Changing Real-World Behavior with Behavioral Requests and Social Norms Messages: A Meta-Analysis of Results from Field Experiments

Imagine you are trying to get a whole neighborhood to stop littering. You have two main tools in your toolbox. The first is a Social Norm, which is like a giant billboard saying, "Hey, 90% of your neighbors already pick up their trash!" This relies on the human desire to fit in and do what everyone else is doing. The second tool is a Behavioral Request, which is much simpler: it's just a polite sign that says, "Please pick up your trash." For years, scientists have been trying to figure out which of these tools works better to change how people act in the real world. They've been testing signs, flyers, and messages in parks, hotels, and streets to see if telling people what "everyone else" does is more powerful than just asking them to do the right thing. This question matters because if we want to solve big problems like climate change or public health, we need to know exactly which type of message actually gets people moving, rather than just guessing.

This paper is like a giant detective story where the authors, Benjamin Douglas and Markus Brauer, decided to stop guessing and start counting. They gathered results from 53 different real-world experiments (involving 290 different comparisons) to see what actually works. They noticed a messy problem in previous research: scientists had been mixing up their labels. Sometimes, a simple "Please do this" sign was being called a "Social Norm" message, even though it didn't mention what other people were doing at all. It was like calling a hammer a screwdriver just because both are tools.

The authors decided to clean up the mess by separating the messages into distinct categories: Behavioral Requests (just asking people to do something), Descriptive Norms (telling people what others do), and Injunctive Norms (telling people what others approve of). Then, they ran the numbers to see which one actually changed behavior.

Here is what they found, and it might surprise you. When they compared these messages to a control group (people who saw no message at all), only two types of messages actually worked to change behavior:

1. Behavioral Requests: Simply asking people to do the thing (e.g., "Please turn off the lights") was effective.
2. The Power Combo: A Behavioral Request plus a Descriptive Norm (e.g., "Please turn off the lights, and by the way, most people in this room do it") was even more effective.

However, the paper explicitly rules out the idea that Injunctive Norms (messages about what people approve of) have any real power on their own. The data showed that telling people "Your neighbors think you should do this" had an effect size of 0.00, which means it was no better than saying nothing at all. The authors also found that while Descriptive Norms alone (just saying "Most people do this") had a tiny effect (d = 0.03), it was barely noticeable compared to the power of just asking someone to do the task.

The study suggests that for a long time, researchers might have been giving credit to "Social Norms" for changes that were actually caused by the simple act of asking. It's as if a coach thought their team won because of a pep talk about how "great teams win," when really, the team won because the coach simply said, "Run faster!" The paper indicates that the "pep talk" about what others think (Injunctive Norms) doesn't seem to move the needle, but the direct request does.

There is a catch, though. The authors note that the results vary wildly from study to study (a statistical measure called I² was over 75% in many cases), meaning the effect isn't always the same in every situation. They also point out that they couldn't fully explain why behavioral requests work so well or how long the effect lasts. But based on the evidence they gathered, the most reliable way to change real-world behavior isn't to lecture people about what their peers think; it's to simply ask them to do it, perhaps with a little reminder that others are doing it too.