Explore/Exploit for Conversations
Something status: I’ve started to try on an explore/exploit lens for thinking about conversation, and I like it. Here’s some rough ideas. On the editing pass of this, a feel like I haven’t properly separated the important ideas explicitly enough, but it’s fine.
You are in explore mode if you are introducing ideas/topics to the conversation and aren’t sure how much the others will enjoy them. You are in exploit mode if you are talking about stuff that you already know everyone will enjoy.
Being in explore mode requires feeling “comfortable” with your group, and being in exploit mode requires having Common Knowledge (CK) of Common Interests (CI).
Examples:
Explore
In your anime club’s board meeting you talk about the meta of how board meetings usualy go.
You got to a LW meetup and talk about a new beat-boxing technique you’ve been practicing.
You tell your grandma about this heavy metal band you’re getting into.
Exploit
In your anime club’s board meeting you talk about what new shows you want to watch at club meetings.
You got to a LW meetup and talk about x-risk.
You tell your grandma about how school/work/family is going.
I tried to pick examples that weren’t secretly saying “One of these is way better than the other!” I also want to be clear that being in explore mode doesn’t have to be super edgy or out there. It just needs to be a situation where you don’t know what response people will have before you go there.
Claim: For a group (2+) of people to have good conversation, you either need to have Common Knowledge of juicy Common Interests to exploit, or you need to be comfortable spending time in explore mode.
Explore Mode
I think most people experience that it is easier to have explore mode conversation with smaller groups, and with people that they know well.
I claim that it feels comfortable to be in explore mode if every group member has:
A) Reason to believe they won’t get trapped in a boring conversation. (i.e anyone can steer the conversation somewhere new)
B) Reason to believe that if they are boring people, someone will let them know in a non-hurtful way.
If A) and B) hold for the group, I’m going to say your group has “explore-trust”. Trying to explore when you don’t have explore-trust can lead to problems:
Someone starts ranting about something far outside the realm of CI, I start to get bored, and I also don’t feel comfortable changing the topic. If this happens frequently, I don’t want to chat with this group of people.
I’ve been in groups where it seemed like everyone was so worried that they might be boring people, that no one talked for more than three sentences in a row. This lead to lots of things I was interested in not getting explored.
If on several occasions me bringing in a new idea is met with, “That’s boring as shit dude,” and I don’t take it well, I’m going to stop trying to explore.
Having explore-trust seems to require the group members having some knowledge of and experience with each other. You can’t just declare, “We all have explore-trust now!”
Say you tell new members, “We have a rule were if you’re ever really bored of the current topic, just blurt out that you’re bored.” Even if the new member sees some other people acting on that rule, you haven’t necessarily convinced them it’s safe for them to try it. It could totally be the case that there’s some weird status thing where the people are using the “i’m bored” rule are the ones no one wants to offend, but if the new person uses the “i’m bored” rule they won’t get as welcome a reception.
Claim: The bigger the group, the harder it is build explore-trust.
From my personal experience, it seems like there is an amorphous changing point at which a group goes from, “some people hanging out” to being A Group
When dealing with A Group I feel extra negative pressure away from steering “too much”. This varies tons based on the situation. I’m generally cool with steering if I’m in an explicit leader position. I’m also more inclined try and steer if I get a sense that “no one else is going to”, but it’s still easy for small inferences on the group (“wait, Joe is frowning, is that cause he doesn’t like that we didn’t vote?”) to make me more hesitant.
(Rough guess, I get the sense that many self identified rationalists are either extra-oblivious, or extra sensitive to “groupyness”)
Exploit Mode
If you’re doing conversation in exploit mode, somehow you’ve created Common Knowledge of Common Interests. This could have happened because your group formed under a strong self selecting context. If I post an add on meetup.com for “people who like to make figurines of the characters from Fight Club out of match sticks” odds are that if anyone shows up, I’ll have some very specific Common Interests with them, and it will be Common Knowledge by the fact we are both at this meetup.
Except, it’s not quite that easy to make Common Knowledge. It’s totally possible for someone to show up at my figurine meetup for reasons other than, “I have a passion for making Fight Club match stick figurines”.
“I like figurines in general and there are no other figurine making meetups”
“I really like Fight Club”
“I wanted to check if this was actually a real meetup”
You could also have either some structure or a leader that maintains a consistent conversation experience, and over time group members will self select and the one’s still a part of the group will likely have pretty good Common Knowledge of Common Interests supported by the structure/leader.
You also have Common Knowledge created from past explore mode conversations your group has had.
The last thing I want to say about exploit mode is that you always have some level of CK about your CI with any group of people, so really it’s a matter of creating sufficiently strong CK about sufficiently interesting CI to have a good group. With any person on the street you have CK about about your CI of not starving to death tomorrow. You just probably aren’t going to be making a club about it any time soon.
The above idea makes me think about interfaces, and specifically the Melting Asphalt post on the personhood interface. Interacting with someone via the personhood interface is less rewarding than interacting via the best-friend interface, but it has the benefit that you can use it with almost any person you meet.
One Particular Failure Mode
Rationality is not one thing, and people have wildly differing goals. I’d peg the rationalist interface as, “vaguely intellectual/insightful/clever, probably contrarian, possible relation to tech, some facet of nerdiness to them”.
I recently attended a meetup where it felt like we never built any more specific CI. Conversation at the level of exploiting the rationalist-interface has a feeling of, “ooh, what’s some clever vaguely relevant thing I can say next to prevent a long awkward pause?”
This can get real old real fast. When I noticed what was happening, I had the thought to try to make an explicit steering move to have us make some CK about more interesting CI, but I felt a bit too unsure about if I would be stepping on the toes of the host of the meetup. The host also wasn’t doing any steering, I’m assuming because they didn’t like the idea of forcing a structure on people.
Questions to Explore
Do you have strong enough Common Knowledge about sufficiently interesting Common Interests for your group to be enjoyable with no explore mode?
Does your group have explore trust? If you think you’re group has explore trust, would others in the group think you have explore-trust?
If you’re bringing a new group into existence, be ready to appeal to some structure, or maybe at least some structured explore to find some sort of CI to exploit. If you find that people have more explore-trust than you expected, ease back on the structure.
Think about your trajectory. Does it feel like your group will build more and more explore trust over time? If not, what could you do to change that?
I really like this framework! I’ve noticed that if someone makes a comment that assumes everyone in the group has CI, but I’m not sure if everyone does, I get a sense of awkwardness and feel the need to model two conversations: the one happening assuming everyone has CI, and the one happening assuming at least one person doesn’t. This has the unfortunate side effect of consuming most of my thought-bandwidth, which makes me boring and quiet even if I would have otherwise been engaged and talkative.
I’d never thought about it clearly, so thanks for this model.
A behavior I’ve observed (and participated in) that you don’t mention: the group can temporarily splinter. Picture 6 people. Someone explores topic A. Two other people like the new topic. The other 3 listen politely for 1-2 minutes. One of the three bored people explores topic B, addressing a bored neighbor (identified by their silence). The third bored person latches on to them. Then both conversations evolve until one dies down or a leader forcibly merges the two.
(By forcibly merge, I mean: while participating in conversation A, listen to conversation B, wait for a simultaneous pause in both, then respond conspicuously to conversation B, dragging most conversation A participants with you. I have observed myself doing this.) (Single participants can also switch freely.) (I have observed this to work with close friends and relative strangers, but obviously strangers need leaders/confident people to start new conversations, because they have to be in explore mode.)
I think this lowers the exploration barrier, compared to your model.