Brett Gelino studies human response times to natural disaster alerts

As climate change ushers in unpredictable weather in many regions, ensuring the public is attuned to severe weather warnings and alerts is important for protecting both property and lives. Brett Gelino, PhD, a postdoctoral researcher at the Johns Hopkins School of Medicine, and his colleagues have taken a unique approach to the concern. Using methods from both behavioral economics and operant psychology (the study of voluntary behaviors that are influenced by rewards and consequences), they have researched how people in Tornado Alley—an area of the Great Plains stretching from Texas to the Dakotas where tornadoes are frequent—react to tornado alerts, finding the sweet spot in timing these alerts so that people prepare and leave immediately rather than wait around.

These research methods can also provide insights on how to help people make choices such as switching to renewable energy for their homes. “There are a number of these metrics that tell us something really interesting for policy, and thus this lends a really nice framework for studying policy-relevant concerns,” said Gelino, who is also a member of APA’s Climate Change Psychology Advisory Group. The Monitor asked Gelino to talk about his research and how the psychology of decision-making intersects with climate science.

In my University of Kansas laboratory under Dr. Derek Reed, we were expanding our scope to look at areas of broad policy concern related to climate change and sustainability.

I’m originally from Connecticut, and Dr. Reed is from Illinois, and one thing we’d noticed is that folks in the Midwest don’t seem to react as quickly as they should when tornado alerts are delivered. That hinders the amount of time people have to prepare if there is extreme weather coming. Climate change is going to cause more unpredictable weather, so that’s something we wanted to home in on.

We used the framework of delay discounting, or the idea that people value outcomes less when they occur farther in the future. An older study by another research group had already raised some significant concerns with respect to lead times, which is the time between when the tornado is likely to strike and when the alert goes out.

So, for example, if I said there’s a tornado about to strike in 5 minutes, you’re probably going to say, “OK, I don’t have much time to prepare; I’m going to get ready right now.” But if I say 30 minutes, you might think, “Oh, you know, I’ve got a couple minutes, maybe I’ll go outside and see if the sky looks that nice pea green.” Hence the discounting of that delayed outcome.

We used a simulated task to create that scenario. Tornadoes don’t strike every day, and it’s not something we can test in the real world; we wouldn’t want to modify the actual alerts just for the sake of our findings. We created approximations of a push alert you receive on your cell phone and asked people, at a series of increasing lead times, the probability that they would immediately stop their preferred activity and prepare for a tornado. We also wanted to investigate the impact of different types of emergency alerts. So instead of just saying “tornado alert” or “tornado warning,” some of these warnings provided a brief description of the impacts expected, such as damage to cars or damage to roofs.

Previous research has found that when alerts precede a tornado by 15 minutes or more, recipients delay preparation, and the storm becomes more deadly. Longer lead times are dangerous because people assume they have more time to prepare. They wait around, go outside, and maybe try to snap a picture or video.

We found that for those who received the standard warning, their indecision point—or the pivotal moment where they must decide how to respond to the alert—came at about 15 minutes before the storm’s impact. They delayed acting until the last possible moment they could. By contrast, for those who received the more descriptive impact-based warning, their point of indecision between seeking shelter or waiting was at about 30 minutes before the storm’s impact—they took immediate action even though they had more time to prepare because they had more information about the severity of the situation. The novel language was enough to create urgency so that folks were considering seeking shelter with twice as much time to prepare.

In 2019, a tornado struck Linwood, Kansas, and Dr. Reed and I collaborated with structural engineer Elaina Sutley and her research group to document residential damage, risk perceptions, and sheltering behavior. We went door-to-door and talked to folks who reported damage from that tornado, looking at the integrity of their home and the security they felt in their home, and how quickly they said they would seek shelter in the future, having experienced a tornado recently. The data generally suggested a poor match between residents’ perceived in-home safety and the actual safety of their homes. Many of these homes were ill-equipped to handle a strong storm, a finding that bolsters the importance of ample extreme weather preparation.

Once we establish a precise model of decision-making, we can test strategies that might be too expensive to roll out without testing in a real community. I’ve looked at people’s willingness to enroll in a green energy source such as solar or wind power. If we can effectively model willingness, we can start introducing wording changes or classic approaches like nudges to get people to adopt renewables. A lot of my work is looking at extending these behavioral economic models to the sustainability effort and climate change.