Squiz Series: Do Aussies trust AI?

This is a squeeze podcast where your shortcut to being informed. Compared to other countries, Australia is low on trust, we're low on acceptance, we're also low on adoption. The majority of people say that they're concerned that elections are being impacted by this, about what AI may over time do to our democratic processes. Through our evolution, we've had millennia of experience that's embedded in our DNA on how to trust other people, for example. Here, this very powerful tool that has some human-like capabilities comes along. Yet we haven't really had much experience in thinking about how do we calibrate our trust in these tools. Good day, I'm Andrew Williams. We're thanks to Minduru Foundation, we're able to bring you this special series on AI, artificial intelligence. Minduru Foundation is an Australian philanthropy that's driven by commitment to create a future where people and the environment we depend on can thrive. And they have a focus on AI, particularly how we can find the right balance between protecting people and unlocking its benefits. So recent research commissioned by Minduru found nearly two-thirds of Australians feel that the pace of AI is too fast. And you, our squeeze audience, have told us in the past via our own polling that trust is an issue. In fact, you trust independent experts over government or tech companies to set the rules for AI in Australia. So in this series, we're going to talk to some of those experts. And first up is Professor Nicole Gillespie. She is a chair of organizational trust with KPMG and a professor of management with the University of Melbourne. And for years now, she's been involved in polling the public on their level of trust in AI. So she's an expert in public sentiment towards it. And she talks about how we're feeling now in Australia, why we rank particularly low when it comes to attitudes towards AI and where you can go if you're keen to learn more. Here's Nicole. Nicole, thanks very much for joining us on the podcast. Let's begin with the basics then. What level of trust or otherwise do Australians currently have in artificial intelligence? Yeah, so look, our 2025 survey of over 48,000 people across 47 countries. It shows that Australians are pretty cautious about AI. So only 36% just over a third of Australians say that they trust AI systems. And we find that that low trust holds across a range of different common applications. So everything from Gen AI, so you know, Claude, touch UBT, Gemini, those sort of tools. But also AI systems just generally, so it's capturing that general attitude. We also ask about trust in AI systems that have specific applications, like AI use in healthcare to aid the diagnosis and treatment of disease or even in human resources to help with shortlisting job applicants. So it's really important because trust is contextual. So we need to really understand it in the context of particular systems. We also find though that Australians are more trusting of the technical ability of AI and its ability to provide a helpful service where their particularly cautious is around the safety and security of using these systems. And also the impact that AI systems, you know, broadly as they scale up, are having on people and also on society. We want to gain the benefits of AI use, but we want that without the negative impacts. So we are really concerned about those negative impacts. And that's also reflected in our mixed emotions. So we find around 45% of Australians say that they're optimistic about AI, you know, wanting those benefits. But we find more 62% say they're worried about AI and quite a lot of people report both. We also find that compared to our earlier surveys, we see a bit of a trend towards concern towards AI actually deepening over time. So for example, we find that Australians willingness to rely on AI and the way that they perceive the trustworthiness of AI systems has actually dropped somewhat. And also worry is increasing probably as people are seeing the impacts more. So you mentioned there that trust is low, how low compared to other countries, where do we rank on that scale as far as trust and AI goes? Yeah, so look compared to other countries, Australia is low on trust, we're low on acceptance, we're also low on adoption. And that's an important part of the mix. In particular, Australia ranks the lowest globally on the acceptance of AI. So we're matched only by New Zealand. So it's just like 49% of Australians say they accept AI use compared to 72% globally. We're really quite significantly lower. And compared to other countries, we're also less optimistic about AI. We report less benefits. It's only just over half of Australians that say that they're realizing benefits from the use of AI, probably reflecting that lower adoption levels. We also find that our trust levels are similar to other advanced economies. So even though we're a little bit lower, we're pretty much on par with other countries like Canada, the UK, Germany, France for example. But we're a lot lower than the emerging economies, particularly like China and India and some of the African countries. Now, when we dig into this a little bit more, we also find that there's no real differences between people across countries when it comes to concerns about the negative impacts of AI. And I think that's significant because it really does show there's quite a lot of global consensus on average four and five people are concerned about a range of negative outcomes from AI. So I think that's a reason for optimism in some way in the sense that we can come together and really work to cooperatively to try to solve some of those issues and address the risks. So really when you tease about where Australia sits, it's not necessarily that we see the risks differently. It's really more that we tend to see the balance between the risks and the benefits differently. I think a really telling indicator there is it's only 30% of Australians that think that the benefits of AI outweigh the risks. And that's again the lowest of all 47 countries we surveyed. It's 69% by comparison in China where they see the benefits as outweighing the risks. Why do you think that perspective exists more in Australia than it does in other countries? Why do you think we have that caution where a country like China doesn't? I think it partly reflects what we consider to be like the drivers of AI. So across four indicators that we see through our modelling is really influencing trust. Australia is actually low on all of them. And I think one that's really important is we're low on AI literacy. So it's only about a quarter of Australians that say that they've had any sort of AI training or sort of education. And it's really hard to trust something if you don't feel that you understand it. So that's definitely one factor. Another is we're not seeing the safeguards to provide that reassurance that the risks are being mitigated. We actually find that it's only 30% of Australians that believe that the current regulations and governance are sufficient to make AI use safe. And that's lower than many other countries as well. So sort of that combination also because we're adopting it less and perhaps because we don't have the training, we're also finding that we're not seeing the benefits and therefore the risks are not being sort of offset by the benefits that we're seeing. It's an interesting chicken in the egg situation isn't it about whether you trust something more like AI because you're using it more, whether you need to trust it to use it in the first place. Which one tends to drive which? Well, I think this is it. Part of the learning with AI comes from its actual use for that. So I often think about you know through our evolution we've had you know millennia of experience that's embedded in our DNA on how to trust other people for example, you know all the micro cues and the different things that we look for. And we're learning that from birth. Here this very powerful human like you know or tool that has some human like capabilities comes along and it is easy to anthropomorphize it and think of it as like a person. Yet we haven't really had much experience in thinking about how do we again sort of calibrate our trust in these tools. So I think that's just a natural sort of part of understanding this. Similarly I think you know we often talk about AI as being having an analogy with something like the car right when the car came along you know people were really upset about this and they weren't trying to slow down cars. You know and cars can be dangerous if we don't have driver's license if we don't have road rules if there's not that sort of appropriate governance and regulation over it. So we're at an early stage I think in figuring out that kind of landscape of how to regulate and govern these technologies well but I'm optimistic that there is a lot of cooperation going on internationally. And all of these things I think do filter through to the public consciousness when they start to see AI being used in delivering benefits and not you know resulting in harms. So you mentioned that the trust sort of needs to be calibrated to the area that's working on are there certain sectors of government or industry where there is particularly big gap between trust and adoption. So what we actually find in our survey is that one of the kind of areas which has been a bit blind to I think appropriate oversight has actually been an employee use of particularly generative AI tools in the workplace. So often you know across many different sectors sort of traditional AI tools that are embedded in products or services for customers for example there's actually been often very good governance over those and where we're quite mature in our way of sort of understanding how to govern those tools. In terms of employees every day use we actually find really high levels of inappropriate and complacent and shadow use and that's often manifesting where we're seeing for example in the media or you know whether it's a consulting company where people have used AI tools for quite prominent government reports for example and then found that it's fabricated whether it's citations or some of the information in it. So we've also seen this happening by lawyers a lot so in the legal industry where again multiple cases where lawyers have put together certain applications to courts and then found out that again fabricated cases have been cited in there. So that does a lot of damage. It does a lot of damage to not only the individual but to the organisation and more broadly across the industry as well. In some contexts, I think that have already been highly regulated. For example, in health and in banks we tend to see less problems, less media headlines about any problems. And I think also in some of the early sort of failure cases, for example, where AI was being used in criminal recidivism, so to actually predict who was likely to commit crimes again. We saw these tools being deployed sometimes with what would be considered to be very low levels of accuracy, sometimes only like 60% accuracy, which is just really inappropriate, whereas I don't think that that would happen now, given the maturity of sort of the governance of these systems. So let's say that I'm someone who's seen these headlines, I don't have like a huge amount of understanding AI and I heavily distrusted just hypothetically. What would you suggest to me if I wanted to improve my literacy or trust in this sort of area? Yeah, so look, I think here in Australia and to get the Australian context, the National AI Centre, which is, you know, part of the government's industry science and resources sort of department, they've got fantastic resources. Everything for say business people who want to adopt AI or put in place appropriate governance, there's really well developed now guidelines on how to do that. Historically, there's also things like the AI ethics principles, for example, which provide a general guide that's been around for some time. There's voluntary AI standards as well, AI safety standards that can be adopted. But I think just for the average sort of person out there, New South Wales TAFE also has, at the moment, their free courses on AI sort of fundamentals. So that's a really great place to start. And the fact that they're free is a bonus too. I do think though that hands-on experience is a really great way to learn with these tools. But first, just being mindful about the strengths and limitations. And I think it's always appropriate to be quite cautious in one's use, and to be really reflective, so to be critically evaluating the outputs before actually using them in your work. One of the things that I often say to people who are worried about, what if I make mistakes with AI is just to really come back to fundamental principles. If you're using it in your work or in your daily life, you're still accountable for what you're doing, whether you use AI or not. So just use AI in a way that you feel confident with and that you would be able to explain to other people and what would pass the SNF test. That's the personal perspective. You recently participated in the AI Roundtable that was held in Canberra where there was more talk about regulation and how the government should approach this. What were your main takeaways from that? Or how do you think the government should look at regulating AI going forward beyond what they're currently doing that you've talked about? Yeah, so look, I mean, I think at the moment we do have voluntary standards and principles in there. I do think for high use cases, using a risk stratification approach, I think it is appropriate that we need to have some enforceable rules. The government's approach to doing this is not to necessarily set up separate laws around AI, like the EU AI Act has, so the European Union. But rather to think about how do we already work with the many laws that can govern AI that we already have, whether that's privacy laws or consumer protection laws or anti-discrimination. So I think we're in a phase at the moment where the test cases for those laws are coming up and we're actually seeing the adaptation and clarification of some of those rules. I think we need to do more on that and I know that there is a lot of focus, particularly by regulatory agencies, to do that. So certainly that's one aspect. I think the government, you know, all governments, not just the Australian government need to be really thinking about how do we tackle sort of misindistimation. One of the, you know, quite startling statistics from our survey is that 87% of people globally and similar numbers in Australia want, you know, clear laws that really combat that misinformation. They want social media and media platforms to find ways that people can be more informed when content is AI generated. They want them to be taking strict action to combat that. It's complex. It's not an easy problem to solve. There's a lot of sort of technical challenges with it, but certainly there's a public mandate to do that. And when we think about what's really at stake there, the majority of people across our survey say that they concerned that elections are being impacted by this. So in the sort of liberal democracies, we do see that very clear statistic where the majority of people are concerned about what AI may over time do to democratic processes, you know, particularly due to just the extreme amount of sort of AI generated disinformation. So if we want to improve adoption in Australia and make sort of a trust at the centre of that, obviously you've talked a lot about how those two things are kind of inextricably linked. What does that look like compared to what we're currently saying at the moment? Well, I think another area that needs a lot of upskilling is industry use and governance of these tools. So, you know, there has been often a speed to market. Maybe one wants to start using these tools and get the productivity and efficiency benefits to keep trust centre stage of companies use of AI. I think it's really important that there's that companies are thinking about the shared benefits they're creating from AI. So it's not just about, you know, productivity and efficiency. It's also translating into real value for customers, better customer service, not just more efficient service, for example. So we find that's a very clear predictor of trust is the extent to which people are seeing those benefits as being shared. And that really plays into I think the Australian government has been very clear that, you know, we want AI to be equitable. So I think a real focus around also copyright and IP and that's something the government has been really trying to negotiate quite hard, I think, with whether it's social media, media companies, even with sort of generative frontier kind of models as well. Like how do we have a fair exchange if we're giving our data, how are we making sure that we're protecting people's IP and copyright through that process? I think another opportunity for government as well when you said what else could government be doing? I mean, some of the, I think most beneficial applications of AI, the kind of applications that can really make a difference to the public, is often in public sector service delivery. There is a lot of patterns, if you like, in that service delivery and AI is very good at that sort of pattern matching and efficiencies around that. We just published a case study showing how New South Wales revenue, for example, used AI in the context of debt collection to actually protect people, to actually identify vulnerable people and give them a different pathway where they wouldn't be, you know, having their bank accounts garner sheet if they didn't pay overdue fines. So that's sort of an example of how AI can actually be used to protect people, to give them extra care, to free up if you like, service providers time so that they can work with cases that perhaps require more care and handling and a bit more complexity. So that was actually an award-winning program and it was highly trusted. And so through that, we've really been able to, I think, show that public sector can be an exemplar of trustworthy and responsible use of AI. So I think that's another opportunity for government is how do they invest in making sure that their own use of AI is exemplary. And I think that then feeds into not only role-modeling for business and commercial organisations, but also at helps the government to then, I think, regulate credibly other organisations as well. Well, you mentioned that's very fast-moving technology, so it'd be fascinating to see what the numbers look like with your next server, which is coming out this year. We're collecting data later this year. It'll be coming out next year. Coming out next year. So we'll be very interested to see what that looks like. Nicole, thanks so much for your time. Thank you, Andrew. Thanks for listening and thanks to the Mindarrue Foundation for making that interview possible. For more on their research around AI, a link is in your show notes.