Trio Walks, Duo Talks

Epistemic Status: Exploration

Related: The Engineer and the Diplomat

Having a great conversation is hard, but valuable. The structure a conversation is built around determines how it will go. How should we engineer more valuable conversations?

At the retreat, it was suggested that we go on Trio Walks. The idea was that three is the ideal number of people for a conversation about important things. With only two, it is easy to encounter blind spots, make mistakes or lose your way. Having a third keeps things grounded.

The walking part was partly that we had nature all around us, partly that at some point we needed to get some exercise, but mostly to get the trio away from everyone else.

I only got the opportunity to go on one trio walk. Once we were clear of other people we did not do much walking, but we did stay a trio. I found it to be an excellent use of time, including the chance to meet two new people. Last week I arranged a trio walk with two people I wanted to connect, and that seemed to go better than it would have with a different number of people.

World of Two

This conflicts with my previous model, which said the right number of people is two. I thought the trade-off was quality against efficiency; with each additional person, things get a lot harder. The way large conversations happen at all, under this model, is most of the people are mostly observers.

With two people, the conversation can stay on track. Pauses to think are more practical. When one needs time, it is often obvious, and if it isn’t, one can ask. You can craft a multi-part argument. You can leave threads hanging to come back to later. If things get off track, it is relatively easy to steer things back on track.

With two people you can retain control of the conversation without the worry it will be hijacked, or you will cease to be participating.

Only one person needs to understand any given thing being said, so explanations can be tailored to that person. Different explanations work for different people. You can know that person’s background knowledge, available reference points and shared metaphors. If they are not sure they understand, they can say it back in their own words so you can determine if they have it right, and often when they do not, what they do say is also interesting. If the explanation does not work, it is relatively cheap for that person to say so, and you can try again.

Only one person needs to be convinced by any argument. Often they already agree and no argument is required. Different arguments convince different people, so making the case becomes much easier. Citations are less often needed.

Within a conversation, status must be managed. With two people, it is much easier to maintain equality, and thus allow communication; this problem becomes exponentially more complex with more people. If inequality is desired or necessary, that is easy. Without anyone else to impress, status can shift to help the conversation without worrying that one is making a status claim, in general. Without anyone else to embarrass yourself in front of, you can risk sounding stupid or ignorant. Tracking such things, and attempting to change them, takes up a lot less of everyone’s brainpower.

Your chances of keeping things private is much better. “Three people can keep a secret if two of them are dead” implies an exaggeration, but not a large one.

Three’s Company

Adding a third person threatens to disrupt all of that. To some extent, this is good, since the two person conversation is in some sense cheating. There are also additional benefits. How do we get the benefits, but maintain what is good about a two person conversation?

I believe that the key to a successful important three person conversation is to be a (1+1)+1 conversation, rather than a 1+1+1 conversation. By maintaining a level with two participants, the good things about a two person conversation can be preserved.

There are two ways I know of to split this group of three: You can have one person assist, or one person lead.


The assistant has several jobs.

The most important job is to track the stack of discussion items. It is easy for a conversation to veer off into a side topic and then into another, never to return. Sometimes that is good but often it is bad. It can happen because both participants get caught up and forget what they were discussing, or because neither party can suggest popping back out without conceding the side point. The assistant can remind the primaries to pop back up the stack to the central point.

The flip side is the assistant can and should observe when the conversation is no longer interesting or fruitful, which is easy to miss and can be awkward to state, and transition to something more interesting. This is especially true when an impasse is reached or the two parties are in violent agreement.

The second job is to help with explanations. If one party does not understand the other, the assistant can either explain the point or note that they too are confused. Often when an explanation fails, a third perspective is in a better position to figure out where the confusion lies, or has a better idea for an explanation. The assistant can also note when they are confused, and often they will not be alone in their confusion.

The third job is to point out mistakes and factual errors, and help settle some disagreements. With two people, the chances are mistakes will be missed at some point. An additional error check goes a long way, and allows more confidence in both data and arguments. This includes pointing out implicit assumptions that need to be explicit. The assistant can also help when the primaries disagree, especially on a factual question (including if participants can’t remember what was said), but also with opinions.

The assistant can also help offer a prompt if either participant is searching for a word or concept, or things are otherwise stuck, but this should be kept as short as possible.

For the most part, the assistant’s job is to listen, absorbing the conversation on all levels, and stay out of it. Both primaries need confidence that the assistant will not lightly attempt to assume a primary role, or the dynamic you want will break down. Sometimes the assistant should tag themselves into the conversation, but never lightly. It must be the only path forward.


Here, one participant is in command, and the other two combine to fill the other role. The same way the assistant should be super careful about tagging themselves into a primary role, the respondents need to respect the leader role. This can take several forms. The two most common are mirrors.

In one, the leader is conveying information, attempting to answer questions and/or explain ideas and concepts. In this context, the leader is automatically high status. Neither respondent is primary and either can ask questions freely; if one steps back for a while, then they become the assistant instead. Stepping up from assistant to respondent is also possible, and should have a lower threshold of action than the assistant attempting to become primary. The leader does their best to explain, with the expectation that they will have the time necessary to do this in multiple steps and to take a winding path if they feel it is necessary or prudent.

In the other, the leader is seeking information and/or feedback, and the respondents work together to provide it. This can be anything from poking holes in (or improving) a new theory to explaining a difficult topic. The leader can be low status, but all are there to help the leader, and allow the leader to steer the conversation to what they need. This security is important, so the leader can take the time to go slow, get things right, be free to ask embarrassing questions, and not worry about going into details without getting the conversation hijacked.

These two sub-modes can combine, especially when the leader wants to explore something but first needs to bring the respondents up to speed on background information and/or the leader’s thinking or new ideas.

Three for All

If all three participants are primary, either more attention must be paid to steering the path, everyone needs to be indifferent to where things go, or some combination. There is nothing wrong with that, but it is better for modes like ‘friends joking around’ or ‘tell random old stories’ than important conversations. If there is what you want, go for it! If there is a goal, my experience says this form imposes large costs.

Four is a Crowd

What happens when you add a fourth person? Risk of hijack goes up, and potential social dynamics and necessary social calculations multiply. It becomes harder to avoid a level of conversation that is not about its stated topic, and that level looms larger. The best way I know to minimize this is to retain the concept of primary roles.


You now have multiple assistants, or three or more respondents, which can be called an audience. Like the assist and leader cases, this is structurally a two person conversation. Sometimes you ask the audience for help, but the audience must be reluctant to interrupt. Each audience member needs to be more reluctant to interrupt as the audience grows, so the rate of interruption stays constant or declines, and audience members need to be damn sure before attempting to become primary participants. Around size eight, raising hands becomes worthwhile.

When the primary or primaries are taking questions from the audience, that is a generalization of lead rather than assist, and can be thought of as a two member conversation between the primaries (the panel), typically as the leader, and the audience as the respondents. In rare cases leadership can flip, with highly variant results.

A special case is where there is one person who does most of the talking, and others who want to listen or interact, so an audience grows up around someone ‘holding court.’ This is especially likely when a higher status person is at a gathering. It can be a failure mode but it need not be.

Topic as Assistant

A generalization of the assistant role is where there is a topic, and the group is discussing this topic, with the assistant only speaking to keep the group on target. This is the default goal mode for a discussion meetup, and is often necessary in larger groups to avoid people not getting a chance to talk or feeling left out. It comes at the price of making it very difficult to pursue complex paths, but I don’t know a way around that except to split up the group. Defaulting to that split more often has been a good change.

Free for Failure

The structures need not be explicit. They do need to be clear to everyone. Otherwise, conversations are mostly about themselves and about social and status relations. Some of that is fine, but overloading on such matters is almost inevitable in life. The good news, from my perspective, is the viability of the group of three.

This entry was posted in Good Advice, Rationality and tagged , , , , . Bookmark the permalink.

2 Responses to Trio Walks, Duo Talks

  1. Pingback: On Automoderation | Don't Worry About the Vase

  2. Pingback: Best of Don’t Worry About the Vase | Don't Worry About the Vase

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s