Societal Resilience & AI Change Management

Good afternoon, everyone. I'm a psychological scientist and neuroscientist by training, and I have a lab at UNSW. We study consciousness, mental imagery, intuition, decision-making, things like that. But a few years back, we started to get more and more concerned about the societal impacts, the psychological impacts of artificial intelligence. So we shifted what we're studying to try and understand that, and to almost take a futurist point of view. And we came up with this phrase, neuro futurism, to try and describe using what we know about the brain and mind to predict the effects of scaled-up intelligence — to predict how it's going to affect us, and what we can do about that, from all kinds of different angles.

I'm not a technologist. I'm not an AI researcher per se, but a neuroscientist. Let me start off with a little bit of a reframing. I'd like to start off some of these talks by introducing the concept of when electricity first came on the scene. Way back when, a lot of people thought about it as a lighting technology, and it was framed that way. They thought: better candles, more light, lighter rooms, light on the streets at night time. And if you thought about it and framed it just like that, it's this lighting

revolution. You would not have predicted all the things that came next — refrigeration and computers and electronics and cinema and all the kinds of things we have today. And I think at least for the first few years and still today, we have a framing problem around AI in a similar manner. We keep calling it a technology revolution when it really is an intelligence revolution. And those two things are different. We've had lots of technology revolutions. We kind of know how they play out. We know what to do.

We've seen them before. But intelligence — and we can debate later on what that means — is something different. When you scale up intelligence, you make it fairly ubiquitous, fairly cheap. It has very different impacts on society. But society is built around this idea of intelligence: what we do at home, what we do at work. And so when you start disrupting that, you start to change everything. And that's really one of the threads we've been pulling on and trying to understand: what happens when you start disrupting intelligence across society?

What would you expect? How can you model this out on all different levels? So what will change? Well, everything. Not tomorrow, not by the end of the year, but over the next 5 to 10 years. Every level of society, what you do at home, the way you play with your kids, the way you interact with technology, the legal systems, finance, the economy. Law, of course, jobs — all these things are going to be disrupted in very interesting ways, scary ways and ways that are hard to predict.

So you could even follow that through. A lot of people have been talking about how capitalism itself will be disrupted and how we're going to end up in 10, 20, 30 years in a world where we probably won't have money like we do now. So change is really something to think a lot about. And humans don't handle change well. So I sometimes call this the everything disruption: given enough time, we are going to see a disruption to every layer of society and to the way we live our lives.

And we're not equipped to handle this kind of disruption, these kinds of changes ongoing for a decade or two. Change is a very threatening thing to people, particularly when we don't have control over it. And you can also think about some of these changes already locked in. So some of the data, as I'm sure you know, is what the frontier models can do and how they're being put to work in different industries and in business generally. There's a bit of a gap there. And so you could even think, if we could somehow freeze where AI is right now, we'd still see disruption over the next year or two, or maybe even longer.

So I'm not going to talk about this misalignment, the Terminator stuff. There's plenty of other technology people to talk about that here and elsewhere. I want to talk more about change now. And change is a really strange thing. When it is kind of uncontrollable and unpredictable, large and very fast, that's when it threatens people. That's when it shifts from something that could be a little bit positive to something that's really threatening and puts us in this fight or flight, this anxiety, this stress, and we don't react well to it.

And it's really particularly this controllability — when we perceive that we don't control change is out of our hands. It can be really threatening and it can push people into this sort of learned helplessness state where they really give up because they really think they don't have control over things. We could talk a long time about change, but let me quickly show you a couple of graphs to make a few points. This is years since someone's lost their job, and this is mortality. And see, in the first year or two, mortality, depending on the different age groups and a few other things, goes up.

So the likelihood of dying goes up between 50 and 100% that year to year and a half after. Then it comes back down again. So it's not saying you're 100% going to die. The chances of you dying go up, it doubles, it goes up by 100%. It's at a low base. But that's one way to think about change. That's an impact to all-cause mortality. It's not just that — there's also a cumulative effect. So you can look at data — the number of stressful events or change within a month.

And as you have one, two, three, four or more, the likeliness of getting a strong onset of depression goes up. You see similar things with all kinds of mental health challenges, anxiety. So it's long lasting, it's cumulative. The number of change events you have in a given period of time affects you more and more. And you can break this down to young people, old people. But generally the facts are there. I think it's important to understand that we're not just talking about negative change. So there's interesting data going back, looking at when people win millions of dollars in the lottery.

You come back a few years later and their lives are really messed up and they regret the whole thing, which seems kind of strange and hard to understand. But particularly young people, when they're given a lot of money or a lot of opportunity — when they get drafted into the NBA or get paid a lot of money as a singer — their lives tend to fall apart. Then what the lottery companies did, they brought in sort of change management, for lack of a better word. So they had teams of people to help manage them through this.

Psychologists, counsellors, financial planners, lifestyle coaches — and then this problem started to go away. So the point here is it's not about AI being good or bad. It's not about change being good or bad. It's about the controllability and the amount of change and the speed at which it happens. In other words, even if you could just fast forward to AI being some kind of utopia, that amount of change would still be massively disruptive to society at scale and cause a lot of ill health to a lot of people around the world.

So it's not really about whether AI is good or bad or having that argument. It's really just, like I said, about the amount of change. So let me introduce the idea of AI change management. There's a lot of research in the corporate world around change management. When a company goes through a particular change, let's say it's going to be Netflix. They're going to go from posting out DVDs to streaming. They bring in people to help with that change. And there's a lot of different change frameworks.

When change is managed, it's helped. It's more successful by about 600%, give or take. So that's management. I won't go into depth here, but I want to give you a flavour of the kind of things we're working on. Here's a very simplified version of an AI-specific change management model. That example I gave with Netflix is really a one-off change from A to B. And in the AI world, we're not seeing that simple one-off change and then steady state again. It's change after change. So we need models to be updated — this cyclic, ongoing change after change, which we are seeing and we're going to keep seeing for the next decade or so.

So here's a very simplified version. Let me take you through it. Let's use an example. This is kind of scalable, so we can think about this for at home, or for a business, or in fact for a whole nation. But it starts at the top here and moves around this cycle. And you start with actionable inspiration. So think of this as: let's say at home, why would you adopt AI? Why would you bring it into your family? The same applies for business. Why would you adopt AI?

So this is kind of the fuel — the understanding of why this change is happening, or why you want to personally do it, or why your company or institute might want to do it. And when we're going through change, people always get to a point where there is resistance, but they don't want to do it. They turn away from it. And so this fuel can help with those moments of resistance to the change. It was important to start there. Then we're going to go through a discovery period — educating, understanding what AI could do, how you might want to implement or what it can and can't do.

Then we're going to map out, sort of build, a roadmap for AI integration. And of course that's going to be very different whether it's a family at home. Are you going to use it for planning your shopping, going on holiday, for everything, every decision? And you can think of similar things in a business, but you're going to map out this integration roadmap. Then you're going to test it out, see what works, what doesn't work. Being very agile here: things that don't work, you're getting rid of.

Things that really work well, hanging on to. And being agile and adjusting as you go. So that's really analysing how that system and how that process works. And then finally here, this last one is really mobilising. And this is a really important thing that often gets left out of these change processes. So whether it's a business or a family or a whole country, the idea here is to socialise, to talk about the experience, what's working, what's not working. This does a few things. It brings down the uncertainty.

It socialises change. By simply socialising change and talking to people about it, it makes it easier. It takes away some of that sting. And whether that's in person or it's through some digital option, it's very important when you're bringing AI into a company to let everyone talk about that experience and learn from each other's successes and mistakes. So that's this kind of cycle, and it's a very simplified version I'm just showing you here. Then the centre here, this big thing — think of it as a sort of centre, this sort of spoke-wheel model.

And this is really around thinking about the balance of different things in terms of mental health, stress, anxiety, well-being, this idea of growth mindset and resistance to change. So when it comes to AI, the kind of change that's happening and we're going to keep seeing happening, not only is continuous and cycle after cycle of change, the mental health, the emotions involved are much stronger than with typical change frameworks. So we do need to put more resources there and do need to support people that are going through this, because some people are going

to respond with anxiety — and they do, and they are — and other people are not. So we do have to pay attention to this. So that's a flavour of the kinds of things we're working on in terms of AI change management models, and I'm a fan of using that as a lens, not just for a family or a business or an institute, but for a whole country like Australia. So here's the question. What would a change framework along these kinds of lines look like for a whole country?

If businesses use change frameworks just for simple one-off changes, a country like Australia is going to go through all these changes over the next decade or two. Then the data really supports the idea that we are going to need help and frameworks to help people through this change without really being hit hard with stress, anxiety, mental health challenges. And when I say mental health, you always get biological and physical health challenges along with that. So we need these AI-specific change frameworks at national scales. This is something that's been missing from governments around the world, missing from most of these conferences and missing from the big Geneva-based AI things that are happening as we speak, I think.

So we've been working on this as well, and I've been talking to different people at different levels of the federal government. What would a change capacity look like, and how could you support the whole country through this? And just to throw out a little teaser, what that might the flavour of some of that. Most people in the room are probably old enough to remember slip, slop, slap, which was a campaign that ran in Australia, which is sort of to protect us from the sun, wearing hats and sunscreen.

And it was a great jingle that went along with that. So imagine running a national campaign about AI. What is AI? What is it not? How is it going to change your life? And that's running it everywhere, across all media. And it brings down the level of uncertainty. It reduces the fear that people are feeling in the workplace. And it sort of solves some of the problem as well, that there's this big inequality in terms of what people actually know about AI in the country. Some people know a lot and use it all day, every day.

Other people know nothing about it still. So we still see that inequality there. So the other thing we're working on — I think you want to package as part of that — is psychological toolkits that are scalable, that you could get out to every person in the country. So what kind of things am I talking about? I'm going to go into some stuff around uncertainty in a minute. But we know a lot in psychology and neuroscience about how to deal with stress, how to control your physiology, how to learn to know when you are in a stress state, and how to bring that down in relatively short periods of time.

But these skills are not widespread. So here's a question: how could you update people's everyday habits — think cleaning your teeth — to incorporate some of these things? Because the skills in terms of controlling our physiology and mental health that got us where we are today aren't the skills we're going to need for the next decade or two of radical change and disruption from AI. So as part of these change frameworks, we need to start scaling up these kinds of things and building them in ways that are very easy to use, and then everyone can adopt them.

So let me just say a few things about why many very smart, very capable people in government and businesses still aren't grasping some of the impact that's coming when it comes to AI. So this is the part of the talk where I'd normally give the example of — I'll do it, but everyone in this room who understands exponential change — anyone not understand exponential change? A couple of people. So let's say I take a piece of paper and I fold it once or twice and the paper gets a little bit thicker.

I folded a few more times. I keep folding that paper, and when I folded about 42 times, that piece of paper now goes from the Earth to the moon. And that's because I'm doubling the thickness each time I fold it. Once something gets big and you keep on doubling, it gets really, really big really quickly. And that's a simple way to explain exponential change. And that's what we're seeing with AI at the moment. And yes, there are bottlenecks with human adaptation and how we integrate into businesses.

But the AI frontier models are very exponential, and some would say hyper exponential. Then when it comes to our understanding of exponential change, we have a blind spot for it. We have a cognitive bias for linear change. And when you ask most people to predict what's going to happen in the future, they predict out in a pretty straight line. They take what's happening today. They may increase it by 10%, and then picture that for next year and the year after. And so you get these very different trajectories where things are sort of jumping up like this.

And most business and political predictions run in this straight line. And so this can lead to some sort of — yes — surprise, but some other things as well, where we're not prepared for what's happening next year, the year after. And these are the kinds of changes that you need, time lags you need preparation for. And when things are happening, changing exponentially, the idea of being on time with them when they're moving so quickly is kind of off the table. You either have to choose between too early or too late.

The idea of getting it just right is not really going to happen. So the other thing that we're seeing now is that some of the capabilities — not necessarily with the frontier models, but including those — are changing lots of other fields. So we're seeing effects in medicine and material science and military applications, and we can go on and on. And so AI is kind of this thing that's lifting the tide, and all the boats are being lifted by that. So we're seeing lots of different exponential functions emerging

in all different fields at the moment. And when this is happening, it gets really, really hard to predict the future. And that's where we bring to this elephant in the room, this uncertainty. So we know from a lot of neuroscience this is linked to change as well. That uncertainty absolutely puts us — in fact, all primates and pretty much all mammals and animals — into this state of fear, this anxiety, this fight-or-flight state. So simply not knowing the future is like a fear stimulus. So apologies if you have a snake or spider phobia, but I want you to think about uncertainty just like that.

That not knowing the future is like a venomous snake slithering past, or a spider crawling past. It activates the limbic system, the amygdala, these fear circuits of the brain. And in some people it's quite strong. Other people not so much. And there are big individual differences here. And indeed, if you're very sensitive to uncertainty, that's kind of this precursor to all different types of anxiety disorders. So we are in this fundamentally less predictable world now that there's structurally more uncertainty. So it is becoming the new normal.

We have to find ways to adapt to this. And when I talk to businesses and employees and indeed students, they're feeling this. And I think this is an interesting way to frame it. This is this pandemic of uncertainty that's spreading. Students are worried about their future. Parents are worried about their students. Future employees are worried about their jobs. And another way to say that would be: you don't need to lose your job to AI, to it, to really affect your life, your quality of life, and your family's quality of life.

All you need is the uncertainty, because it pushes you into a stressful state. Innovation takes a hit. Your health takes a hit. You're more likely to get all kinds of diseases. So uncertainty is quite a scary thing. But we do know how to deal with it to some degree. And unlike the digital revolution and back in the day, when it comes to uncertainty, it's really more a human problem. So a couple of quick things we can do. First is to embrace the situation. This structural state that we're in — the world is very uncertain.

So we don't want to fight that. We can try and fight that. You can try and surround yourself with a sort of bubble of certainty, but it's not really going to work that well. You're going to burn yourself out. It's like trying to fight your way out of quicksand. It's not going to work. So the first step is really to embrace the fact that the world is structurally uncertain. Now, the second thing is to understand what's happening. And this is the simple thing I just talked about: that it's not your fault if you're feeling stressed or triggered by uncertainty.

It's simply the nature of having brains like we do. We get triggered by uncertainty, and to understand that, so no one feels at fault, no one feels it's some kind of intelligence fault or personality issue or anything else. It's just the nature of having brains like we do. Another thing I like to talk about is cognitive reframing, finding ways of exposing ourselves to uncertainty every day and reframing that, trying to see it as something that's full of opportunity rather than something scary. So changing that story — and it's something that takes practice over and over.

But if you do this more and more, it gets easier and easier. And there's data on this. So there are examples of roller coaster rides, horror films, fancy restaurants where you have no idea what the chefs are going to bring out. And people love these types of uncertainty. So you can practice reframing the uncertainty as something like this. So this is one example. When I say psychological toolkits we can scale up — and that is absolutely needed in society at the moment — is something like this: new tools to dealing with uncertainty, because it is the new normal.

So how can we embed that into these large-scale, society-based change frameworks that can help people everywhere? So I'm not going to have time to go into all the other stuff. We're looking at the moment in terms of the psychological effects of deepfakes or AI and human relationships, how that's changing a whole generation of young people, or the cognitive offloading, the brain fry, all these other really concerning and deep impacts in terms of neuroscience and psychology the AI is having. But I think you can sum up, what's the idea of change and uncertainty [inaudible] as this overarching challenge right now?

And if we can find ways to ease that stress state that people are in, all the other challenges will be much easier. So that's why I chose to talk about these two specific things today. So I think if we do pay attention to the human side of everything that's happening — this resilience — how can we build more societal resilience into the next decade or two? Yes, there's a lot of work to be done on the tech side on lobbying the frontier companies and making that technology more humane.

But there's the other side of that coin. How can we prepare people everywhere to be ready for disruption, to be ready for AI and all the things that it's doing to us and to our society? And I think that's a very important piece of work that is not necessarily the role of technologists or economists or AI scientists, but it's the role of behavioural neuroscientists and psychologists. And I really wish that more of my colleagues would jump onto this bandwagon and really start building up these tools and outputting things into society that can really help.

So that's now my mission. I have my own lab, but I'm also part of the UNSW AI Institute, and we're working to try and build out these tools, both at a state level and at a federal level, and embedding some kind of transition functions into government that can take on some of these roles. So I think we can absolutely thrive for the next decade or two if we really pay attention to this human side of what's happening. So I will leave it there. And I thank you very much.

Greg Sadler

We'll have time for one quick question.

Audience question

Thank you for that talk. That was beautiful. I really like the slip, slop, slap idea. How do you communicate to the public and help? What's the role of government in bridging that? I thought that was a good idea. So the question that I have in the app is someone's talking about change management where the destination is uncertain itself. So if you're restructuring and you're moving from position A to position B, maybe you can think about I'm changing to this thing. Whereas if you're just losing your job, is change management different?

Joel Pearson

So running if you can it's going to be different. But when I talk to people in government, that's the thing they say: well, we don't want to make predictions about the future, so it's too early to do anything yet. But I think we do know there's going to be change. We don't know exactly what — we can make predictions about the types of change, but to know there's going to be a lot of change, to know that people don't handle change well. We know we need some functions and frameworks and infrastructure, probably from the government top down, to help people through this change.

So you can do a lot to help people without having to have precise predictions or accurate predictions about what this particular flavour of change is going to be. You don't even know whether the change is going to be good or bad. As long as we know there's going to be a lot of change and it's going to happen relatively quickly, then we know we need something like this. So I don't think you need to make accurate predictions. You just need to know that things are going to change.

I think they're going to change.

Greg Sadler

We are on time. But sneak in one more question.

Audience question

Thanks for a great talk. In addition to a slip, slop, slap style campaign, if you travel into the future five years and all of your dreams had come true on this topic, what would we be seeing, do you think, in terms of how this change management?

Joel Pearson

How long do we have? We would have next 12 months. We would have limited all the downsides that we're seeing with AI at the moment. So everything I've mentioned, the stress, the anxiety, and including the more deeper individual verticals from deepfakes and conspiracy. So the other things that I haven't mentioned are that we can predict from uncertainty that we'd see these big political shifts. You predict that there's going to be a spike in conspiracy theories, for example. And so those have national security and political implications.

So we would have, by putting a lot of resources and the right expertise and all the knowledge we already know about the brain and about psychology to work in ways that are scalable, we can minimise all those downsides while leaving all the upside there. So really providing this bridge to get us over this uncharted territory. So we'd be living — it's a way to maximally get to some kind of AI utopia.

Greg Sadler

Great. Thank you so much for the talk. I think that was really beautiful. So thanks, Greg. Thank you everyone.