“What are you DOING?” I yelled. My four-year-old son froze, his tiny hand poised above the tower of toilet-paper rolls he’d been building … right on top of a hot electric heater.

“You can’t DO that!” I exclaimed, whisking the rolls onto a high shelf. “You could burn down the house!” Since it was the third time I’d found him building a tower in the same spot, I decided to get serious, launching into a long, vivid rant about the potentially dire consequences: All our stuff would go up in flames! We’d have nowhere to live!

As I wrapped up my tirade, I paused, looking down at his little bowed head, suddenly feeling terrible. Had I gone too far? Had I scarred him for life? Would he someday recount to his therapist about how his awful mother had created his lifelong fear of fire?

Just then, he raised his head and looked up at me with delighted smile, his innocent brown eyes sparkling with joyful anticipation. “If there’s a fire,” he said breathlessly, “will the FIRE TRUCKS come?”

So, yeah. That’s the thing about telling another person what they should or shouldn’t do: Even when the intended message seems crystal clear, sometimes what the listener hears is something entirely different. It can happen in our families, but on a larger scale, too—in our policy efforts, classrooms, workplaces, and more.

First, as my son taught me, the message can be undercut when the bad result is inadvertently made to look appealing or fun. Early antidrug campaigns were sometimes spectacularly guilty of this, showing glamorous young people partying at fun-looking parties before their not-that-terrible downfalls. (And I still remember watching the movie Heathers and thinking, wow, that Christian Slater character is an objectively awful, awful person … but he is kind of cute…) An ill-advised TV spot intended to teach young children about the dangers of prescription medicines instead featured smiling pill-shaped puppets singing adorably about how you shouldn’t eat them (which, as one writer noted, is very much like the ad campaign for M&Ms: "Please don't eat us … wait! You ate me and I was delicious! You're so clever!” Probably not what the Long Island Poison Control was going for.)

Another TV ad—this one intended to fight childhood obesity—showed kids racing around their house, gorging on cookies, building forts out of cookies and making cookie smoothies; the point was supposed to be the unpleasant subsequent sugar crash, but most kid viewers must have thought, “Cookie forts? Cookie smoothies?! Where is this magical wonderland?!” (At least that’s what I thought.) And those instructional videos at trampoline parks that demonstrate what the kids shouldn’t do seem destined to backfire—because honestly, that long-earringed guy doing the sideways triple flip while chewing gum looks like he’s having a pretty great time.

Research shows that guidance can also backfire when it seems too bossy. If your company forces you to attend mandatory training on, say, diversity, not only are there likely to be plenty of rolled eyes and grumbles, but studies show that making such training mandatory will—ironically enough—actually lead to the opposite result over time. (In one study, after five years of compulsory diversity training, companies had fewer Asian-American workers and African-American women on staff than before.)

And good advice also goes awry when it inadvertently points out that an undesired behavior is popular and widespread: i.e., “everybody else is doing it.” Surprisingly, the antidrug D.A.R.E. program was found to actually increase the rate of student drug use because it gave students the impression that drugs were everywhere and therefore many of their peers must be using them. When the IRS increased tax penalties because so many people had cheated on their taxes, cheating actually increased the following year—because people figured, heck, if so many people cheat, I might as well, too!

When researchers put up two different signs in the Petrified Forest National Park—one stating, “Please don’t remove petrified wood from the park” and the other stating, “Many past visitors have removed petrified wood from the park” so please don’t make things worse by taking more—the second sign led to more theft, because the sign made stealing seem normal (“many past visitors”). And every election season, voting advocates shoot themselves in the foot by loudly bewailing the number of people who don’t vote—inadvertently sending a message that not voting is the social norm.

It’s a surprisingly easy mistake to make. Just last week, I heard a well-intentioned public-service ad stating the high percentage of kids who start experimenting with drugs before age 12—which gave an unintentional heads-up to 12-year-olds everywhere that they were missing out on something many of their peers were into. The announcer at my public radio station noted that “only 9 percent of our listeners donate to the station!”—unintentionally setting up the idea that the social norm is not contributing.

The scary part is, mistakes like these can have life-or-death consequences: When a group of teenagers took part in a suicide-prevention program that emphasized how often teen suicides occur, it actually made the teens more likely to consider ending their own lives. If everybody else is doing it …

What this all means is that we have to be careful—as parents, educators, journalists, policymakers—about the messages we’re sending. True, we need to tell kids about the dangers of vaping, but if we do so while bemoaning the “epidemic” of vaping in their age group or at their school, we inadvertently set up vaping as the social norm, which they’re likely to emulate.

Yes, we need to warn our kids about the dangers of sexting and online bullying—but if we launch into the topic with a speech about how “everyone is doing this, it’s everywhere—but you shouldn’t do it,” we undercut our own message. And while we need to talk to our kids about the importance of exercise, we undercut our message if we do so while bewailing the “sedentary, phone-obsessed lives of all you kids today,” because that just reinforces the those traits as the norm.

The good news, though, is that there’s a flip side: Each of these insights can be put to use toward positive goals.

On the “undesirability” front, we can make unwanted actions seem disgusting or potentially fatal. Emphasizing the gruesome effects of smoking through photos of diseased lungs on cigarette packages really does make smoking less appealing. Emphasizing the recent increase in teen deaths resulting from vaping (rather than bemoaning its widespread use) could help combat the appeal of e-cigarettes.

On the “don’t be too pushy” front, we can encourage and empower positive action without demanding it. Research shows that diversity programs are most effective when they’re voluntary—so company leaders can put out the word that these programs are helpful, but it’s okay if you can’t make it. Encouraging employees to take positive action regarding their health—rather than scolding or penalizing them financially—can lead them to adopt healthier lifestyles.

Messages to college students about binge drinking can be effectively framed as information to help students make their own decisions, rather than scolding or shaming. One organization hit the nail on the head when it recently changed its name and branding from the tut-tutting “National Campaign to Prevent Teen Pregnancy” to the empowering, how-can-we-help-you-reach-YOUR-goals moniker, “The Power to Decide.” (Gives it a whole different feel, doesn’t it?)

And on the “everybody’s doing it” front, we can frame choices in terms of the positive social norms we want to encourage. When 800 doctors were sent letters stating that they were prescribing more antibiotics per patient than the majority of their peers, the letter caused them to subsequently prescribe 73,000 fewer antibiotics than they otherwise would have—a valuable step in reducing antibiotic resistance. When drivers in Louisville, Ky., received letters telling them that “the majority of drivers” in their town pay their fines within 13 days, they paid their own fines twice as readily. When hotel guests are told that the majority of other guests reuse their towels—rather than simply told that reusing towels will save water and energy—they are more likely to reuse their towels themselves. When voters are told that most of their neighbors are voting, it makes them more likely to vote.

Using social science to influence the behavior of others can be a delicate business, but done right, it can be incredibly effective—keeping our kids safer, encouraging better health, building civic engagement, helping to save the planet … and maybe even keeping our toddlers from burning down the house.