(content note: discussion of war and mass death; also a long aside about the philosophy of apologies)
After 100,000 people were killed in the Bosnian war, the three sides eventually met for mediation in the United States. To be clear, the attempted negotiators for most of the war were the EU and the UN, but after the Srebrenica massacre (8,000 men and boys killed) in an area that the UN had designated a 'safe area', the US was shocked and felt that if they didn't take responsibility for ending the war, the war would never end. (It had also shaken things up enough that NATO got the political backing to take independent military action in the region, so all sides involved felt more threatened and like they might lose.)
Richard Holbrooke, the assistant secretary of state for Europe, was experienced in getting agreements, and felt the US had an obligation to act. He requested meetings with the Presidents of the different sides, which they accepted. To then get them to the point of being willing to meet and hash out an agreement, he traveled between the Premiers some dozens of times in a 6-8 week period, what's known as 'shuttle diplomacy', where you play go-between. This alone was enough to agree on some broad points—a single Bosnia (rather than being split up between Serbia and Croatia), two governments, and splitting the land in a 51/49 split (the Serb forces had controlled like 60-70%, so they were losing a lot; they wanted to still be able to say that they got the majority of the land).
However they still had to hash out a lot of the details, amongst sides that hated each other. They had to figure out the details of the boundary lines between the two territories, the timetable for ceasefire and troop withdrawals, the rights of refugees, and both sides had to come up with the constitution for this new country with three presidents.
So Holbrooke brought them all to a US air-force base in Dayton, Ohio.
The secure site was chosen in order to remove all the parties from their comfort zone, without which they would have little incentive to negotiate; to reduce their ability to negotiate through the media; and to securely house over 800 staff and attendants. Curbing the participants' ability to negotiate via the media was a particularly important consideration. Holbrooke wanted to prevent posturing through early leaks to the press.
The ground rules included no speaking to the press, and also that all parties had full permission from their congresses to make agreements; all the decision-makers were "in the room".
After 21 days of negotiations, they reached the Dayton Peace Agreement, which has held to this day.
I have had a similar experience to this story. I was once in a conflict where I felt another party had wronged me. Our contract specified a mediator, who we brought in. That person said that the other side owed my side recompense, but nothing much happened.
Later on, a different party with significant power over the both of us (though relatively more power/responsibility for the other party) heard about the situation, and came to my side to check what they'd heard. They then arranged some meetings where actual details were determined, where we argued about whether any wronging had occurred, and (when the mediator decided it had) a settlement with recompense was agreed.
I'm grateful for the intervention; nothing was happening otherwise, and I had been wronged. I'm still a little angry about it, but I feel that the counterparty ended up honoring our contract.
(Coincidentally, it was also the case that in my story the first mediator was European, and the second mediator was American.)
To explain one reason why I think mediators help, I need to talk a little about apologies. In my opinion, many rationalists have a mistaken philosophy of when to say 'sorry'. As a single datapoint, a friend of mine showed me a draft blogpost where they explained that they should apologize when they executed the wrong algorithm—when they should have known better—but not just if your decisions imposed costs on others. They say that an apology means the following:
- I made a mistake in calculating the expected value (EV) of an action I took;
- I understand what my mistake in that calculation was; and,
- I’ve updated my decision-making process to correct that style of mistake.
This is all independent of whether harms were caused; having bad outcomes is totally irrelevant to an apology (or acceptance thereof).
I think this is missing half of the reasons to say sorry. (And no, I'm not talking about the sympathy sorry, where you say "I'm so sorry for the loss of your family member", which is entirely separate from responsibility.)
I think of the two reasons for saying sorry as taking responsibility for process or for outcome.
The above is a process apology—the process I executed was mistaken and caused harm, and I am responsible for my process.
The other kind is an outcome-based apology. It's never really demanded, but often a good person will do it anyway. It means "We interacted, and it imposed costs on you. I'm going to take responsibility for imposing those costs." For instance, if I ran a big conference, and then on the day of it there was a freak rain or earthquake rendering the conference venue unusable, a process-purist might say "Oh, well, it seems that even though I made entirely reasonable decision, we simply do not have a conference any more. How terribly sad! Okay, goodbye." Whereas outcome-apologist might say "I'm so sorry that you trusted me to bring you a conference, and it has not worked out. I'm going to make it up to you, me and my team will work extra hard to find another conference venue for tomorrow, or else we will reimburse all of your tickets, and/or do some secret third thing to try to make it up to you."
I think it's often good that people take responsibility for costs that they have imposed on others—even if it was unintentional, and not through any fault of their own. This helps show that the other party cares about your interests through interacting with you. And that helps you develop a trusting relationship, and improves your ability to make positive-sum trades.
Outcome-based apologies are a common kind of conflict, where either side could take responsibility for the cost, and the counterparty is often arguing that you should bear their cost, even though you didn't actually make a mistake. When they're asking for such an apology, it is a situation where there are incentives for the other party to make this claim, so it's hard to trust their view to be unbiased, but if some neutral third-party (without this incentive) says that it looks to them like it's natural for you to take responsibility for the cost, you're much more likely to believe them—and care about acting well according to their perspective.
Insofar as you're in such a conflict—especially so if you're arguing this symmetrically back at them!—then adding a mediator is a great way to get out of this stalemate. When they say "My read of the situation is that it would be reasonable for you to take responsibility and apologize here", you can take that at face value. And then it's much easier to apologize.
There are a few reasons why I do not see mediators help resolve conflicts more often.
Lately I have been looking for opportunities for mediators, and trying to overcome these problems. I have often seen people dismiss third parties as relevant, and say they just want to talk directly with the person they're in a conflict. I think this is kind of dumb, and that has reliably been the point whereafter I felt that the situation was unrecoverable, as the two parties had been utterly failing to successfully communicate and won't take this natural step for getting back on the same page.
(Do I have any advice for picking a mediator? Well, pick someone with a backbone who'll tell you if they think you're wrong, and good enough judgment that you'll take them seriously when they say so.)
Mostly this post is a plea to accept and use third-party mediators more often (i.e. to solve problems 1 & 3). Hopefully if the first reason is the problem, then just thinking to do it will help resolve more conflicts.