see my wife and I get married, and to meet both our kids.
So if Research Screener can help bring the right study to light even one month earlier, if it helps just one more person get early access to a life-changing treatment or supports someone working in domestic violence… well I’ll die happy if that’s my impact on the world.
I wish I could help more people directly, but both Kevin and I are full-time researchers. We do what we can, and we’re proud of this tool.
Shauna: That’s so thoughtful. You don’t think of tech tools having such a personal story behind them, but for you it’s so clearly about helping people and saving lives. Are you optimistic about AI being used to diagnose and treat patients, like the AI physio clinics being trialled in the UK?
Leo: I’ve followed those trials pretty closely. To understand the current context, you need to look at what they’re replacing.
In the UK’s public system, it’s not uncommon to wait six months or longer to see a physio. So the choice isn’t between the new AI clinic and a human physio. It’s between AI and nothing. If you’re in pain and someone says, ‘Wait nine months to see a physio or use this tool now’, I think most people will try the AI.
If those patients had the option of seeing a physio the next day instead, I’d be inclined to think 90% would choose the human. That’s where our biggest strength lies – our ability to connect, to communicate, to build trust and therapeutic rapport. That’s not something AI can replicate. At least not yet.
We can already see ways in which AI can help with ‘red flags’ in physio. For example, cauda equina syndrome is a serious condition that can cause loss of bladder and bowel control and requires surgery within 24 hours. An AI system that spots those symptoms can tell a patient: ‘Go to hospital, now’. Without that, someone might wait for days and the outcome could be much worse. So if you think about that in the context of the UK trial you can see value.
That also goes for radiology. Traditionally, radiologists read scans in the order they’re taken. But what if the 13th patient has an aneurysm? Their scan might not be reviewed for hours. Now, AI can review the whole pile of scans in seconds, flag potential life-threatening issues and say ‘read this one first’. That kind of prioritisation could save lives.
So yes, these are some of the ways I see AI offering support. But there’s always human oversight. And there needs to be. If I hear something that doesn’t sit right, I’ll dig deeper. A teenager with back pain might mention Year 12 exams, and I’ll ask about stress. They might say, ‘yeah, when I’m really stressed, the pain’s worse’. That’s not in any checklist, it comes from clinical experience and pattern recognition. Will AI get there eventually? Maybe. But we’re not there yet.
And here’s the thing. Even among physios, there are different schools of thought. Some focus on biomechanics. Others, like me, lean more toward the psychosocial model. So the question becomes: who trains the AI? What philosophy does it reflect?
That’s why I like programs like Therapha in the US. It offers multiple possible diagnoses depending on your approach – biomechanical, psychosocial, or a mix. It doesn’t pretend there’s one ‘right’ answer. Instead, it prompts the clinician to think again. Maybe you didn’t consider that psychosocial angle, but now you do.
That’s the kind of support I think AI should offer. Not replacing us. Expanding our thinking.
Shauna: What about when it comes to AI making mistakes and ‘hallucinating’?
Leo: It depends on what we’re comparing it with. We keep comparing AI to perfection, when perfection doesn’t exist. People say, ‘AI got this wrong’, as if humans don’t make mistakes.
But I’ve seen systematic reviews done entirely by humans that miss key papers. If AI helps us catch even one thing we might have missed, that’s already an improvement.
The real question should be: is it better than what we have now? Is it helping? That said I do think healthcare will always need human oversight. But in other areas, maybe AI can operate independently one day. It’s hard to say exactly when or where that will happen, but what’s clear is that the more we work together across disciplines – healthcare, computer science, education, the humanities – the better outcomes we’ll realise.
Shauna: You mentioned out the outset that healthcare professionals were worried about AI’s impact on employment. Do you think a robot will take your job/s?
Leo: No, well at least not in the next decade. I don’t think AI will replace me as a teacher, a physio, or an innovator, as long as I stay curious, communicate well, and use AI to support what I do, not substitute it.
Right now, I’m working at about 160% capacity, according to my wife. If AI can help me scale that back to 120% and I can spend more time with my kids that would be great.
Shauna: And just finally, if you had to sum up your thoughts on the future of AI and humanity in one word what would it be?
Leo: Collaboration. It has to be a partnership. Humans and AI, working together to improve effectiveness and efficiency in everything we do.
Awards galore: Here Leo shares more about the innovative work that earned him Curtin University’s Aspire Award in 2021.
I’m a freelance writer/strategist who shares complex concepts, ideas and insights along with stories from the extraordinary people behind them. If you’d like to work together, contact me.
