Apu Gomes / Getty

As Americans have taken to the streets, protesting the killing of George Floyd, Donald Trump has repeatedly urged police to “dominate the streets,” pressing law-enforcement agencies and the military to act aggressively against the protesters. In some ways, Trump’s calls are unsurprising. Many Americans believe that, when a violent protest occurs, police should use any means necessary to get things back under control, arrest the protesters, and allow people to return to their regularly scheduled program of activity. These calls are even more predictable when race is involved. One study of thousands of protests found that, even after controlling for the use of violence by protesters, police are much more likely to use repressive techniques when the protesters are black.

But aggressive approaches frequently backfire. All too often, the spark that ignites a violent protest is the use of force by police. And when law-enforcement personnel adopt a combative stance, they tend to escalate the conflict, which may quickly spiral beyond their control. If police around the country actually want to maintain or restore order, my own research suggests that they would do better to ignore the president and try negotiating with the protesters.

Social scientists have learned a lot about police-protester interactions and crowd behavior over the years. They divide police responses to protests into two main approaches. The first is a get-tough model of crowd policing, the sort championed by Trump, called “escalated engagement.” The second is a more moderated style called “negotiated management.” Many have objected that Trump’s approach is not legal, justifiable, or right. But escalated engagement has another problem—in general, it doesn't actually work.

In some instances, if police demonstrate an overwhelmingly powerful presence either before a riot breaks out or at its very early stages, it is indeed possible for them to control the situation. But it is next to impossible to predict exactly where and when a riot is going to break out—by the time enough police are mobilized and in position, it’s usually too late.

Much more frequently, a show of force by police can make a situation far worse. The protests of the police killing of Floyd have unfolded in a manner that has become too common in recent decades. Protests emerge in reaction to a police killing. A march or rally proceeds peacefully until the protesters block a street, go down an unplanned route, or stay in an area longer than expected. The police intervene, the protesters resist, and then a confrontation ensues, typically spinning into something neither side expected. Police can quickly find themselves outnumbered, their show of force producing the circumstance it was designed to avoid.

Increasing the number of police who are trying to contain protests—particularly with inexperienced officers from outside the community who are not trained for crowd confrontations and lack the discipline necessary for such situations—often fans the flames instead of tamping them down. When the National Guard or other troops are called in to help police crowds, they usually escalate the combative nature of the situation, or play a central role in producing a violent confrontation in the first place.

Negotiated management may not satisfy the president’s demand to “dominate,” but the evidence shows that it produces considerably less violence and destruction. In a 2004 study, I looked with a colleague at the equipment 400 police departments had stockpiled and the training they had received about how to handle crowds. Some departments were focused heavily on stockpiling repressive equipment—tear gas, smoke grenades, riot helmets, gas masks, riot batons—and providing their officers with the training to use it, while other departments had greater stores of assistive equipment—such as stretchers, ambulances, mobile communication devices, and lighting—and focused their training on understanding crowds and predicting and preventing riots. As it turned out, preparing for escalated engagement was a self-fulfilling prophecy: Repressive equipment increased the chances that a jurisdiction would experience a riot, and then that it would have additional riots. Assistive equipment had just the opposite effect.

This isn’t terribly hard to understand. In the typical escalated-engagement encounter, police show up at a protest, confront the protesters, and try to force them to do something they do not want to do. Contrast that with negotiated management, in which the police actually facilitate the protest—even help plan it—and in doing so, get concessions from the protest organizers that make the protest less disruptive and less likely to spin out of control.

But it’s not just how departments prepare for, and choose to handle, protests that makes the difference. The history of relations between the community and the police matters too. When police-community relations are already strained from heavy-handed encounters and acts of brutality, and when the police lack formal connections to the community and channels for feedback, almost any incident may trigger feelings of mistrust developed through a long history of negative interactions. Under those conditions, any confrontation is much more likely to spiral into larger protests or even riots.

Escalated engagement and the use of repressive force have been repeatedly shown to be failed approaches to quelling protest and rioting. Threats of domination made by police chiefs, mayors, and especially the president do not deter action—they merely build community and protester resentment, providing additional grievances to motivate protesters and rioters.

Acting tough, it turns out, is the approach that ultimately undercuts law and order.

We want to hear what you think about this article. Submit a letter to the editor or write to letters@theatlantic.com.