---
id: "workshop:using-behavioral-science-to-revolutionize-policy-making-and-service-delivery"
type: "InnovateUS workshop"
title: "Using Behavioral Science to Revolutionize Policy-Making and Service Delivery"
url: "https://innovate-us.org/using-behavioral-science-to-revolutionize-policy-making-and-service-delivery"
date: "2024-05-16"
duration: "90 minutes"
language: "English"
instructors: ["Leah Everist","Anna Keleher"]
recording: "https://www.youtube.com/watch?v=IJjgUWxPKvk"
transcript: "automatic, from the recording"
source: "InnovateUS (innovate-us.org)"
licence: "CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/), as stated on innovate-us.org"
---

# Using Behavioral Science to Revolutionize Policy-Making and Service Delivery

InnovateUS workshop · 16 May 2024 · 90 minutes

Join us for this 90-minute virtual workshop with Leah Everist and Anna Keleher of the Behavioural Insights Team where participants will explore the fundamental principles of behavioral science and uncover its potential to drive positive change across various sectors.

## Learning goals

- Receive an introduction to behavioral insights— the science behind why people think, feel, and act the way they do— with multiple applied case studies across sectors
- Explore the fundamental principles of behavioral science and uncover its potential to drive positive change across various sectors
- Dive deep into real-life case studies that illuminate how behavioral insights can revolutionize policy-making and service delivery using examples from a range of sectors including transportation, sustainability, health, and justice
- Gain valuable insights into leveraging behavioral science to enhance effectiveness, efficiency, and outcomes across state government roles
- **Watch the recording [here](https://us02web.zoom.us/rec/share/kyfJhtJsS26EnmEn3np54mE1OAgiyRl3KGvAKFWB52phbqbRtdpyaLdmLxPxMIj8.ww6RN6a_DaN14zSf)**

## Speakers

- **Leah Everist**, Senior Advisor, Behavioral Insights Team
- **Anna Keleher**, Senior Advisor, Behavioural Insights Team North America

Recording: https://www.youtube.com/watch?v=IJjgUWxPKvk

## Transcript

Automatic transcript of the recording: names and terms can be misheard, and the times can be a few seconds off. Each time links to that moment of the video. The transcript alone: https://innovate-us.org/using-behavioral-science-to-revolutionize-policy-making-and-service-delivery.txt (plain text) and https://innovate-us.org/using-behavioral-science-to-revolutionize-policy-making-and-service-delivery.json (every segment with its start and end second).

**Unknown speaker**

[0:00](https://youtu.be/IJjgUWxPKvk?t=0) Leah, over to you.

**Leah Everist**

[0:05](https://youtu.be/IJjgUWxPKvk?t=5) Thank you so much, Hannah. And just to say, kind of on the back of that introduction, Hannah and I have both worked at the Behavioral Insights team for going on five years. And so, even though a lot of our work is focused in the areas you just heard from Hannah, I lead our portfolio of work with states, and my background's in public health, so I work on a lot of health projects, and Hannah leads a lot of our sustainability work right now. We have both worked on all kinds of projects and challenges kind of across sectors, so please feel free to ask us about Behavioral Insights applied to anything under the sun, because we have a lot of stories we would love to share. Before we fully dive in, just so that I have a sense of who is in the room, go ahead and put a 1 in the chat if you have heard about Behavioral Insights before, or a 2 in the chat if this is a new topic area for you. Okay, great. So, I'm seeing a lot of 1s, a couple of 2s, a couple of 1.5s. I love the creativity. I would never try to put anyone in a box, but it sounds like we've got kind of a range of exposure here, which is always fun.

[1:25](https://youtu.be/IJjgUWxPKvk?t=85) In terms of the—let's see if I can get my slides moving here. Okay, here we go. In terms of the next hour or so and kind of how we want to spend our time, we're going to start with an introduction to the Behavioral Insights team. You already heard a little bit about us from Hannah, and behavioral science, behavioral insights in general, and then we'll move into four different policy area-specific case studies, so stories about how we've applied behavioral science in the past to real-world challenges, and then, as Hannah said, we have some time saved at the end for a discussion, but again, feel free to ask questions throughout. You can drop them in the chat, whatever works best for you. So, in terms of who the Behavioral Insights team is, in a nutshell, we are an organization that takes behavioral science and insights, research and evidence, and we apply it to real-world problems, and what we do pretty much falls into these three categories you see on the slide here.

[2:36](https://youtu.be/IJjgUWxPKvk?t=156) So, all the way on the left, there's behavioral science. That's really our core, our bread and butter, conducting research and designing interventions and solutions based on how humans behave in real-world contexts, and that's going to be the focus of what we talk about today and the examples that we're sharing. We also do a lot of evaluation, so because what we do has a lot of, you know, taking research from one area, tweaking it for a new context, applying it for a new population, it's really important to us to understand what works, and not just in kind of a binary sense of, do we see a causal connection between our intervention and our outcome of interest, but additionally, you know, how well did it work? For which populations did it work best? How can we improve upon this design when we think about scaling it up for a wider population? And then the last thing we do, which this session is a great example of, is capability building.

[3:39](https://youtu.be/IJjgUWxPKvk?t=219) So, supporting leaders and teams to apply behavioral insights to their everyday work. It's really exciting to us to see behavioral insights teams pop up in other organizations, because I think it just sort of reflects more people taking this approach or thinking about behavioral insights and human-centered design. Okay, so what is a behavioral insights approach? What is this whole behavioral insights, behavioral science thing? Before I tell you, I'm actually going to make you play a quick game with me that will help illustrate what behavioral insights are and how they work. So, in just a moment, I'm going to flip through a set of slides. It's imperative that everyone is on mute for this exercise, because what you're going to do is read the color of the font on the slide out loud. So, for example, on this slide, you would say white out loud to yourself. Okay, I'm going to put myself on mute so that no one has to hear me talking, because I'm going to do it at the same time as everyone, and in a moment, we will begin.

[4:58](https://youtu.be/IJjgUWxPKvk?t=298) Okay, I hope everyone was playing along here. In the chat, drop in a word or two about how you felt once we got to the last three words on the right there. It could be something about your reaction time, whether you accidentally said the wrong color and then you had to switch it, any kind of difference between the first set of words and the second set on the right. Yeah, so it took another second to figure the word out, slower reaction time, had to think more, pause, reading the words instead of following the directions and saying the color.

[6:00](https://youtu.be/IJjgUWxPKvk?t=360) Yeah, absolutely. Somewhat, we have a superhuman in here, no difference, able to follow directions incredibly well, got angry. Okay, so what we just experienced were two different styles of thinking, and we're going to refer to those styles throughout the presentation today as System 1, or fast thinking, and System 2, or slow thinking. So System 1 is our automatic system, our fast thinking. It's intuitive, and it's relatively effortless. So it operates with basically no effort required on your part, and the reason it's able to do that is because it's looking for the quickest patterns available. So in this case, in our game, it's looking for the word, because these are relatively short words, they're very familiar words. It's almost impossible to look at a short word that you know in a language that you know and not read it. So if you look at some of the other examples on here, two times two, it's kind of similar in the sense that, of course, you can do this problem very easily, but even further, it's so easy that you almost cannot not do the problem. Another example I sometimes use is, if you think about looking around a room and you see an exit sign, you don't really decide to read the word exit. You just, you read it the way that you would see the doorway. It's so intuitive and quick that you don't even have to try. So when System 1 runs into difficulty, it calls on System 2, on the other hand, to support the more detailed and specific processing.

[7:48](https://youtu.be/IJjgUWxPKvk?t=468) So this is what we had to do in the other example, or in the later three words, where we were telling our brains, okay, wait, don't read the word, say the color of the font. That's probably not, you know, the hardest thing that you've had to do all day, but I wasn't giving you a lot of time, and I also kind of tricked you into this pattern where the word was also the color. So your brain was trying to just take a shortcut, and then you had to really backpedal, call in System 2 to account for that. The other example here is driving. So especially if you're someone who drives every day and drives in similar patterns, for example, drives to work, drives to drop off your kids, then that's probably a behavior in the System 1 realm, fast thinking, you may not have to put very much effort into driving. But on the other hand, when you're learning to drive and you don't have these kind of automatic intuitive senses about where you are in the lane or how much time it takes for you to brake, things like that, then it's going to be a very deliberative exercise, or deliberate, rather, exercise, so it falls into the System 2 category. Or you can think about, you know, driving in inclement weather, driving a U-Haul, if that's something that you've never done before, where you have to really focus and pay a lot of attention. It requires a lot of sort of perception.

[9:18](https://youtu.be/IJjgUWxPKvk?t=558) The guy in the middle of the screen here, his name's Daniel Kahneman, and he won a Nobel Prize for Economics a few years ago for his work in behavioral economics, is one of the two people who developed this system. So the quote here is something that we will come back to a lot in this session, that environmental effects on behavior, so basically the things happening around you in the circumstances where you might perform a behavior, are a lot stronger than most people expect. And that has a lot to do with these patterns and shortcuts, especially the patterns and shortcuts associated with System 1. So a little bit more about System 1 and System 2. At this point, we sometimes get the question of, which one is better? You know, which one are we trying to get people to use System 1, or are we trying to get people to use System 2? And the answer is that they are both great. And that's because the division of labor between System 1 and System 2 is highly efficient. It minimizes effort, and it optimizes performance, because we give System 1 essentially as much work as possible, and it's able to act as this sort of cognitive autopilot that replaces more complex decision-making with what we call heuristics or shortcuts or simple rules of thumb. So it's looking for the kind of the path of least resistance. If you imagine you're creating a profile on an app, System 1 is the thing that's telling you to click the button in the lower right corner, because System 1 knows that that's where the kind of next page button is going to live, right? That's a pattern that it knows, and so you don't even need to read that button to know how to navigate that situation. System 1 uses all kinds of things as shortcuts. For example, if you're, let's say, you're at a train station and you can tell something funny is going on, you can't really tell what's happening, but then you see a bunch of people

[11:22](https://youtu.be/IJjgUWxPKvk?t=682) are moving in one direction, you might just move in that direction with all the other people, right? That's System 1 kind of saying, like, you're probably going to have to do what all these other people are doing. So again, kind of taking shortcuts. And that's really advantageous for us for a lot of reasons. If we didn't have System 1, we would burn through all of our energy thinking through the minutiae of every single decision that we make. That's probably something that everyone in this call has experienced. If you have a really long, stressful day where you had to make a lot of decisions, and then you get home and someone asks you what you want for dinner, you have this experience of, I just cannot make another decision. I'm so spent. That happens to people all the time. And that's because System 2 wears us out. But equally, System 2 is really important. And that's because it can be deployed, really, at any time to help us learn new things, even when they're challenging. System 2 is what we have to call in to help us exercise self-control and to make calculated decisions by evaluating options, weighing up the pros and cons of each. Another example I sometimes like to use is, imagine if I'm going over to a friend's house for dinner, and I want to bring them a bottle of wine. But I don't know very much about wine or the wine preferences of the person whose house I'm going to. I might basically use what we call a proxy or a substitute for how I choose that bottle of wine. Maybe I choose based on price, because that's just a lot easier for me to think about than the much more difficult question of what goes into a high-quality bottle of wine, what does high quality even mean, what might the taste of this person be, et cetera, et cetera. So I could employ System 2 and do a lot of research,

[13:24](https://youtu.be/IJjgUWxPKvk?t=804) but it would take me a lot more time and effort. And so in a lot of cases, that's the type of situation where I might just say, I don't want to buy the cheapest one, I don't want to buy the most expensive one, I'm kind of gauging quality based on what's somewhere in the middle. That's the kind of shortcut that System 1 allows. They both have their downsides, which I've touched on a little bit. On the System 1 side, as I said, the heuristics and the shortcuts that we get with System 1 serve us incredibly well a lot of the time. But because they're using these simplified rules and shortcuts, they can go wrong or lead to poor outcomes sometimes. So a minute ago, I used that example of when you're creating, let's say, logging into an app or something, and you're just clicking that button on the right. Well, the app might, because the app designers know that that's what System 1 is going to make you do, they might put something there that's a little bit sneaky that maybe gives them access to your data or allows them to send you notifications or something like that. So sometimes System 1, because it's acting on these patterns and it's so predictable, can lead us astray. It also means that we're prone to cognitive biases. So those are systematic errors in decision-making where we take predictable shortcuts. So let's say I'm talking to a friend who has a young child, and my friend says, do you think it's safe for my child to swim in the ocean?

[15:00](https://youtu.be/IJjgUWxPKvk?t=900) I'm worried about sharks. Well, I don't know anything about shark hazard statistics, so what might happen in my brain, without me even necessarily realizing, is that my brain substitutes that hard question, which I don't know very much about, for a much easier question, which is, do I know anyone who's been eaten by a shark? The answer is no, and so my response to my friend will be kind of unduly painted by that personal experience. That's what we call the affect-heuristic or availability bias, or not affect-heuristic, rather, availability bias, because that was my experience, so it's more likely to affect the advice that I offer my friend. System 2 also has its limitations. As I noted earlier, it's really labor-intensive in terms of what's going on in our brains, so we don't necessarily have sufficient mental processing power to make all of our decisions using this slow, deliberate system. We have limited cognitive bandwidth, and that means that we need System 1 to be kind of taking control some of the time.

[16:16](https://youtu.be/IJjgUWxPKvk?t=976) The other challenge that System 2 can run into is that if it's not aware of the biases of System 1, then it can't intervene, so if you don't realize that maybe a certain bias is at play, then System 2 can't cut in and take over. So, that's all the kind of theoretical side of things, but the question for the folks in this room is how this actually impacts important decisions, right? How does this play out in the real world? So, I want you to humor me for a moment with kind of a thought experiment here, where the year is 2040. Each person on this call is a state health commissioner, and unfortunately, the unthinkable is happening. So, your people, your top advisors are telling you there's a new disease that's infecting people in your state, and it has an extremely high fatality rate, and unfortunately, 600 people already have it. So, what's going to happen if we don't already have it? But then one of your most trusted advisors rushes into the room and says, I have some good news. We went through our seed stock for medicines and immunizations, and there are a couple of promising treatment options. So, here's what they are. There's the first option, Program A, which is going to for sure save 200 of these 600 people who will otherwise die. Option B is a little bit more of a gamble. There's a one-third chance that it will save 600 people, but there's also a two-thirds chance that none of the 600 people will survive and they'll all die. So, before I continue, I'd love to see in the chat which one would you choose. Remember, you're the health commissioner, and the governor has said it's up to you. Okay, seeing a mix, but it looks like more, it looks like the A's have it so far at least.

[18:35](https://youtu.be/IJjgUWxPKvk?t=1115) A couple of hybrids of A and B here. If you put A, you are not alone. So, this was a question posed to a group of research participants, and this is the outcome here. You can see that the vast majority of the participants chose option A, the certain outcome where, for sure, 200 people will be saved. Now, I want you to wipe that from your memory. We're rewinding the situation. Again, it's 2040. You're the health commissioner. We have this new disease. It's extremely fatal. 600 people have it. They're all going to die, but this time, you turn to your most trusted advisor and you say, what have you got for me? Are there any possible treatments out there? The advisor says, well, we do have a couple, but they're not very good. In the first one, for sure, 400 people are going to die.

[19:46](https://youtu.be/IJjgUWxPKvk?t=1186) Out of the 600 people, definitely at least 400 will die. Then, there's also an option where we have a little bit of a gamble. There's a two-third chance that all of the 600 people will die anyway, but at least there's a one-third chance that no one will die at all. Which program would you choose? Okay, seeing a mix again, but it looks like the Ds have it this time. Once more, if you put D, you're not alone here. When the researchers posed this question to the group, the vast majority of people went with program D, the gamble here, where there's a two-thirds chance that everybody dies, but there's a one-third chance that zero people die. As we were going through this, you probably noticed that this is the same situation. It's just framed differently.

[20:53](https://youtu.be/IJjgUWxPKvk?t=1253) In one case, it's framed as a gain. We're thinking about the lives saved. In the other situation, we're framing it as a loss, where we're really focused on the people who are going to die. In this study, rather than showing these scenarios to people sequentially like I've done with you here, the researchers showed this to groups independently, so they didn't have the benefit of seeing, okay, it was framed this way in one situation, it was framed this way in the other situation, and this is what they found. There's a more than 50-percentage-point swing of the same situation, just depending on whether it was framed as a gain or a loss. The gain frame promoted risk aversion because people thought, I want to at least make sure that I save some people in this situation, whereas the loss framing induced risk-seeking because people were more likely to roll the dice and say, well, I want to try to at least save some people.

[21:57](https://youtu.be/IJjgUWxPKvk?t=1317) I am going to do one more game with you that, again, illustrates how system one can get us into trouble sometimes with things like framing effects. Oh, you got a little bit of a preview there? I have one more hypothetical situation for you, which is you've just parked your car, you're getting out, and you hear this commotion, so you look up and you see at the bank that you parked your car next to, there's clearly a heist happening. There's broken glass, there's alarms going off, people are shouting, and then you see three men run out of the bank full-on carrying sacks with dollar signs on them. You're like, this is a robbery. You see, because you're standing right in the perfect place at the perfect time, you see all of their faces, but just for a moment. I'm going to show you the faces of the three robbers, just for a second, and then later in the presentation, we'll come back to them and see if you can identify them. I'll show you each of them for a second. Try your best to remember what they look like. I see someone in the chat saying Ivan's testimony is super shaky, shouldn't be the primary source that leads to convictions. We're totally on the same page, and what we're going to do is kind of illustrate one of the reasons why that is the case.

[23:20](https://youtu.be/IJjgUWxPKvk?t=1400) So, get ready, you're about to see who the robbers are. Okay, and with that, I'm going to pause and hand over to Anna to talk a little bit more about what behavioral science looks like in the real world.

**Anna Keleher**

[23:43](https://youtu.be/IJjgUWxPKvk?t=1423) Thank you, Leah. So, these games are a fun way to feel system one and system two actually working in our brains, but behavioral science can have really big implications for how we design policy as well. So, behavioral insights is not here to replace traditional policy tools that many of you work with, like regulations and incentives and providing information to residents. Instead, we're here to enrich those tools and increase their effectiveness, both when we're trying to improve existing services to residents and also when we build new ones. So, on this slide, there's two examples of how policy design change using behavioral insights made a huge difference. In both examples, we see that defaulting people into a program makes a huge difference because when people are in that system one autopilot mode, they are more likely to stick with the default option. And so, these are just quick examples, but on the left, you can see that when the UK started having employees automatically enroll into a pension program as opposed to having them elect to be in it, the number of people who signed up was just massively larger because people were just on system one doing whatever is recommended to them, and it really makes a huge material difference in their lives. So, on the next slide, just to hammer home the crucial point here is that understanding how people behave and practice helps us design better policies and services. And on the next slide, we can see that behavioral insights can

[25:47](https://youtu.be/IJjgUWxPKvk?t=1547) be applied to a lot of different areas. So, first, information provision or communication. There's an example here. Last year, we helped the Louisiana Department of Health to refine a postcard reminder that would be mailed to people who are eligible for a booster shot. It's a little hard to see here, but the original was very dense and text heavy. So, we used behavioral insights to make it much easier for people to understand what they need to do and to take action without requiring them to delve into their system two brain. Behavioral insights can also inform changes to the environment. So, in the example you see here, the footprints are salient clues that model the desired behavior and help people who are in system one brain just do the right thing. I live in New York City. Another kind of similar environmental change that I see a lot is on the subway platforms, arrows pointing to where you should stand in order to be able to board the train. Things like that make it really easy to go to the right place, do the right thing. And then, lastly, we can also use behavioral insights to improve processes. So, this is an example on the right where we removed a tiny piece of friction from a process. So, in this example from Oregon, we removed just one click in the process of getting people to track their ballots. So, instead of sending people to a website with a homepage where they had to find a form, we sent them directly to the form

[27:47](https://youtu.be/IJjgUWxPKvk?t=1667) and that increased signups by almost 10%. So, it's that kind of tiny piece of friction that can have an outsized effect on the behaviors that people actually perform. But every population and every state is different and it's hard to know what works in practice in your state. So, there's a few ways that we can make this work really evidence-based and practical. And the first is to do the work to understand the behavior that you're trying to change and the context around that behavior. So, we conduct different activities to understand the barriers that residents are facing to performing a behavior. And there's different ways we can do this, including observing people while they're attempting or performing a behavior, looking at data, seeing where people might drop off throughout a process, conducting interviews with people to get their firsthand experiences, and also looking into the academic literature, the secondary research, to see what do we already know about how people behave and why they behave the way that they do.

[29:05](https://youtu.be/IJjgUWxPKvk?t=1745) And then behavioral insights also has a strong emphasis on learning what works. There are many ways we can do this, including doing field trials and experiments, doing simulated online trials to see how people react to different stimuli, or process evaluations to determine how something is rolling out in practice. So, the methodologies are different. I'm not going to get into them right now, but they're all focused on generating evidence about whether a solution is working. And you'll see through some of the case studies that I'm about to go over how collecting that evidence is really helpful for being able to determine whether a solution is effective, and it's whether it should be something that continues. So let's quickly look at an example of why evaluation is so important.

[30:08](https://youtu.be/IJjgUWxPKvk?t=1808) So starting in August 2012, the Texas Department of Transportation decided to show fatality messages along busy highways for one week each month. And these aim to seize people's attention and encourage safer driving. So it sounds like they're trying to get people out of that System 1 brain into System 2 and really consider the consequences of what they're doing. So can everyone please in the chat write whether they think that these signs were impactful? So either write yes, if you think that they were impactful, or no, if you don't. Okay. We have nos, we have yeses. People seem to split. All right. So those who wrote yes are correct, because as we'll see on the next slide, the signs do impact people, and they make them worse drivers.

[31:18](https://youtu.be/IJjgUWxPKvk?t=1878) So a couple people in the chat guessed that, so congratulations to them. But this is an example of why evaluation is so important. The solution might have seemed intuitive, but when researchers actually looked at it, they found that there were 4.5% more crashes within 10 kilometers of those signs. So it seems like it was actually seizing too much of drivers' attention and distracting them from conditions on the road. All right. So next up, we're going to look at some real-world examples of applying behavioral insights. So in this first one, we're going to be in the world of public health. We were tasked with increasing home lead testing completion rates in Chicago, where many people ordered lead testing kits for their home, but few people actually completed them.

[32:30](https://youtu.be/IJjgUWxPKvk?t=1950) So on the next slide, we'll see the original kit instructions. So in the chat, can everyone write if they think that completing this kit can be done in System 1, or if it requires System 2 thinking? So a 1 for System 1, 2 for System 2. Okay, yes, lots of 2s. So we actually conducted some qualitative research with residents. And Leah, I think if you hit next, you'll see some of the barriers that we uncovered. So we found that residents sometimes incorrectly complete the kit, incorrectly record information on the form because they don't understand where it should go. They find the materials confusing and overwhelming. They complete the kit, but then they never schedule a pickup because they don't realize that there's another step in the process. They forget about it, or the kit is moved or damaged before it's picked up because of that time lag.

[33:33](https://youtu.be/IJjgUWxPKvk?t=2013) So in our redesign, we made everything easier. So there is an infographic that helps residents understand one of the most commonly misunderstood steps. We simplified the process into three easy steps to help residents understand what they should do. We highlighted contact information and QR code that makes support options really obvious to people. On the next page, you can see that we chunked out the information and used visuals to make it easier to understand. We also used the colors of the corresponding bottles in the kit, matched those up with the instructions to make sure the resident is using the correct bottle. So just making that really intuitive for people. And then really emphasizing the scheduling and pickup part of the process so that residents know that they need to complete that immediately.

[34:37](https://youtu.be/IJjgUWxPKvk?t=2077) And then on the next slide, we also included some new stickers in the kit. So there's don't use stickers that residents can place on taps and toilets and places that they might try to turn on the water when it needs to be stagnating for 12 hours. So that's a reminder for them. And then a do not move sticker to remind other members of the household not to move the box once it's complete. So in the chat, please write if you think that these stickers are helpful in System 1 or System 2. So are they helpful when you're in the kind of autopilot System 1 or doing System 2 thinking? Yeah. Yeah, lots of ones. Exactly. So when it's the middle of the night and you walk into the bathroom and you're about to turn the water on to get a drink of water, you're definitely in System 1.

[35:43](https://youtu.be/IJjgUWxPKvk?t=2143) And a big red sticker telling you don't do that is super useful. So let's see the results on the next slide. Basically, the kit redesign worked really well. So we basically doubled the return rate for these kits, and it's now the highest in the nation. So the redesign kits you can see are those top lines up there, 40 to 50% of people returning, as opposed to the original kits where just 20 to 30 people were returning. So this kit redesign made a huge difference. All right. Now we're going to get into an example from sustainability work. Here we were tasked with trying to figure out how we can increase residential heat pump uptake.

[36:44](https://youtu.be/IJjgUWxPKvk?t=2204) So on the next slide, I think there's a picture of a heat pump. Basically, this is a home heating and cooling system that runs on electricity and is much more efficient than traditional furnaces. And so I know a lot of state governments right now are thinking about how they can increase uptake of heat pumps, especially with kind of federal dollars coming through for them. And so we had in a previous trial already identified some barriers to heat pump uptake. And here we wanted to see what are some of the messages that are going to work in order to make these more attractive to people. So on the next slide, I think there's some examples. So the experiment we ran here was randomizing people to see either no message or a message that we had written that was designed to overcome the barriers that we had already identified through research to adopting a heat pump.

[37:59](https://youtu.be/IJjgUWxPKvk?t=2279) So this was something that we ran in the UK. So some of these were pretty specific to the UK context. But the messages that we wrote were one around heating that works. So, you know, your house is going to be warm because that's a concern that people have sometimes. Also advertising the lower emissions of heat pumps. On the next slide, there's three more. One is about basically future proofing your home. So making sure your home has all the newest modern technology, emphasizing the property value that can increase if you get a heat pump, and then also energy security. And this was run in 2022 in the UK. So gas was being able to import gas was a huge concern at the time. So that was one of the message that was tested.

[39:00](https://youtu.be/IJjgUWxPKvk?t=2340) So basically we took the sample of about 8000 homeowners, we showed them at random, one of these messages, and then we asked about their future intentions to buy a heat pump. And this gave us very good evidence about what would work in the kind of public messaging campaigns that the government was running. So on the next slide we see our results. As you can see here, the top performing messages were the heating that works, property value, and energy security. All of those were statistically more impactful than the other messages. So this gives good evidence about what kind of messages are actually likely to move people's behavior. We also crucially asked participants, which messages they thought were appealing, and which messages they thought were accurate.

[40:09](https://youtu.be/IJjgUWxPKvk?t=2409) And this is also important because as a government, you are, you know, you're putting these messages out in the world, you want them to be impactful, but you also want people to like the messages that they're seeing, and to build trust with your residents as well. So that's why we like to often gather as much evidence as possible when we run these kinds of trials. So I think then that is my two case studies, and I'm going to pass it back over to Leah now.

**Leah Everist**

[40:47](https://youtu.be/IJjgUWxPKvk?t=2447) Thanks so much, Anna. And one last thought I would add to what Anna was just saying about this point of, you know, what do we see effective for behavior change, as well as what do people like, is there's kind of the inverse of that as well, you know. So if we're measuring behavior change, but people don't like it, then as Anna was saying, that has maybe implications for the long-term sustainability of the intervention. So what we sometimes see in some of our collaborations with different teams is that if we over rely on what people say that they like, sometimes people don't always have a very clear sense of what is going to change their behavior. So they might see some type of message in kind of a test environment, and they say, I really like the look and feel of this. This would definitely encourage me to, whatever it might be, go out and get my vaccination recycled properly, whatever it might be. But then we sometimes don't see that follow through. So when we think about kind of why we take a behavioral lens rather than a different lens to different types of challenges, that's a big part of the reason, kind of parsing those two pieces apart.

[42:00](https://youtu.be/IJjgUWxPKvk?t=2520) Okay, remember bank robbers? The time has come to try to identify them. So in our scenario, the detective has brought some folks in, and I'm going to show you a couple of slides where I want you to say yes in the chat if you think that one of the men on the screen is a bank robber. And say no if you think that the bank robber is not, none of the bank robbers are on the screen. All right. First one, yes, if you think a bank robber is there. All right, seeing a lot of yeses roll in. Someone, someone trying to protect themselves, got some street smarts. Okay, so do we think that the person on the left is a bank robber, yes or no?

[43:01](https://youtu.be/IJjgUWxPKvk?t=2581) No, right. And then do we think that the person on the right is a bank robber? Nice work. Okay, so far so good. Next screen. Do we see the bank robber on any of the bank robbers on the screen? Okay, we have some mixed yeses and nos here. This time I'm going to start on the right. Do we think that the fellow on the right is a bank robber? Seems like most people are in agreement that the fellow on the right is not a bank robber. And you are correct. What about the guy on the left? Is he a bank robber? Okay, some yeses, some nos. He is not a bank robber. But, well, we'll enlighten kind of what led so many people down that path in a moment.

[44:06](https://youtu.be/IJjgUWxPKvk?t=2646) How about here? Do we see any of the bank robbers? Yeah, okay, so we have a more unanimous answer here. And which one do we think? Is it the guy on the left or the guy on the right who's the bank robber? Yep. Okay, good memory. It is the one on the right. So why did this happen? This is an example of how framing is important. So I gave a couple of examples earlier of how our brain sometimes does this thing when it's in System 1, where it takes a hard question and substitutes it for an easier question. So because I only showed you the bank robbers for a quick second, you didn't have a lot of chance to really, you know, imbibe the details of their appearance. Then it's kind of difficult to look at one person and say, It's much easier for us to make judgments by comparison.

[45:06](https://youtu.be/IJjgUWxPKvk?t=2706) So in this simultaneous lineup, when I was showing you two people at a time, it's easier for your brain to say, which one of these two people looks more familiar? That's kind of what's going on in the background in system one, the substitution that's happening. And what researchers have found is that in situations like this, when folks see the two suspects simultaneously, like we saw in our group, they're much more likely to falsely identify someone as one of the bank robbers than they are if I were to show them to you sequentially one at a time. Okay, that's a good segue into our third case study, which is relevant to the justice sector. And in this case, it's about reducing source of income discrimination in the housing context. So in terms of the context here, this was a project we did with the New York Office of the Attorney General to identify key behaviors, barriers to compliance, and behavioral insights tactics to reduce source of income discrimination by landlords.

[46:17](https://youtu.be/IJjgUWxPKvk?t=2777) So for folks who are unfamiliar with this term, source of income discrimination refers to discrimination that might take place against someone who's applying to live in a house or in an apartment based on the way that they're paying their rent. So it's most commonly seen if someone is using some type of subsidy, like a housing voucher, which happens to be illegal. So lawful source of income is a protected category under New York state law. And that includes any kind of housing assistance, including HUD's housing choice vouchers. Despite the illegality, however, we see this type of discrimination very often. And this is not just a problem in New York. It's a problem in cities across the country, states across the country. And we see it at the level of landlords, as well as the level of real estate brokers who continue to turn away potential tenants who rely on subsidies to pay their rent.

[47:18](https://youtu.be/IJjgUWxPKvk?t=2838) So you might be wondering, why do they do this? Well, we looked at the research, and we also interviewed several people with subject matter expertise on this topic. And we found these four major barriers to equitable housing access. The first one concerns landlords' belief about rental assistance recipients and programs. So these are biases about individuals who receive housing vouchers and that lead landlords and brokers to behave in a discriminatory manner toward those folks. The second barrier has to do with the economic landscape and competition in the rental market. So for example, in New York, if you imagine an apartment is listed, and there are 20 people who are interested in it, if the majority of those folks are paying in a way besides, in another manner, not using housing vouchers, then the landlords might think, well, I might as well just choose one of them.

[48:21](https://youtu.be/IJjgUWxPKvk?t=2901) Basically, it's a good market for landlords where I have my choice of tenants. So there's no sort of economic pressure for compliance. Then we have the challenge, the third one here, about landlords' expectations about enforcement of source of income nondiscrimination law. So essentially, they think it's unlikely that they will be caught. They also think that even if they're caught, the penalties are unlikely to be significant. And within this, I should say that there were also some landlords who didn't fully understand the illegality of this type of discrimination, or felt that they were not liable for it somehow. And then finally, we have this challenge of existing social dynamics among landlords and brokers that have normalized source of income discrimination. So basically, landlords who think this is something that everyone does, this is just part of our industry, and therefore, it's totally fine. I'm not acting in a way that's out of the ordinary.

[49:24](https://youtu.be/IJjgUWxPKvk?t=2964) So within this project, we proposed three different communication strategies to change people's perceptions of enforcement. So really focused on that third barrier. And then we also proposed two potential partnerships and one policy solution to influence existing social dynamics. I know that that's a lot. So I'm just going to talk about one of them for now for the sake of time, which was framing noncompliance as a deliberate act. So remember, the behavior that we're concerned about in this case is landlords discriminating against people because they're paying their rent, they're applying for an apartment and proposing to pay their rent with housing vouchers. A challenge with that behavior is that it can be seen, not choosing a person who's in that situation as a tenant, can be seen as more of a non-behavior than a behavior. So landlords and brokers who have a pattern of not offering apartments to voucher holders may feel like they're somewhat protected because it's an act of omission.

[50:28](https://youtu.be/IJjgUWxPKvk?t=3028) They're not doing something wrong. They're not doing something right rather than doing something wrong is sort of a framing that we see. And this is something we see in other behaviors as well. In the vaccine space, for example, we see people who view getting a vaccine as a behavior, but they wouldn't view not getting a vaccine as a behavior in the same way because, again, it's this kind of act of omission. So we wanted to change that. And what we proposed was an intervention where we assume best intentions, and then we provide some education, and then we try to tackle that omission bias. So we're kind of saying, OK, all is forgiven, but now you know, and we won't forgive you again. So at the top of the screen here, you can see that this is something, an approach we've seen effective in other areas. So in the domain of tax compliance, we tried this on the top right. There's an example of when we did this in Poland. So we said, you know, so far we've thought of your payment delay as an accident.

[51:31](https://youtu.be/IJjgUWxPKvk?t=3091) However, if you disregard this notice, we will consider it an intentional choice of yours and will treat you as dishonest, et cetera, et cetera. So again, there are a couple of reasons why this works, and they're all kind of wrapped up together in this example. So first of all, it increases cognitive dissonance. So it's hard to still feel like a good person when you're being told that your actions are not good. The way that we frame this is, OK, look, we're not saying you were bad before, but now that we're letting you know that you're breaking the law, if you continue to do it, then that's bad and we'll come after you. It also suggests that the problem is being that the behavior rather is being closely monitored. So if you remember from the last slide when we were talking about barriers, one of the challenges was that landlords didn't think that this was a behavior that they would get in trouble for. And so this kind of language corrects that because it has this sense of like, OK, there, someone is keeping an eye on you.

[52:34](https://youtu.be/IJjgUWxPKvk?t=3154) It also leverages a concept that we call reciprocity, which is this sort of, you know, you scratch my back, so I'll scratch yours attitude that people have where when someone does something for us, it kind of endears them toward us or makes us want to do something back for them. So the way that we leverage this is by extending that benefit of the doubt for the noncompliance up to till this point. So instead of framing this this notice in a way that's like, we know you've been breaking the law, we're saying, look, we think it was probably an accident because you're such a good person. However, you know, moving forward, we won't be so so lenient, which brings me to this fourth point about why it's effective, which has to do with timeliness. So a lot of times people are more motivated to change their behaviors in these what we call fresh start moments. You can think about, you know, New Year's resolutions or your birthday if you if you think, OK, this year I want to do X and Y differently.

[53:36](https://youtu.be/IJjgUWxPKvk?t=3216) That's what we're trying to leverage here. OK, this is, you know, kind of create this fresh start. This is the moment for you to change. This one has not been evaluated yet, but we actually just started working with the attorney general's office and a couple of other stakeholders again, and we're hoping to deploy and evaluate this. So if this is something you're working on, please shoot me an email and I'll do my best to keep you looped in to how that is going. OK, we have one more case study for you, one more story to tell, and then we'll switch over to questions. I saw some really interesting questions in the chat already, so definitely want to make sure we have plenty of time for that. But this one is for the transportation folks in the crowd, and it has to do with decreasing highway fatalities by reducing speeding reoffense. This is one from the UK, and I should have said in the beginning when I was telling you about the behavioral insights team, we actually started as a government team within the UK.

[54:37](https://youtu.be/IJjgUWxPKvk?t=3277) And because we started as a government team, a lot of the work we do is with the public sector. We've now been working in the U.S. for many years, hundreds of projects on the U.S. side and worked with, as you can see from some of these case studies, all levels of government, cities, states, federal. I think our collaborations with states are particularly exciting because states do so much really excellent and important work for folks. And we work to states through direct partnerships, through subcontracting to other organizations, through universities, a lot of different ways, but definitely a lot of exciting work on the state side. OK, back to the story. So this is about 10 years ago when the rate of deaths and serious injuries on the roads of a particular UK county was 64 percent higher than the national average at the time. So reasons for this, why are drivers breaking the speed limit?

[55:42](https://youtu.be/IJjgUWxPKvk?t=3342) One of the big ones that came up in our in our research was that a lot of drivers didn't think the speed limits were legitimate or that they were specific to the different parts of the road where they might change. So we conducted a couple of different types of research here to to discover some of these insights. We did a survey, we did some interviews, and in the survey we found that more than 80 percent of folks thought other people regularly exceeded the speed limit. So there is this kind of negative social norms indicating that respondents perceive that speeding behavior is normal. It's it's this kind of everyone is doing it problem. We also had this challenge whereby drivers underestimate the probability that they will get caught. So kind of similar to the landlords and the brokers, there's this optimism bias where people underestimated the likelihood of negative events. And that's something that we see in a lot of different contexts for a lot of different behaviors. And then finally, we have this challenge of drivers being overconfident, particularly on familiar roads.

[56:53](https://youtu.be/IJjgUWxPKvk?t=3413) So when we were first talking about system one and system two, we even had driving on the screen as an example. And this is like the number one situation right where you're in system one mode, when you're right by your house, you're familiar with the roads. And so you're more likely to be kind of in an autopilot rather than paying close attention to the elements of the road around you. So in in some of the research that we did, we saw that 90 percent of drivers in the county thought they drove average or better, obviously, from a mathematical perspective that doesn't check out. But it demonstrates sort of the confidence of drivers on the road and and particularly when they were close to home, even though our data science work showed that the likelihood of a serious collision was actually higher. The closer people were to home, probably because of this this system one being in control and people potentially not paying as much attention or being a little bit more comfortable and relaxed.

[58:00](https://youtu.be/IJjgUWxPKvk?t=3480) So we hypothesize that a speeding citation packet was a good moment to change speeding behavior that was currently being underutilized. So what was happening is when someone got caught speeding, they would be sent this packet of materials. And in that moment, they might be having some regret because they have to pay a fine or maybe they got into a minor accident or a major accident. And so it could be a moment where they're kind of taking stock and there's an opportunity for us to get in there and and shift behavior. So we said, hey, can we redesign, redesign your letter and some of the other resources in this packet? So we started off modifying this leaflet that would go to drivers who had been caught speeding, and it starts at the top with boosting the salience of the consequences of speeding. So there's a photograph here of a roadside shrine and this emotional header that, again, tries to stay away from being directly accusatory.

[59:02](https://youtu.be/IJjgUWxPKvk?t=3542) You're a bad person, but but really drives home the potential consequences of actions like the one that this person would have just taken. Then corresponding to that barrier around kind of legitimacy, why do we have these speed limits that we do? The letter explains the rationale for speed limits and links their existence to the prevalence of serious collisions in the county. So kind of providing some justification, this is where we've had accidents. This is why we have speed limits, et cetera. And then finally, closing with what we call primacy or recency to to help folks recall some of the key elements here. So primacy and recency both have to do with how easily something comes to mind for you. And the goal with a letter like this is that it's not only apparent to someone why they shouldn't speed while they're reading the letter, but that those messages stay with them. Right. So we want we. I want some of the information in this letter to come to mind really quickly and readily when folks are driving next time.

[1:00:09](https://youtu.be/IJjgUWxPKvk?t=3609) I won't go through this one in as much detail for the sake of time, but the packets also included a Notice of Intended Prosecution, or NIP, which we also redesigned. So you can kind of imagine what this looked like before in the sense that a lot of these types of documents can be really, really text-heavy. They have a lot of legal language on them that's not necessarily easy to decipher. So what we wanted to do is really be clear about the call to action, clearly state what the next step or action we wanted people to take was, and we also have this sort of behavioral insights lever in here asking people to sign a declaration or nominate another driver, so incorporating some social norms in there as well. And what we found here, so to kind of explain what we're looking at, all the way on the left we have the old leaflet and the old Notice of Intended Prosecution.

[1:01:12](https://youtu.be/IJjgUWxPKvk?t=3672) So in that case, the likelihood of re-offense was 4.7%, and then we tried a couple of different variations where we had the old leaflet versus the new Notice of Intended Prosecution, the new leaflet and the old Notice of Intended Prosecution. You can see those ones kind of bump that re-offense down a little bit, particularly with the new Notice of Intended Prosecution, but then when we combined both of those things and sent people the kind of full package of behaviorally redesigned tools, there was a 23% reduction in the likelihood of re-offending, and that in this particular county translated to 170 fewer re-offenses within the next six months compared to business as usual. And it would have been, because of the way that we did this evaluation, during that same period it would have been almost, let's see, 560 fewer re-offenses if all four of those categories had had the same treatment.

[1:02:17](https://youtu.be/IJjgUWxPKvk?t=3737) So in an evaluation like this, what we do, if something works, is obviously we want to scale it up. Ah, Jocelyn, NIP is Notice of Intended Prosecution, so that was this form that we redesigned. And this is our last case study, so I'll just close by saying, again, in a situation like this, when we see clearly that the new leaflet and the new Notice of Intended Prosecution are more effective, then we want to basically scale that up so that we reduce speeding re-offense as much as possible. So I'll pause there. That was our final case study. Thank you for playing along with all of our mind games and for asking good questions. I know Hannah mentioned earlier that this is the moment to pause the recording, I believe, but the floor is now open for questions.

**Unknown speaker**

[1:03:14](https://youtu.be/IJjgUWxPKvk?t=3794) Thank you so much, Leigh Ann and Anna. I am going to stop the recording.

* * *

Original page: https://innovate-us.org/using-behavioral-science-to-revolutionize-policy-making-and-service-delivery · Licence: CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/), as stated on innovate-us.org