(epistemic status: fairly speculative)

Recently, I've been thinking about where the main benefits of conversation lie, and I've come to some interesting early conclusions.

Most of the benefits of private conversations come from effects they have on the immediate participants. However, this doesn't necessarily hold true for other forms of conversation. For instance, in a public debate, there is often little benefit to the debaters themselves, preparation aside [1], but often a benefit to the audience. On the other hand, in a public double-crux, the participants as well as the audience can benefit.

One area that I think this bears special interest for is that of online discussion. Online discussions are in some cases private, but because of the permanence of online conversations I feel they often have much more of a "bystander benefit" than conversations in person do - the arguments being offered are available not just to those who happened to be in the room at the time, but to many others who can view the archives.

For instance, I think most of the benefit of comments on LW doesn't come from the immediate interaction between the commenter and the original poster, but rather from their discussion being read by others. One piece of evidence that helps support this is how frequently many more people vote on a particular comment than are active in the conversation itself - this indicates that the people active in the discussion are often a small fraction of those reading, and when you consider people who don't vote, who "lurk", or who don't have accounts the number of readers probably increases further.

Real-time text conversation, like that offered by Slack, Discord, or IRC, is perhaps "in between" in-person and asynchronous online communication. However, I tend to think that similar assumptions hold, at least on public servers - there are many more "lurkers" who read but don't necessarily comment than there are active discussion participants. On Slack or Discord, which allow emoji reacts, this is especially apparent - I frequently see reactions added to conversations well after they have concluded, indicating that people have read the conversation and reacted asynchronously. [2]

One potential counterargument to this theory is that, while many more people may interact with arguments or statements than those directly involved, those who are directly involved probably derive much more of an impact from interaction. I think that this is likely somewhat true, but even if you assume a large multiplier - let's say 10x more value, which I think overstates things considerably - for actively participating vs. "lurking", there are very easily many conversations online where more than 10x the number of people lurk or observe than actively participate.

One important result of this is that conversations that may not seem all that productive or informative to the immediate participants can actually be highly valuable to less active users. This can be especially true for "101" type discussions that might seem basic to the participants (likely more experienced people), but can provide helpful context to lurkers. For instance, I recall commentating on a live stream of a strategy game where the players were taking a while to get into the action. In order to fill space in my commentary, I started explaining basic aspects of the strategy and composition of this game.

Surprisingly, multiple people later commented or messaged me saying that the information I had been broadcasting to fill time was very helpful to them and they wished that more content like that was out there! The most vocal parts of my usual audience might have been more experienced and maybe even bored by these sorts of "basics", but there were also people who hadn't encountered much of this before, and so the "101 content" I was broadcasting to fill time ended being actually very useful - maybe more so than my usual "higher level" stuff!

The end result, in my view, is that one should keep in mind that, when having conversations in more permanent media, much of the benefit of the conversation is likely not with your direct interlocutor, but with others who may be watching or reading without commenting themselves, even after the fact. As a result of this, conversations - especially basic or "101" conversations - can be helpful and worth having even if they don't necessarily feel that productive in the moment.

[1] I cannot recall any instance where a participant in a formal debate I am aware of changed their mind as an immediate result, even when they very clearly lost - though it's of course possible that one could mull over the arguments for a while and eventually change positions.

[2] Interestingly, on a more active channel or server this effect is probably diminished somewhat, as with more and more messages being sent it becomes more difficult to view the entire backlog so people are less likely to have viewed a particular conversation. However, this is likely mitigated by an increased number of "lurkers" who see things in the moment.