Growing up, I was taught by my karate sensei never to start a fight unless I was prepared to be the only one standing at the end. That’s how I approached debates at work and with friends: I thought the key to victory was to go into battle armed with airtight logic and rigorous data. The harder I attacked, though, the harder my opponents fought back. I was laser-focused on convincing them to accept my views and rethink theirs, but I was coming across like a preacher and a prosecutor. Although those mindsets sometimes motivated me to persist in making my points, I often ended up alienating my audience. I was not winning.

For centuries, debating has been prized as an art form, but there’s now a growing science of how to do it well. In a formal debate your goal is to change the mind of your audience. In an informal debate, you’re trying to change the mind of your conversation partner. That’s a kind of negotiation, where you’re trying to reach an agreement about the truth. To build my knowledge and skills about how to win debates, I studied the psychology of negotiations and eventually used what I’d learned to teach bargaining skills to leaders across business and government. I came away convinced that my instincts—and what I’d learned in karate—were dead wrong.

A good debate is not a war. It’s not even a tug-of-war, where you can drag your opponent to your side if you pull hard enough on the rope. It’s more like a dance that hasn’t been choreographed, negotiated with a partner who has a different set of steps in mind. If you try too hard to lead, your partner will resist. If you can adapt your moves to hers, and get her to do the same, you’re more likely to end up in rhythm.

In a classic study, a team of researchers led by Neil Rackham examined what expert negotiators do differently. They recruited one group of average negotiators and another group of highly skilled ones, who had significant track records of success and had been rated as effective by their counterparts. To compare the participants’ techniques, they recorded both groups doing labor and contract negotiations.

In a war, our goal is to gain ground rather than lose it, so we’re often afraid to surrender a few battles. In a negotiation, agreeing with someone else’s argument is disarming. The experts recognized that in their dance they couldn’t stand still and expect the other person to make all the moves. To get in harmony, they needed to step back from time to time.

One difference was visible before anyone even arrived at the bargaining table. Prior to the negotiations, the researchers interviewed both groups about their plans. The average negotiators went in armed for battle, hardly taking note of any anticipated areas of agreement. The experts, in contrast, mapped out a series of dance steps they might be able to take with the other side, devoting more than a third of their planning comments to finding common ground.

As the negotiators started discussing options and making proposals, a second difference emerged. Most people think of arguments as being like a pair of scales: the more reasons we can pile up on our side, the more it will tip the balance in our favor. Yet the experts did the exact opposite: They actually presented fewer reasons to support their case. They didn’t want to water down their best points. As Rackham put it, “A weak argument generally dilutes a strong one.”

The more reasons we put on the table, the easier it is for people to discard the shakiest one. Once they reject one of our justifications, they can easily dismiss our entire case. That happened regularly to the average negotiators: they brought too many different weapons to battle. They lost ground not because of the strength of their most compelling point, but because of the weakness of their least compelling one.

These habits led to a third contrast: the average negotiators were more likely to enter into defend-attack spirals. They dismissively shot down their opponents’ proposals and doubled down on their own positions, which prevented both sides from opening their minds. The skilled negotiators rarely went on offense or defense. Instead, they expressed curiosity with questions like “So you don’t see any merit in this proposal at all?”

Questions were the fourth difference between the two groups. Of every five comments the experts made, at least one ended in a question mark. They appeared less assertive, but much like in a dance, they led by letting their partners step forward.

Recent experiments show that having even one negotiator who brings a scientist’s level of humility and curiosity improves outcomes for both parties, because she will search for more information and discover ways to make both sides better off. She isn’t telling her counterparts what to think. She’s asking them to dance. Which is exactly what Harish Natarajan does in a debate.

DANCING TO THE SAME BEAT

Since the audience started out favoring preschool subsidies, there was more room for change in Harish’s direction—but he also had the more difficult task of advocating for the unpopular position. He opened the audience’s mind by taking a page out of the playbook of expert negotiators.

Harish started by emphasizing common ground. When he took the stage for his rebuttal, he immediately drew attention to his and Debra’s areas of agreement. “So,” he began, “I think we disagree on far less than it may seem.” He called out their alignment on the problem of poverty—and on the validity of some of the studies—before objecting to subsidies as a solution.

We won’t have much luck changing other people’s minds if we refuse to change ours. We can demonstrate openness by acknowledging where we agree with our critics and even what we’ve learned from them. Then, when we ask what views they might be willing to revise, we’re not hypocrites.

Convincing other people to think again isn’t just about making a good argument—it’s about establishing that we have the right motives in doing so. When we concede that someone else has made a good point, we signal that we’re not preachers, prosecutors, or politicians trying to advance an agenda. We’re scientists trying to get to the truth. “Arguments are often far more combative and adversarial than they need to be,” Harish told me. “You should be willing to listen to what someone else is saying and give them a lot of credit for it. It makes you sound like a reasonable person who is taking everything into account.”

Being reasonable literally means that we can be reasoned with, that we’re open to evolving our views in light of logic and data. So in the debate with Harish, why did Debra neglect to do that—why did she overlook common ground?

It’s not because Debra is eight years old. It’s because she isn’t human.

Debra Jo Prectet is an anagram I invented. Her official name is Project Debater, and she’s a machine. More specifically, an artificial intelligence developed by IBM to do for debate what Watson did for chess.

They first dreamed the idea up in 2011 and started working intensively on it in 2014. Just a few years later, Project Debater had developed the remarkable ability to conduct an intelligent debate in public, complete with facts, coherent sentences, and even counterarguments. Her knowledge corpus consists of 400 million articles, largely from credible newspapers and magazines, and her claim detection engine is designed to locate key arguments, identify their boundaries, and weigh the evidence. For any debate topic, she can instantaneously search her knowledge graph for relevant data points, mold them into a logical case, and deliver it clearly—even entertainingly—in a female voice within the time constraints. Her first words in the preschool subsidy debate were, “Greetings, Harish. I’ve heard you hold the world record in debate competition wins against humans, but I suspect you’ve never debated a machine. Welcome to the future.”

Of course, it’s possible that Harish won because the audience was biased against the computer and rooting for the human. It’s worth noting, though, that Harish’s approach in that debate is the same one that he’s used to defeat countless humans on international stages. What amazes me is that the computer was able to master multiple complex capabilities while completely missing this crucial one.