Conflicted Page 17
List and his team had already established that poor service was costly for Uber. Customers who were delivered to their destination 10–15 minutes late spent 5–10 per cent less on future trips. Ho and List wanted to find out if an apology would boost a customer’s spending back up. Along with two fellow economists, Basil Halperin and Ian Muir, they devised an experiment to help Uber figure out what makes an effective apology, and how much one is worth.
The researchers had a big, real-time dataset to play with, garnering information from 1.6 million passengers across America’s major cities. They were able to identify which passengers had recently had a bad trip and ensured that these people received an email containing an apology within the hour. The economists divided the passengers randomly into eight groups and sent different apologetic messages to all but a control group, which got no apology (the control group represented the status quo, since at that point it was not Uber policy to apologise for bad trips). Some received a basic apology with no elaboration. Some received a ‘status apology’, which included the phrase, ‘We know our estimate was off.’ Some received a ‘commitment apology’, which said that Uber would work hard to give the customer arrival times they can count on. All four groups (control, basic apology, status apology and commitment apology) were then split in two, and half of each group received a $5 coupon they could redeem against a future trip. The economists tracked the passengers’ Uber purchases, the number of trips they took and how much they spent for the subsequent eighty-four days.
Ho and his co-authors discovered a few things from their results. First, apologies are not a panacea. They found almost no effect for the basic apology: just saying sorry had very little impact on the number and length of trips people went on to take. Second, the most effective apology was a costly one: giving people a coupon along with an apology actually led to a net increase in their spend with the company, by comparison to the period prior to the bad experience. Third, apologies can be overused. Some of the customers had more than one bad experience and so received multiple apologies. Those customers punished the company more than customers who never received an apology at all.
This echoed something the hostage negotiators interviewed by Miriam Oostinga mentioned. ‘Saying sorry five times in five minutes won’t make for a positive relationship,’ one of them told her. The more apologies you get from someone, the less costly those apologies seem. At some point they start to feel cheap, even insulting.
* * *
Knowing how to apologise is far from simple, because the same apology can have different effects depending on who we are and what we do. Larissa Tiedens, a social psychologist at Stanford University, has studied how the emotions that politicians display in public influence the way that voters perceive them. In one experiment, Tiedens showed respondents one of two video clips of President Clinton, both extracted from the grand jury testimony he gave on the Monica Lewinsky scandal in 1998 (the fieldwork took place in 1999, when Clinton was still president, and his opponents had begun impeachment proceedings). In one clip, Clinton is visibly angry. He describes his treatment as inappropriate, wrong and unfair, and questions the motives of the investigators. He looks straight at the camera, slashing the air with his hands to emphasise his points. In the other clip, Clinton reflects on his relationship with Lewinsky, and his demeanour is very different. He says the affair was wrong. His head goes down, and he gazes off to the side.
At the time, there was a consensus among media commentators that Clinton needed to show remorse and guilt instead of anger if he wanted to repair his relationship with voters. Tiedens found the opposite: the respondents who viewed the angry clip were more positive about Clinton than those who viewed the remorseful clip. The reason, according to Tiedens, is that ‘anger communicates competence’. Social psychologists have consistently found that people expressing anger are seen as more dominant and competent, even as they seem less friendly, warm and nice. Angry people are more likely to be perceived as high status than sad or remorseful people are. It wasn’t that Clinton’s apologetic style didn’t have a positive effect on how the respondents saw him; they liked him better for it. But the people who saw the angry Clinton respected him more.
This trade-off between respect and warmth makes it hard to judge the right tone to strike when you’re apologising. If you make a status apology (‘I’m sorry, I’m an idiot’) you’re trading away some of your reputation for competence – that is, respect – in exchange for likeability. That can be risky. Whether or not you should do so depends on whether competence or likeability is more important to the relationship in question. Nobody wants to hear a doctor say, ‘The thing is, I’m basically a bit crap at this,’ but a husband or parent should prioritise warmth over authority.
For those who need a measure of both, like hostage negotiators, the question of when to admit error can be a fine judgement call. Some of those Oostinga talked to told her they were reluctant to own up to a mistake unless they had to because of the need to be seen as competent by the person in crisis. For others, an error was an opportunity to equalise an inherently lopsided power relationship. By apologising, the negotiator can show themselves willing to be submissive, which lowers the suspect’s guard and opens up a pathway to intimacy (one of the negotiators told Oostinga that she might even return to an error later in the conversation if she feels its effects lingering: ‘I have a feeling that what I said is still upsetting you’). An error has the potential to bring negotiator and suspect together into a ‘bubble’, where the relationship has time to incubate and deepen. The stakes, the onlookers, the future – the weirdness of the whole situation – all can be forgotten for a while, as the participants pore over what the negotiator got wrong and why. ‘They can bond inside that bubble,’ Oostinga told me.
* * *
Disagreements should be full of mistakes. A disagreement in which the participants plot their every intervention like a chess move and take great care not to say the wrong thing is an arid and passionless affair. It is unlikely to be productive, either; as Oostinga’s negotiators remind us, a conversation without errors is either trivial or robotic or both. Of course, that doesn’t mean you should be happy when you realise you’ve been deaf to the emotions of the other person, when you find yourself talking down to them – or when you get their name wrong. But if this book helps with your disagreeable conversations it won’t be because you have eliminated all the errors you can make; it will be because you are better at recognising those errors and at knowing how to respond to them.
Once you understand how and why disagreements go wrong, the prospect of disagreement’s bumpy, uncomfortable ride seems less intimidating than it might do otherwise. First, because you realise that it’s not just you – that people make similar mistakes all the time, except that usually they don’t recognise them as mistakes. Second, you come to see your mistakes as opportunities in disguise. By correcting your own error – resolving your bum note – you can strengthen your relationship with the other person and make the conversation richer.
An error shakes things up. Or at least it should do. It’s a mini cyclone blowing through the conversation, rearranging the landscape, creating fresh perspectives. It also gives you the opportunity to apologise well, which, as we’ve seen, is much more than a matter of courtesy. An apology should cost you something. I don’t mean that every time you grossly misinterpret what the other person is saying you need to whip out a coupon promising her the next five opinions for free. I mean your acknowledgement of a mistake should be emotionally costly. When you say sorry, it has to mean something other than ‘Let’s move on’; otherwise it’s hard to move on, at least it is for the person who feels offended or badly treated. When you back down from a position, it’s OK to let the other person see how hard it is for you to do so – in fact, it’s better that way.
One of the worst ways to apologise is to say, ‘I’m sorry if . . .’ The ‘if’ immediately renders your apology cheap and insincere, because you’re not admittin
g to a mistake. If you’re not sure that you’ve made a mistake, best not to apologise at all, until you’re absolutely convinced you have.
If it feels bad, that’s good.
11. Disrupt the Script
Hostile arguments get locked into simple and predictable patterns. To make the disagreement more productive, introduce novelty and variation. Be surprising.
In the autumn of 1990, a Norwegian sociologist named Terje Rød-Larsen and his wife Mona Juul, a diplomat, made a trip to Gaza, the battle-torn strip of land on Israel’s border that at the time was home to a million Palestinians. Rød-Larsen was preparing a survey of the living conditions in what was the most densely populated spot on earth. As they were being escorted around a Palestinian refugee camp by a UN officer, the two Norwegians stumbled into a skirmish between Palestinian youths and Israeli soldiers.
Rød-Larsen and Juul froze, terrified, as Israeli bullets whistled past and Palestinian rocks crashed around them. While their escort attempted to calm the situation, Rød-Larsen and Juul found themselves transfixed by the faces of the young men fighting on both sides. They looked scared and defiant and unhappy. Most of all, they looked similar.
Over the next three years, with Juul’s help, Rød-Larsen embarked on a series of visits to the region, meeting with Israelis and Palestinians. Although he met them in his capacity as a social scientist, he had decided to attempt something way beyond his remit as an academic. Rød-Larsen wanted to find a way to help the two sides recognise that the gap between them was not as great as they thought, that the hostile image they had of each other, forged over decades of fighting, was stopping them from seeing each other as they really were – as people with a common interest in peace.
It was moonshine, really, but Rød-Larsen was a stubborn optimist. Jane Corbin, a British reporter for the BBC, wrote of Rød-Larsen that he had ‘complete conviction that anything that is worth doing can be done’. Self-confident without being arrogant, with a ready smile, Rød-Larsen was instantly likeable and had a way of inspiring trust in anyone he met. His maxim was that ‘Sometimes the impossible is easier than the possible.’
Rød-Larsen believed that the road to peace led through the Palestinian Liberation Organisation (PLO). Officially, neither Israel nor the USA dealt with the PLO because they defined it as a terrorist organisation. When a Washington-led peace process opened in 1991, it involved other Palestinian leaders instead. After the fall of the Berlin Wall, there was optimism for a new world order, but by 1993 the talks were already foundering. The USA was not able to play the role of an even-handed mediator, because of both its alliance with Israel and its sheer military and economic might. The Palestinians mistrusted the Americans, while the Israelis railed against the pressure that the Americans put on them.
Rød-Larsen began to wonder if Norway could offer something the USA couldn’t. Norway was a small country, incapable of pushing anyone around. It had good relations with both sides in the dispute. It had its own oil supply, and therefore minimal economic interest in the Middle East. With a population of just over four million, Norway’s comparatively modest size gave it another advantage: small groups of influential people could introduce political innovations.
Mona Juul was friends with Jan Egeland, a former social scientist who was now Norway’s deputy foreign minister. The co-author of Rød-Larsen’s social study of Palestine was Marianne Heiberg, another sociologist, who was married to the new foreign minister, Johan Jorgen Holst. The fluid, informal nature of public life in Norway was in sharp contrast to the vast, bureaucratic and hierarchical structure of American government. In Norway, Rød-Larsen knew everyone and everyone knew him, and some of them were prepared to listen to his crazy idea.
Rød-Larsen proposed peace talks in Oslo, separate from the official negotiations in Washington. These talks would be hosted by the Norwegians, and they would be secret. There would be no grand diplomatic ceremonies, press conferences or fleets of limousines. Above all, there would be no playing to the gallery. Rød-Larsen had observed that the bright public spotlight on the Washington peace process had the effect of polarising the conversation. The Israelis and Palestinians were always aware of the audiences back home. There was intense pressure on them to maintain their ‘face’: the negotiators felt compelled to project strength above all, which made it very hard to exhibit flexibility. As a result, the two sides did not truly engage with each other. Instead, they staked out their positions and dug in, making the same predictable moves and counter-moves as in previous negotiations. The script might as well have been written in advance.
* * *
Peter Coleman, a professor of conflict resolution, runs the Difficult Conversations Laboratory at Columbia University. Coleman and his team have analysed hundreds of encounters between people with opposing views. They study the emotional dynamics of conversations; how they flow and how they get stuck. Then they graph them.
The lab employs a similar methodology to the one pioneered by relationship scientists. It matches strangers with strong and opposite views on a polarising issue and invites them to talk the issue through. Afterwards, each participant is played a recording of the conversation and asked to say how he or she felt at each moment. The results are coded for positive and negative emotions, and for particular thoughts and behaviours. The conversations can get heated; sometimes they have to be brought to an end early. Others go much better.
Coleman and his colleagues have identified a key difference between the more destructive conversations and the more constructive ones. The destructive conversations get locked into a tug-of-war dynamic early on and stay that way, becoming progressively more bad-tempered. Each person aligns themselves with a side, and each blames the other’s side for the world’s ills. Meanwhile, the constructive conversations are not necessarily serene or well mannered – they can involve verbal attacks and bad faith, and the participants can report feeling hurt and annoyed – but, at certain points, the participants are able to escape or subvert the dynamic. Positive emotions, like amusement, empathy, and insight make appearances, if only fleeting ones. The conversation is more expansive. It has variety.
‘The more constructive dyads’, Coleman reported, ‘thought about the issues in more complex, nuanced and flexible ways. They felt many different types of emotions, both positive and negative, over the course of the discussions. And they behaved in more varied ways that demonstrated a greater degree of openness, flexibility, and curiosity in addition to a strong advocacy for their positions.’
When Coleman’s team map the data on the emotions from each conversation on a grid, the shapes generated by constructive and productive conversations look very different. The constructive ones are, literally, all over the place, messy constellations of dots. The destructive ones are straight lines, like ruts. It’s as if the participants in the destructive conversations have had their emotional range reduced to a narrow band. They have become wholly predictable.
In session 1 (destructive), the dots are tightly organised along two lines, indicating a narrow emotional range of the discussion. In session 2 (constructive), the dots show that the emotions of the participants fluctuated back and forth as the discussion unfolded. Coleman measured thought and behaviour as well as emotion. On every dimension, the constructive conversations were more complex than the destructive ones.
Software engineers who design chatbots distinguish between ‘stateful’ and ‘stateless’ conversations. A stateful exchange is one in which the participants retain a memory of what is said during the conversation. Stateless dialogues are ones in which little or no conversational history is retained, and each new remark responds only to the last. They are so low context that they don’t even have the context of the previous conversation.
For obvious reasons, it’s easier to design bots for stateless conversations: picking an appropriate response to a single cue requires less processing power than attempting to engage with the flow of a conversation. The trade-off, for programmers, is that stateless bo
ts sound ‘robotic’, dispensing pre-scripted replies without any indication that they really know what the conversation is about. But it’s not as a big a trade-off as you might think, since a lot of human conversation is stateless.
Have you ever had an argument like this?
A: I really enjoyed that book.
B: Oh really? It’s very poorly written.
A: Why do you have to make me feel bad about what I like?
B: Why do you have to play the victim all the time?
A: Oh, that’s rich – you’re always playing the victim.
B: Look, you’re clearly just in a bad mood today.
A: I’m the one in a bad mood?
And so on. In an exchange like this, each remark is only about the last remark. The conversation has almost no memory of itself. Neither partner is learning anything from the other and both are becoming increasingly bad-tempered.
In 1989, an undergraduate computer programmer at the University of Dublin called Mark Humphrys wrote a chatbot program he called MGonz. Whenever MGonz lacked a clear cue for how to respond to a remark, it threw in an insult, like ‘You are obviously an asshole,’ or ‘OK thats it im not talking to you anymore,’ or ‘Ah type something interesting or shut up.’ When Humphrys left the program connected to his university’s computer network overnight, he returned in the morning to discover that somebody had spent an hour and a half engaged in an argument with MGonz, obviously convinced that he was debating with a real person.