Moving from nudges to evidence-based system change

Recorded on February 18, 2026

Dr Bethany Jones, Acting Managing Director BETA

Good afternoon and welcome to today's event. My name is Bethany Jones and I'm the Acting Managing Director of BETA or the Behavioural Economics Team of the Australian Government. On behalf of us and the Australian Centre for Evaluation, welcome to today's special presentation by Professor David Halpern.

I'd like to begin by acknowledging the traditional custodians of the lands on which we meet today all across Australia. I'm joining you in Canberra on the lands of the Ngunnawal people, and I'd like to pay my respects to their elders, past and present. I also acknowledge and welcome Aboriginal and Torres Strait Islander people joining us today.

For those who are not familiar with BETA, we're the Australian Government's central behavioural science research team supporting evidence-based policymaking. Since 2016, BETA and other behavioural science teams across APS, many of whom are joining us today, have worked to put Australians at the heart of policy.

I'd also like to acknowledge our co-host today, the Australian Centre for Evaluation, or ACE. ACE was established to help improve the volume, quality and use of evaluation evidence across the APS. And we're lucky today to have Eleanor Williams here, ACE's Managing Director, and she'll run the question-and-answer session at the end of the presentation.

Today's presentation feels very timely. Some of you know that BETA is celebrating its 10-year anniversary this year and internally we've been reflecting on our journey to date. Our mission has always been clear, to put human behaviour at the heart of policy making and we and other behavioural science units do this by engaging with the public, to understand what drives and influences people's actions, how they make decisions, and how they interact with government. And these insights are then used to help design policy and programs that go with the grain of people's lives. But we don't stop there. Where possible, we use a range of methodologies to evaluate the changes we recommend, to find out what works and why and for whom.

As part of all this work, we rely on the great research that has happened and continues to take place in the APS and across the world. Certainly, for us, we appreciate all the significant work done by BIT and other research units to help us understand how people interact with Australian Government services.

And with that, I'd like to welcome our special guest, Professor David Halpern. David is the President Emeritus of the Behavioural Insights Team, having previously been the team's CEO when it launched in 2010 through to 2023. He's recently been appointed the Director of the Downing Battcock Institute, a new initiative based at the Downing College in Cambridge. And is designed to strengthen bridges between academia and policy makers. David is also a visiting professor at King's College London, at El-Erian Institute in Cambridge and Imperial College London. Prior to BIT, David was the first Research Director of the Institute for Government and between 2001 and 2007 was the Chief Analyst at the Prime Minister's Strategy Unit. He was also appointed the What Works National Advisor in July 2013, a position David held through to 2022, through which he led efforts to improve the use of evidence across the UK. David was awarded a CBE in the New Year's Honours in January 2022 for public service in his role as What Works National Advisor.

So David brings a wealth of knowledge from both behavioural science and evaluation domains. I encourage you to take advantage of our time with Professor Halpern today and actively engage in the Q&A session, either in the room or using the chat function. So please join me in welcoming David.

Professor Halpern

Wonderful to be here. I should also pay my respects to elders past and present. It's really good to be. And also, for a particular kind of part of the extended family for BETA and ACE, as you'll appreciate, we've gone on many of the same battles and it's been incredibly, I've certainly personally appreciate it over the years to have, you know, colleagues in other places where we can collaborate with and also tell our ministers, look, they're doing it now, you know. So it's really good to be here. I'm going to focus on some relatively new stuff today and run past with you amongst friends and hopefully it’s useful, there's a lot of stuff I won't cover actually about some of the where next on evidence or whatever. Maybe we can talk about some of those things when we get to the questions. What I'm going to talk a bit about to those who don't know about, let me just say we're goig to cover the kind of three things, some 101 stuff about behaviour, in case you don't know already. I think most people will do.

In the room and online. And then we'll talk a little bit about essentially a kind of anomaly, why we got stuck in some areas, particularly economic policy. And then I'll talk about this notion of boosts in particular as a sort of supplement to some of what we've been doing. BIT has been around and there's folks, there's Alex here of course and Karen who's based here in Canberra, they're doing a fantastic work. I only do a couple of days a month now with BIT, but the team continues to be fantastic. Thank huge numbers of projects across the world. And one of our intents was always, I'm sure you share this, which is that by moving out from within British government, it was a way of us learning from other countries and learning from each other, right? And why can't we do that better in many domains, including sharing evidence. OK, and there's a previous visit and there's Karen and Alex. Look at that, looking very cool and beautiful. So yeah, and of course, Australia was one of the first places back in what was it 2012, when we started working in New South Wales at request of the Premier.

So it's a very deep and long-standing relationship. Yeah, very briefly, I'm now basically split roles and I spent quite a bit of time back at Cambridge. I used to say I was recovering academic and now I guess I'm a partly relapsed one, but still, you know, trying to build stronger bridges between this world of policy and academia.

and ideas and so one of the key things we'll be doing that and hopefully maybe some people over here will join in sometimes is to try and bring together key policy makers and some of the key experts ahead of the policy cycle to figure out what might work better.

Right so today as I said I'm going to do a brief overview some stuff about updating economics and then talking about this shift from nudging to boosting, hopefully do both those things. For those who don't know, most of you will know this very, very well, but at some speed we figured out some time ago that human beings, how they actually seem to make decisions and a key part of it is through fast systems. You know, Danny Kahneman, later getting the Nobel Prize, as we tend to think of us as making these sort of slow, reflective decisions, whereas actually a lot is done in this fast, intuitive, automatic kind of way, but is also often error-prone as well. So if we're building policy, why wouldn't we build it about actually how humans make decisions? Many examples, this still, I think, a favourite one for me. I can think of it as a...

We ask people how many calories in a burger and they give you an estimate. If we ask them how many calories in a burger, which has got some salad next to it, and magically, magically forget those laws of physics, some of those calories disappear. So there's less. We can sort of see why that is. People might be doing some kind of averaging, etcetera, but you can also see why it's potentially problematic.

But actually, that's how we make decisions and how we think. Behavioural insights increasingly here as well now have been used to so many policy areas from more like public health, actually, can we reformulate food to take the sugar out? Can we guide in lots of areas? And with respect to getting people pay their taxes and many, many other areas, there are also some areas that actually hasn't been used so much. And one of the nice things actually just being here, it just came out of a conversation about cohesion and conflict, which has been for a long time. I'm not unique, I felt like, well, it's great to get people to just use behavioural science to get people pay their tax on time, but you know, really big issues, like if we're falling out with each other in terms of conflicts and cohesion, that is a massive, extremely behavioural issue. I don't know how many trials ACE and BETA have got on it, but it's like it merits almost certainly more than we are doing.

But today I'm also particularly going to talk about economic policy. Those who don't know, and I remind it for a particular reason, it becomes obvious when we were doing this in 2010 in Downing St. we had the new incoming coalition government and this statement was made in the coalition agreement. Our government will find intelligent ways to encourage, support and enable people to make better choices for themselves. And this was partly the raison d'etre to help explain why we would do the Behavioural Insights Team, in order to do this. Rather than just passing a law and assuming that that would have the impact, it also was relevant politically in the sense that it was a government which was trying to say maybe we should reduce the volume of regulation and activity and trust our citizens. And so behavioural science also fitted in that model.

As many of you know, one of the roots we also pursued, which is very well continued as a tradition here, which has always felt that the secret sauce often in history would judge us for, which wasn't even if you didn't get about behavioural science, is just use the empiricism. And some of you know, the White House itself pivoted on the back of our approach to become more empirical with the bringing of Maya Shankar in over there too. I'm not going to dwell on it, but yup, if you don't know about it, if you haven't heard it, where have you been? But as we worked with many policymakers, we found ourselves trying to simplify the very simple heuristics you might think about in any given policy area. If you want people to do something, make it easy.

see attractive, social, think about social influences, and timely, when's the moment? And of course, in many policy areas, or indeed in policy practice, we actually haven't been doing these things and we still don't do them enough. We had a very, one of the early examples actually in New South Wales, a beautiful a little illustration it's still just a reminder, kind of part of the greatest it's this um enforcement order that people got you know traffic fines etc and if you don't pay you can end up losing your license um you know these documents written by very well-meaning people in government and lawyers which citizens have no idea what the hell you're saying. You know it's like what does it say? What is an enforcement order? What are you asking me? Why have you been sent this enforcement order? What's your main message, you know?

Rewriting it to be easy, etcetera. You know, Oh my God, what are you asking me? Oh yeah, it's an unpaid fine. Pay now, right? Would it make a difference? Of course, it turns out makes quite a big difference. People are much more likely to pay their fine by the due date.

My memory was you know, just for saving on the not having to print extra letters was more than the cost of the, you know, the whole trial. And literally thousands of people who didn't basically lose their driving license or have them suspended because they had paid the fine on time.

So these kinds of very simple examples help to break through, and of course since then we can use other techniques. An obvious one in this area is using AI to predict where people will look and then use that without having to run a trial. You know, can we continue to upgrade the methods? So if it works, as I said, for tax and some of these micro examples, one of the questions is where have we not used it enough?

Spreading across the world now, Faisal, who does this on a regular basis. He swears he's never going to do it again because it takes too much time. But this was his latest addition. More than 600 of these units, including here, of course. That's quite a long way from where there was a pretty lonely affair back in 2010. So that's something very much to be celebrated. So as I said, there are some areas in which we haven't gone far enough. The one area in recent years has bothered me more and more, which is that what about, you know, particularly when we call it behavioural economics, it's surprising that a lot of our economic policy hasn't changed that much.

And so my basic thesis in this, is going to for this next section, is to say, look how many, kind of the key assumptions, of economics have been quite firmly sort of knocked out or challenged, and yet the underlying practice doesn't seem to change much, which is at least pause for thought.

And this is what Richard, Richard Thaler here, of course, has called and he called long ago, “supposedly irrelevant factors”. That seem like actually empirically they turn out to be really important to economic behaviour, but they're supposed to be irrelevant with respect to models.

So Richard and I, a little over a year ago, actually we did a seminar in Downing St. not least for the new government and also for key economic advisers coming in, to sort of go through some of this, about well, what is the implication if you think through,

you know, if these things aren't true, where does it take you? And it also coincides, by the way, with Richard republishing or an updated version of Winner's Curse, which was itself, you know, 30 years, 30 years old work from his early essays, which were basically, these kinds of falsifications, in economics. And if you read the original text, the expectation was, what we were showing, this doesn't work, this doesn't work, and then we're gonna update our models. But actually, we haven't in many areas, including a lot of policy, updated those models. We did a bit of it and people wrung their hands in the wake of 2008. It's one of the things that helped to create BIT originally was that, in the wake of the economic shocks is that people were saying, well, our models didn't predict this, you know, that no one saw it coming, as Greenspan said. But then after a couple of years, basically even those big institutions essentially went back to the models anyway, right. So I want to quickly just go through a few examples, as long as you can read about our paper, fixing the holes in economics, and better theories for better growth. I'm just going to pull out a few examples and then we'll pivot back to what do we do about it.

So first one is just sentiment effects. So this is Bob Shiller, again, one of those good, original, you know, key figures who broke in to help introduce Behavioural science more generally. Talking about the fact, you know, epidemics of popular narratives, if we don't fully understand these changes in the economy and if we don't, basically it's a key driver in respect to economic behaviour.

The classic example everyone thinks of is things like tulip mania. So you get this.

this, you know, famous massive bubble, this huge spike in price, insane level, and then it crashes down. Or Australia's own version of it were in nickel prices a few years ago, quite a few years ago now, but it's slightly misleading actually, and these are sentiment driven things, but in this, you see it going back to zero, where it's actually a lot of sentiment effects don't return to 0, right? So what actually happens, there can often be a coordinating mechanism in the economy. So if you think about it, if I think, well, Eleanor is going to need a whole load more evaluation skills and then I'm going to start more FE colleges, produce more of them, etc. People will respond accordingly. You know, I'll build more of these machines or more PCs. So it essentially creates a form of synchronisation where it can drive a surge. Of course, it can also work in reverse, and if you've got reason to think the economy is going to fall apart, you might reduce investment and so on. And in fact, there's some evidence that those effects can be even larger because when people talk down an economy, particularly ministers, that's a pretty unusual thing to do. But oh, if you look at it empirically, our kind of crude estimate, if you look across the literature is maybe 20%, for example of US growth is pure sentiment based. Right. And often upwards, because it is doing some of the synchronisation and driving activity. Those are, those are real effects. There's a that's real growth that is occurring and yet it's not really featured. It doesn't feature in, you know, in a lot of our practice. You know, you might have an army of people in your Treasury or your business department, but you don't have an army of people thinking about the way in which we're affecting sentiment, you know, beyond a kind of art, turning into a science.

In some domains, oh, come on. There we are. Did it work? Yeah, so we see it in lots of areas, you know, as markets move according to fears and sentiment. Some areas it has been, and a famous example is with respect to central bankers. Very, very good studies on this you may know, but very, very formally studied now that the things, the character, the sentiment which is used by central bankers has impacts on the market reaction over and above the objective decisions that are being made.

And so if that's true there, it's true in lots of other domains. Interestingly, in the British government, famously with this new administration, talk very, very negatively about the economy, about how bad it was. And there's quite a lot of evidence that in fact had real effects with respect to the actual economic activity.

So the sentiment, that's pretty big. 20%, that's worth getting our bed for. A bit more abstract in some ways. In fact, one of Richard Thaler's favourite areas of, if you like, an anomaly. Often we talk about, you know, we've got more sophisticated and clever, you know, economists have, but at the same time, consumers remain decidedly human.

Please open the question of whose behaviour we're trying to model and this is particularly true around something like fungibility. So fungibility, in case you're not familiar with it, is means a dollar is a dollar. It doesn't matter where I received the dollar from, right? It's the same kind of thing and you know it's an amazing thing, and it makes models very simple.

It's just not true. There are lots of areas where the way in which you receive the money, including from government, affects how you spend it. We happen to have, again, I'm going to use a UK example, an amazingly dramatic example of it. When you get changes of governments, I know it's true here, I'm sure.

The new minister comes in and then you've got lots of they said, what should we do on XYZ? And you can go to the minister. Well, here's something we prepared earlier, Minister. We couldn't persuade the last one, but really we should do this. Things you should cut, etcetera. And one of those in Britain for a while has been what was called the winter fuel allowance, right. So people, it's not quite the same in Australia, because you know, why would you want heat ever? But in Britain, cold old Britain, this was introduced by Gordon Brown quite a while ago, where at winter, coming into winter, older people are given a chunk of extra money and it's called literally the Winter Fuel Allowance, right?

So from first principles, it's like, this is a bit, it's not a great policy, right? Because it doesn't, it's not progressive. We give it to rich and poor alike, right? It doesn't have other kinds of, you know, why wouldn't you just shift it on, to sort of, we spend a lot of money on this, just increase pensions and also we can do it in a progressive way, right?

And we could do it, if you like, at that time of year, fine, but just do it on pensions. So the Treasury has long thought this and, you know, and from an economic analysis it makes perfect sense. And so Rachel Reeves announced this, particularly because she felt under great pressure to say, you know, we've got to be tough on the budget. What am I going to kill? We'll use this as an example.

It was kind of, there was a massive backlash against it and I think the Treasury and indeed Rachel Reeves was a bit dumbfounded by why everyone's so angry about this, but a telltale against, to look at fungibility. So it turns out there was a beautiful study done quite a few years ago before in fact, on fuel prices, but a discontinuity design. I can say that and you guys know what I'm talking about. So you're looking at people who are just below the threshold and just above and you compare them and you say, well, how did, how did they, by the way, spend money on, on the fuel? So at that time, OK, it's a decade ago, in fact we found that, they would spend, people that kind of age, right? They spend about 3% of their fuel, just below the threshold, oh sorry, of their money on fuel, right. So if we give you a few extra hundred dollars, what would be a reasonable estimate how much you might spend it on? Well, that's a pretty good starting point, say 3%. But they don't, by using this discontinuity design, you can see, they would spend, if they were just (unclear) and they get these funds, 47% of this extra money. So it's a massive sort of what we call a jam jarring effect, right? It's got the label, that's what I spend my money on. So when you come along and you say, well, don't worry, we're going to put it in pensions, et cetera, et cetera. That makes perfect sense, but it doesn't make sense for humans, because we jam jarred it, right? It's non fungible. This is a massive effect. A key question of course, can you use a jam jarring for also positive effects like with respect to investment or saving? The answer could be yes.

Preference inconsistency. Massive issue, we’ve known about for a long time about humans seem to think, oh, if only I was over there or had that car or that cart, I'd be happier or even mis predict their own utility in short time frames. It's a real problem, right? Because we say people are utility maximisers, but which utility are they maximising since they seem to be inconsistent. Lots of famous examples. One many of you recognize in BETA, an old study now, but still beautiful with Danish workers. You win this competition. Congratulations. Here's your prize. You can either have. We can deliver it next week. Just tell us what you want to have. You want fruit or do you want chocolate?

You might be sceptical about this, it was Danish workers, but remember a clear majority, around 3/4 choose fruit for being delivered. A week later the guy turns up, says here's your prize. I'm really sorry. One thing, we lost your paperwork. What is it you said you wanted? Now three quarters of the workers now say chocolate. And of course even affects what time of day. If you get them before or after lunch, it affects as to whether they say chocolate or fruit. So we're inconsistent. That's a really serious issue with respect to utility maximizing.

We're irrationally cooperative, only irrational, of course, to mainly to economists, I say, as many of my best friends are. Again, another Nobel Prize winner, Elinor Ostrom, about this idea that, you know, humans wouldn't cooperate and we've got this problem of the commons and therefore we have to intervene and use markets, et cetera. Studied literally more than 5000 examples across the world. And it turns out humans are actually remarkably good often at cooperating around commons type problems. They're more cooperative than models predict, because most famously, they're more cooperative than the game type model, which we all love. This is Joe Henrich out there, ultimatum game type phenomenon. You know, here's the money, we'll split it. I get to decide how it's split, your only power is to reject or accept. And so, you know, well-trained economists, I should say, there's only two populations who conform to the model, which is what you're “supposed to do” in inverted commas. Even if Eleanor says she's going to keep 99 cents. I'm only going to get one cent. I'm supposed to still say yes because my utility has improved. But of course the vast majority of people will say thank you, but no thank you, we're going to get nothing. And in most populations, you know, people want at least a quarter, generally more like 40% before they say yes. You're not doing. That's the wrong answer. Wait a minute. So there's only two populations, of course, those being formally trained in economics. The longer you're in economics, the more likely you are to behave like an economist says you're supposed to. And then, interestingly, very rare tribes in the middle of nowhere. They've had almost no exposure to market economies, interestingly, also perform like economists say they're supposed to. But most of us who've been in market economies don't do that. We are, we are cooperative and we're, you know, nice to each other.

It's got a large scale thing. And again, we were talking about this actually a few times over this visit around social trust. So this is a question we've been asking for 50 years.

Generally speaking, do you think other people can be trusted? Yes or no? It's really important and it's also massively variable. Just a point for those who want to see. Here, by the way, is Australia. Australia is quite a high social trust country and it's a massive asset for you.

Andy Haldane and I did a paper about a year ago. He's an economist, revisiting some of this literature. It remains a really powerful predictor of economic growth at national level. It also affects your, actually even the individual level in a similar way what happens. So there are loads of countries, you know, if Australia's past, you know, 50% or around about 50, there are also countries where less than 10% of people would say most can be trusted. It has huge impacts, right? So it means information flow is more difficult. It means that your transaction costs are much, much higher. Or put it this way. In Australia you can just say, hey, can you do this? Will you fix my roof? Will it? Yeah, et cetera. You could do a handshake, relative to other countries where you'd have to have a contract. It also means it affects your labour market, for example. So you're trying to decide who should you employ for the job. If you're on a low trust equilibrium, you're going to say I'll employ my cousin.

Someone I know, but if you're in Australia, you're going to say I'll employ the best person for the job, right? That's really consequential. You can see how it drives growth. So it's really. Is it in your Treasury models? I don't know. You tell me. I can tell you it is not in our Treasury models or business. This is a more powerful variable predicting economic growth and levels of human capital.

And it's not in there, partly because I think it's not supposed to be there. But with humans, it should be. It has to be.

Going to drive forward, shrouding. So this is David Laibson. Actually, he hasn't got a Nobel Prize yet, but I'm sure it's a matter of time. These are attributes which are basically normally hidden, actively hidden, but they could be costlessly shared. Why does it matter is because we came to the conclusion in a number of studies that it's very, very hard in a lot of markets for consumers to actually tell the difference between good and bad. And this is a massive drag on the functioning of an economy.

Now you normally think in fact of responses, well, if there's a market failure like this, well, the market should solve it. But it turns out it often will not. And for lots of reasons.

This is actually a piece of work which we've done by the team BIT in Australia on greenwashing. These are sort of fictitious companies, but you're asking consumers, you know, how do you feel about this company? Do you think it's green? Do you think it's not? And you see sort of an advert. Basically, it just needs to put a bit of green or a tree in there and people think that it's green. Green washing, in other words, is phenomenally effective. Not only is it more phenomenally effective, it works best, that is the kind of efficacy. It's much even more effective with people who care about green and carbon issues, right? Which is like, what? So for people who really care about the planet, if there's a picture of a tree there, that's enough to make them think great, super. I'll stick with that. Why is that a problem? Imagine you're a company which is actually genuinely is green and you're going to a lot of effort with your green products. Well, how do you differentiate in the actual world of this is how consumers are swayed through these kinds of shrouding effects? The big, big issue and back of an envelope, our estimates where if you address some of these de-shrouding issues, you can roughly double productivity growth, right. I'm not in detail. I'm just trying to give you a glimpse of, we might be wrong, but if we're even close on the order of magnitude of some of these effects, not only is it, why haven’t we updated our model? It takes us to some really powerful policy levers. And my suggestion is we're still not using them. We're still stuck, weirdly stuck in an overly Chicago model, which we haven't updated.

So yeah, this productivity, a classic example of it is the rise in zombie companies, these companies where, you know, I'm making widgets, but I'm not very efficient where Alex here is making, you know, he's much more efficient. So either his widgets are better or they're cheaper.

So why would somebody still be investing in my company? Why is people still buying my products, not his widgets? Key issues? Why are people still working for me? They should go down the road and work for Alex, right? But again, it makes much more sense when you realise that we're actually human beings and markets aren't clearing very well. Or another example is consumers.

It's true in lots of countries. So how many consumers actually pay for Choice and probably should pay for Choice? So for us, it's Which magazine, in the US it’s Consumer Reports. In all three countries, a lot less than 1% of the population will pay for a product like this.

It may not matter that much to you, as it happens for some people, it matters greatly, right? So if Eleanor decides whether she should get an AEG or a Miele washing machine, it probably won't be the end of it. If you're a low income individual, it's a really important decision as to whether should I pay a little bit more. Will it be more reliable? How would you know?

Right. Really crucial question. But also if you look at it on the other side, if I'm a company trying to produce a better product, the market isn't differentiating effectively, right? And so it's slowing the growth of your good companies and products. So I'm trying to say this looks like a massive issue and if you fix it.

You're basically getting high productivity and high growth.

Right. What does all this mean? Well, one of the themes that runs through this, because it was the Behavioural Insights Team, even though everyone said, you know, it would refer to us as the Nudge unit, including myself. So that's my fault too.

But if you remember the wording which was used for us, and I don't know if BETA. actually interesting had a very similar wording but was to support people to make better choices for themselves. Was one of the ideas of the political logic. And you can argue that one of the things about some of the interventions we've done, you know, the classic thing around the canteen or water enrolment for pension or whatever pensions. Is that we actually have been able to change choice architecture a number of ways and spectacularly guide people to what look like better choices, right? And by the way, even people who opt out of some of those things still say it was the right thing to do, right? But it's not obvious we made people better decision makers. That was the greater ambition. And so this is, I mean, sometimes put as a rival, but I would say a complementary agenda, which is are we doing interventions which also enhance people's capacity to make good decisions? And that also echoes if you think about some of that earlier work I just went through.

I mean, sometimes this is Sendhil Mullainathan. He says this image captures everything you need to know about behavioural science in one image. You know, this is what our human beings, we need to go and get some fitness. But as you can see, the way we get there is we put an escalator to get you to go to the gym and it's kind of funny, but it's like.

Yeah, but actually it sort of seems to be somewhat true. So you're using a kind of nudge, you're changing choice architecture to try and get people to go. I mean, obviously, why don't they just run up the steps because they're going to go and get some exercise, but it turns out the escalator helps.

It's quite a negative view in some ways, you might say, of human behaviour, but in principle we want to do more than that.

So in, you know, nudges, you know, anchors, defaults, social proof, etcetera, dark patterns, there's a whole fringe of these which we might call educative.

Educative nudges or nudge boosts, which, you know, help people learn in various kinds of ways. But there is some kind of fringe which is definitely not captured, but we would narrowly call a boost. I would say our team, BIT was certainly using a number of these, but probably didn't think about them especially overtly and we should do so more so.

Here’s an everyday example of it. We all recognise the happy family group of today having a good conversation and a meal together. I mean, the numbers on this are actually quite stunning and I know it's quite topical here. I mean, for to give you numbers just from the US would be pretty typical. You know, people have spent a lot of time already on their screens circa 2010 when BIT and others were being set up. But if you just go forward even a few years, it moves from about three years, sorry, three years, three hours a day to more than six hours a day driven by mobile phones, et cetera. We know this pretty well. We know that these devices have been built to be super addictive, right? We've been hacked in various kinds of ways, including use the scroll and so on and so on. Well, what do we do about this, right? How do we break the cycle? And so this is quite a well-rehearsed and particularly linked to Ralph Hertwig I should say, who's been a big champion of boosting.

One of his favourite examples. So if you haven't come across it, how do you prompt system two? So the everyday example of it, just to remind you, I'm sure you've all been there. You're sat on the sofa you think, oh, you know, I'll just take my phone out. The example in Britain doesn't really work over here. In Britain, everybody checks their weather all the time, right? See what the weather is. What's the point here? I know. But anyway, you imagine rainy countries, changeable weather. You pick up your phone or something else. You're going to check the news or whatever. And then half an hour later, you put your phone down. It's like, wait, what? What did I mean to do? Oh, I didn't actually even do the thing I meant to do. All right.

We’re sort of caught in that way. We recognise that cycle. So Onesec was developed by a group of psychologists and you can put it on your phone, download it and when you click on it to, you know, look at, you can choose what it is, but you know, it might be Facebook or WhatsApp and it tells you.

How many, how much time or how many attempts you had yesterday to use the platform? And then really annoyingly, does anyone have Onesec in the room? Anyway, but it then forces you to wait by default 6 seconds, which by the way, feels like a really, really long time. It's really quite annoying. You kind of think change it to a slightly shorter time set. Ohh It makes a massive difference. So both in adults and by the way teenagers, it more than halves the use of social media and indeed anything else you put underneath Onesec within a month and it's kind of sticky. So all that time back, what did it take? You can see what you're doing, you're breaking the cycle of the automatic fast brain, you've been trained into a habit now of, oh crap, what am I doing? Oh, maybe I should actually talk to my wife who's sitting next to me or whatever, right? So we would call that a boost, right? It's an individual level boost to kind of enhance your capacity to make it. You're having to make a decision more actively.

There are lots of forms which are deeper, and I'll give you one. This actually from work in Chicago. You might know it from other areas, but this was the very sad case, actually, which drove the creation of what became the Chicago Crime Lab and later the Chicago Urban Lab in the university. So anyone who knows Chicago, you might.

Lovely city in lots of ways, but extraordinarily socially divided. So around the university where actually Richard Thaler is, is a beautiful, lovely place and a great medical school. And then you just go a few blocks over and it's pretty rough. So this young man, Amadou Cisse, was coming out of the library. He just, he’d done his Ph.D. Viva. He just was going to graduate a few weeks later. He comes out of library, book bag on his shoulder. It's just got a few books in it, has a little water bottle. He gets stopped by a couple of kids, in particularly young men, I should say.

Who basically must have asked for the bag. Who knows what happens exactly? He probably was confused and paused for a second. The young guy who was challenging him was a 17-year-old, picks out a gun and he shoots him and he dies.

Demetrius Warren, you can see. Now what is going on? What nudge could you introduce into this moment to make it less likely that Amadou would have got shot? It's pretty hard to think of one. The classic incentive issue is well, this young man will albeit four years later, gets a sentence for 120 years. You know, if it had been in similar cases, if a sentence had been 200 years, would that have changed the behaviour? And it's hard to think that it would do, because that's not what's driving it, right? It's basically an impetuous behaviour which is habit-based.

So this case inspires particularly what's known as a program called Becoming a Man. You may have heard of it. And a tiny bit of detail. What is it? How does it work? You'll see why I think it's a boost. We would call it a boost. This is a bit, the guy over here, by the way, he's not in the program. He's visiting, Obama. So you take these young men who are basically drifting into increasingly violent activities and there's a kind of ex-gang guy who leads these activities. So what he does. One of the early exercises, like it's a really powerful illustration is that they get, they go into pairs, right? So we might say, Alex and Karen on the front here, they're told that they're in this group. So we give Alex the ball and we say to Karen, okay, here’s the exercise, you’ve got a minute to do it, see if you can get the ball from Alex, right? You can imagine how this plays out with these young people. You know, they're scrapping, punching, rolling around the floor, etc. By the way, they also then redo it, flip around the other way. But anyway, so you've done that thing, so you've done the exercise, etc. So calm down everyone. OK, let's have a quick discussion. So we then say.

To remind you, Alex had the ball. Karen was trying to get it. So we say to Karen, why didn't you just ask Alex for the ball? Almost never, almost never do it. OK, that's kind of interesting. So she says normally something like, well, I wouldn't do that because, you just think, often the phrase is he’d just think I was a punk. He's not going to respect me. So we said, well, let's just ask Alex. Alex, if she'd asked you for the ball, what would you've done? And then almost always they said, like, what's the stupid like ball? I don't, I don't care. I just would have given it to her, right?

But illustrates this point you're trying to draw out. And what's happened is these kids have grown up in conditions where, you know, if when they're 10, someone says, you know, give me your lunch on the way to school, they've learned that if they say yes, then the next day what will probably happen is someone will say, you know, give me your phone, give me your money. So they learn it's a tough environment. They better push back pretty hard or it’s not going to be great. But the problem is that then becomes over generalised. When they are threatened, they become angry. They react very, very violently. So essentially Becoming a Man is cognitive behavioural therapy. It's like when you're becoming angry, what's the alternative?

How else might I respond if I got something else on the menu, like I could just try and rip this out of Alex's hand or I could say Alex, can you just give me the ball right? And so, what are my alternatives? And in fact the kids, I think what I loved about it, they themselves nicknamed this Control Alt Shift, like for those of you, remember when your computer's not behaving well, like this earlier, you do that.

And then, you know, your badly behaved program, you can kind of break out of it. So it's like the mental equivalent of it. If you think it's like the Onesec, it's like, OK, I'm emotionally angry as a signal to me, just draw a breath. What else might I do? Beautiful program, 2 RCTS, I should say, associated with basically halving more than halving in the number of violence and arrests associated with these kids from doing it. So that is what we would definitely call individual level boost. Lots of those.

The key thing I want to get to here though, is one of the things about, often instead of our weird psychology is we're often drawn to the individual level intervention. And some of us know the I versus S slightly annoying paper, but you know there's an element of truth to it.

We tend to underestimate the system effects, and this is definitely also true with respect to how can we boost your capability. Oh, I haven't got that. Oh yeah, we've got it there the other way around. That's fine. Let me just use the example. I think it's a beautiful one. Actually, Nicholas Gruen actually drew my attention to it a number of years ago.

So this interesting thing. In the Second World War, the B 17s, if any of you know this example, an amazing piece of technology, fearsome really. But these crews would go into Germany or whatever, get shot at lots in different ways, amazing bravery, fly back, etcetera. One of the things that was happening a lot, was these guys would finally make it back, been to hell and back, land the plane, but they'd crash it. They'd crash it on landing. And particularly one of the very common things is they forget to put down, they didn't put down the landing gear. And like this wasn't a few, it was like hundreds of times this was happening, these planes were crashing. And so of course you do investigations. Like you would today, trying to figure out and it was pilot error. They forgot to put the wheels down. So what do you do? You try and retrain and get your, you know, your pilots to be, you know, more awake, you get more coffee, whatever. One of the things of course they notice is one of the famous early examples of ergonomics is that controls look really similar.

So basically what is happening is they thought they did put the wheels down, they reached the flap, they put it down, but in fact they adjusted something else. So the major insight was, well look, for example in this case you can see over there this is the suggested new design for the landing gear control, which looks a lot like…

A wheel, right? Make it look like a wheel. And then often what they in fact were pulling was the flap control. Well, let's make the flap look like a flap to differentiate them. And this massively reduces and largely eliminates the problem. So this is also a boost if you think about it, but it's in the context, it's in the environment, not just in the individual, right?

And there are lots of examples. Let me just go back to this other one. And one contrast, way of thinking about it is lots and lots of work about financial literacy. Can we boost people's financial literacy? And they and they work a little bit. You send someone to a training program, you can do stuff with kids. They work a bit, but not a lot, is the truth of it. But they do something. This is a very old trial BIT did with respect to Forex trading and we asked people to choose between different currency exchange rates and you when you get airport, it says, you know, no Commission, blah blah blah. But there's lots of other, you know, ways of concealing the costs.

So we basically dial up, you might call the transparency, so that it's easier to see what is the net cost, how much is, you know, what's that total fee for $1000. And guess what you get from less than half of people being able to choose the right option as the best deal. By making it slightly more transparent, easy to compare, you can boost it enormously. One of the shocking things at the end, however, was also, well, what happens if you do a mixed model? So some of them are really clear and some of them are old-fashioned. Well, people's performance really collapses again. So we can spend a lot of time trying to build people's financial literacy or we can improve the legibility and the environment and we can get a massive boost. And we see there's lots of examples. There's some beautiful ones in the years since. Quite often you can double people's performance by improving the legibility. So these also count as boosts. A lot of this is done deliberately.

One of the arguments, it's just information asymmetry. No, it's a lot of it is quite actively done. Here's an example in Europe, which I think is kind of almost funny but shocking. So if you've got electrical goods and you buy them in Europe, to make sure they're safe, which must be equivalent here in Australia. But and so you have this thing you see on a product on a plug, the CE emblem that's to be mapped. Conformité Européenne. The Chinese have, it turns out now, a very similar symbol, you might think suspiciously similar, which happens to mean “China Export”, which can be stamped on. I mean, you don't have to be like a PhD to realize what's going on here with respect to essentially active confuse-ology.

So how can we help the consumers in this? A great local example, one simple thing is just on price, right, with Victor Dominello work there and some of the BIT work more recently is like just improve the transparency. So enable consumers like with respect to fuel to be able to see ideally in advance, you know what's the relative price. So that is a simple example of de-shrouding right as important is to also de-shroud on quality. So for fuel it might not matter. So if you're going off to New York trying to decide which you know and come to Canberra, I don't know very well. There's lots of guides you can turn to which will give you some indications about quality. I mean it's kind of in some ways incredibly simple, but you think about the power of it, particularly. I mean, Mike Luca was early in his work on Yelp in the US where it gave us estimates, causal estimates of the impact. So to remind you that know the study it was that he found that, we could talk about how it works, but for one extra star on a on a five star rating per restaurant, it causally boosts the turnover by 5 to 9% in the year following. That might not sound a lot, but it's actually huge if you think about it. Also more detailed work since, there's lots of other aspects to it. What it does is it actually has twice the effect on boosting the good restaurants than it does on compressing the bad ones. You see what I mean? Can you even picture that? So what it's doing is it's growing disproportionately the good restaurants. And of course that means the overall sector is getting bigger as well because you're becoming more confident, you can go out and you can buy good food, the whole sector becomes larger. That's a pretty neat thing to do and if you can do it across the whole economy. So we've used some of those models to try and work out, well, if you do the equivalent of that de-shrouding elsewhere, think about what you're doing, you're boosting the growth of your better players, you're suppressing the weaker ones and you're accelerating that pattern. It also has a nice twist, which is it works disproportionately on the smaller outlets, which is a key detail.

Because you're trying to drive innovation, and it makes sense if you think about it. If it's a big chain like McDonald's, you know what you're going to get anyway. If it's a small outlet, you're not sure, there's much more uncertainty. You're that much more influenced with respect to this extra information. But that's exactly what you want to do if you're trying to drive innovation and productivity growth.

It's true in lots of other areas. So this is getting to a topical issue over here, and the whole world is watching Australia right now, as you probably know. But this is something from a decade ago, we did a smaller version of it in the US, Australian, in Australia with kids too, which is asking 15-year-old kids, how do different forms of social media affect you? So here's a simple contrast. This is again the responses of lots of 15-year-olds, one and a half thousand of them and you can see YouTube. This is on the negative side is here. It's bad because you scroll, it's bad for your sleep. But on lots of things, community building, self-expression, awareness, it's lots of positives.

But if we look across there, that's Instagram at the same period. All these negatives, you know, body image, sleep there as well, anxiety, depression, loneliness, bullying, fear of missing out. I think it's really bad. Basically, YouTube keeps you awake at night. But it doesn't make you feel like utter crap as a 15-year-old girl in the way that Instagram did. So this is a very powerful example of de-shrouding right, is you enable consumers to be able to see this difference and a key detail, we might find it means a parent gives the kid the phone, smartphone, blah blah blah. Let's agree YouTube's fine, but let's stay away from Instagram. You then make it Instagram’s problem. This is a key thing, right?

So you're effectively constructing a race and we could do that in a lot of areas. In fact, we should still do it in this area. So you can say, well, this is a relative rankings, you know, in the same way that we don't make cars, but we are able to put our rankings around you know, fuel economy and safety, which demonstrably affect consumer behaviour, even small numbers, which then reshapes the market, right? That counts as a boost, right? It absolutely should do. One of the key questions now we're all wrestling with as I kind of draw to a close is AI, right?

There is AI going to be the ultimate boost, that was called Bicycle of the Mind by Steve Jobs for computer, when we use this phrase. And in some ways it's got the potential to be that, right? It's an incredible aid to be able to answer any kind of question, blah, blah, blah, blah, give us advice. But it's absolutely subject to the same issues and forces, right? On what basis? If there was, in an interview you read it a little while ago with Altman in the FT, one of those short things where the interview you have over meal, and he chose to do it in his own kitchen. So there was loads of obsessive, you know, who cares what he said, but what is he using? What kind of knives is he?

Like in the background, there's a coffee machine, so loads of people were analysing what coffee machine it was. You know, is it good or bad? The consensus seemed to be it's a really expensive coffee machine, but actually it's not that great. However, guess what? It was the one most recommended by ChatGPT at that time.

Now that might be innocent right now, but pretty rapidly you're going to have the question of, well, on what basis will it make a recommendation when you say what coffee maker should I get? We know the existing platforms are completely distorted at the moment by paid for advertising. So to give a real example, Amazon, right? Amazon not only Amazon used to be a beautiful, lovely company, enables you to do delivery etcetera for lots of small companies, could get to new markets.

Over time, of course, the charges go up and up and up. But equally important, Amazon also charges companies for their position in the ranking when you do your search, right? And you are not going to go through more than a few pages. So not only they're taking more money or whatever, it's making the search much worse.

Because it's no longer driven by quality. So why is this relevant? It’s in the same way you we should be in the business of designing the races so that these become boosts, right? And that they are actually enhancing our decision making rather than detracting from it.

We were behind the curve on social media. We better get going on AI pretty damn soon, otherwise it's not going to end up where we want it to. And one thing on that. So figures like Mustafa Suleyman, who was at DeepMind now at Microsoft, he's talked quite a lot about this recently about we can try forward the technology, but.

We better figure out is it enhancing our well-being, right? And then who is judging that? OK, for those who know that behavioural science and one of the origins of BIT is in the early days for Prime Minister Cameron, not only we're doing big society, but we're doing well-being as well, and that fell by the wayside, which is too bad. But one of the things, again the behavioural scientists amongst you will know,

is that we're actually not very good at predicting what drives our well-being. We systematically are prone to error. So Australians trying to decide, shall I buy a bigger house, you know, further commute, etc. It looks like people systematically mispredict. So they think the bigger house will make them happy. It turns out they'd be better off in a smaller place with a smaller commute in terms of the wellbeing literature. And what kind of things in there? Some of the kind of, as you expect, like health problems on the negative side, this is basically a sort of summary regression across, particularly using longitudinal data. Unemployment, you know, we sort of kind of know that now, but unemployment effects are so much more damaging than the income loss. Because of connection and all these other things, by the way, what will AI do? You know, divorce, not great for you, blah, blah, blah, blah. Being a woman is a pretty good idea. It turns out in your well-being age effects, etcetera. But look at the top of it, by the way. One of the key ones is this massive variable about social support, which is basically a question. You know, is there someone you can turn to if you need help, right? Or is there someone who will support you if you need it? Massive, massive effects on our well-being. Arguably, we massively underplay it with respect to our policy activity, etc. So are we training AI to enhance our well-being on these kinds of variables? Is it going to say to us, like Onesec, you know you know David, you've been talking to me for half an hour really, for God's sake, go and get a life, go out and talk to someone else. Will it guide us towards wiser, better behaviour? It could do, but by default, that is definitely not where it's going to end up, right?

All right. In conclusion, I've tried to connect these things, which is this puzzle, even for those who are doing behavioural science, which is that we've had some amazing successes right across the world now in all kinds of domains. I think we haven't made anywhere near enough progress on some, including on mainstream economic policy and my thesis to you is because we're actually still quite stuck in many administrations on a sort of outdated, still very Chicago rooted model of what drives economic behaviour. So updating Keynes, you know, Keynes, you might know famously wrote about the fact that economists in government were basically following long dead, you know, economists now of course, including Keynes himself, scribbling away. We should do something about that. I think a common thread through all of it. This is just, you know, draw to a close.

There's a different view essentially on information and cooperation. The presumption is often certain kinds of failures. The market is optimally placed to do it. Don't misunderstand me. I'm really in favour of markets, but I'm in favour of how can we make markets work better right in terms of enhancing cooperation and also particularly do the winners win in terms of products and services. And at the heart of that, I think that's where you end up situating this notion of, you know, nudging to boosting. Are we boosting the capacity of individuals to make better decision-making, especially with respect to the field effects, right? And I guess I say to you, I think we are way off the pace, but we could do much, much more.

The final point I’ll just say is that one of the arguments about this, some of you again will know well. I mean Cass Sunstein put it very eloquently many times about, Richard, around libertarian paternalism is that we're trying to do things. It's choice enhancing or at least not choice removing so that if you don't really want the salad that's put for us, you can just walk on by and you can get those chips. They're still there, right? But when you look at some of this stuff, particularly around, you know, environment, but even that canteen, there's a question of, well, who decides, right? Who decides? Is it parents or someone else? Or is it someone like me or Alex or Karen or Eleanor making that decision? And the key thing is do we?

Do we have governance interventions which enable people to come together? And that's one of the ways we can handle the preference inconsistency. So this is a real example. Some of you might know we did some work with Meta a few years ago when Nick Clegg was there, which is that rather than say the Australian government deciding what the rules should be, this is particularly in this case with respect to how AI should behave online with people, you know, or Mark Zuckerberg, what happens if we got the users themselves across countries to say what the rules should be, right? And it turns out, I'm a big fan of the deliberative mechanisms. I think when you bring people together and ask them to reflect, they come up with pretty sensible ideas and they did on this one too. And this was literally done across multiple countries, multiple languages to figure out, well what would we choose together like as a governance boost and that is also a tool which we underutilize, which is to say, you know, we're not going to set the rules, but we're going to say in your governance there has to be a mechanism which enables the community or the users to make the decisions rather than in particularly which is sort of public good of this kind of consequence in terms of its choice architecture. So on that I'm going to end. I'm sorry for taking too long, but hopefully a useful stimulation for some debate and disagreement. But about, you know, we've done well, but boy, we've got lots more to do. Thank you. Right. How do we do this?

Eleanor Williams, ACE Managing Director

I'll jump in, I'll hand to you. Thank you so much. What a rich presentation. So much in there. Now I'm conscious we only have about 5 minutes for questions and answers, which means I'm not going to get to all, in the chat, it’s been sort of rolling in and we've got some real crackers here. So I thought I might just pick one or two that particularly stood out to me and particularly probably go to the heart of your expertise. So one which came from Riley, is about, you mentioned the supposedly irrelevant factors and lack of updating theories in classical economics. Do you think there's an equivalent in behavioural economics? Is there something we often miss in our theories of behaviour and is there growing empirical evidence somewhere that we're not paying enough attention to?

Professor Halpern

Great question. There might be one main thing, always look out for where we get stuck. One of the areas I think is more and more effective, particularly for policy makers, what's called the illusion of explanatory depth, which is that things that are familiar to us, we think we understand very well. That's true. I think in lots of areas of policy, it almost certainly is going to be happening to us in other domains. So we should be suspicious of our own, where we're over clear about it and you'll know one issue particularly, Richard has also championed this, is so-called permission bias where especially where behavioural science gets trapped within a narrow range of interventions and we're not allowed to go further. Now sometimes that is because that's what the brief that comes in, but sometimes it seems to me we are imposing it on ourselves. We're not being expansive enough about what our range of, you know, options are we exploring, right.

Eleanor Williams ACE Managing Director 

Amazing. I'm going to give you one more question and after that I've got an idea about how we might wrap up the remainder of them. But we did have one question come in just about your new role. Obviously it’s quite new, but just that question of do you have ideas about how we might bring together policymakers and academics more effectively and really bring academia into the policy-making process.

Professor Halpern

Yeah. OK. Thank you for that prompt. I mean, look, one of the common threads, right, is fixation. So once we think we know something, you know, if you like kind of classic literature around people get stuck on a solution and I think that happens often in policy, we think we literally can't see these other possibilities and I don't know if BETA feels this, but we do one of the key roles of BIT in those early days is and still I think, to function as a skunk work right to expand your possibility. So expose what other ideas are out there. I think you have to do this early in the policy cycle. That's the intent and the design is to get people before we become too rigid and the policy is fixed. Are there some other options which we haven't considered? So maybe I'm wrong, but I feel like I'm naively optimistic. But if we we'll find out, right, I'm an empiricist by bringing together policy makers early on in the cycle, look at what other countries and places are doing which is different. And you often see that, right? If you go to other countries, you see that they're doing something which we haven't even considered.

Right. Is it opens up that sort of possibility mind set and it will kind of create, you know, other options. And in particular, I think behavioural science can have a particular win is, you know, sometimes is it where the conventional solution we can see a possibility, but it's unbelievably expensive or difficult. And you are you sure there's not another alternative, right? Are you sure there's not another possibility of tunnelling through to find a possibility? And at the best, I mean, that's where you're, you know, the real policy magic and genius comes is, I mean, sometimes it's marginal, but it can be also, it turns out there is a way of doing this, Prime Minister, which we haven't explored. We're not sure if it worked, but Eleanor's going to go and test it and see if it does work and then we'll know. So I think it's trying to crack that fixation. Are we sure there's not some other option? Let's going to go the extra mile.

Eleanor Williams ACE Managing Director 

Wonderful. Well, thank you so much. Now I know we've got a couple of 100 people online and I know we didn't get to all your questions. So I'm going to suggest I'll walk David through those questions afterwards. And if there are some concrete answers, I'll see if we can maybe look back with a little bit of a debrief on this session. But thank you to everyone who joined. We did also want to do a little plug for Um.

Anyone who, if you do wanna hear more about what BETA's up to, we've got the website up there and same for the Australian Centre for Evaluation. Thank you so much for making the time to attend today and please do continue to come along to BETA and ACE events as we progress into 2026. Thanks so much.