I've noticed that one baby step towards this that I like is to be less 'Markovian' and try to check if the leaf nodes of the conversation are relevant to their great-great-grandparents or something (really, whatever ancestor it's easiest to check this for).
For bulk exchange of information and state, holistic is really good. I strongly prefer the holistic approach, but I've found that it only works for entirely friendly conversations. If it's adversarial, I find that branches get aggressively pruned to just the things that the opposing side can most easily attack.
And if you think about holistic being optimized for "exchange of information and state", this makes perfect sense: adversarial conversations are rarely if ever about information exchange; they're about "winning".
It's also perhaps worth mentioning that aggressively holistic comms can require more mental horsepower than some people have readily available or are willing to invest, so it's best to tune the level of threading based on audience and discussion type.
For bulk exchange of information and state, holistic is really good. I strongly prefer the holistic approach, but I've found that it only works for entirely friendly conversations.
This is a good point. The holistic approach can come off as really adversarial if the conversation is just a little adversarial, because my blindsight guess can be pretty ugly, or at least, can come off that way. I have witnessed a fairly friendly seeming conversation where a holistic-responder responded to a decision theory point by bringing up trauma and abuse (bringing up episodes in the past of both conversation participants which you'd normally expect to be really sensitive) and offered a perspective which could at least very easily be confused with "it's your fault you were abused". If the conversation hadn't been very friendly, this could have gone extremely poorly. Holistic responses have a tendency to
If it's adversarial, I find that branches get aggressively pruned to just the things that the opposing side can most easily attack.
Yeah, I often find that I have to prune less-important threads preemptively, out of concern that my response on branch X might seem like an easy target and thus cause branch Y to get pruned by the other person, when I think branch Y is more important. To be honest, I probably do not do this as much as I ideally should. (I want to respond to all the threads!)
So a level of non-adversarialness is also required to really maintain good branching conversations.
A kind of conversation which I very much enjoy is one where everyone has time to formulate long thoughtful responses, and most of the conversational points are remembered and addressed later.
For example, a long email thread between Alice and Bob, where Alice and Bob write their next emails as point-by-point replies. In caricature:
(email 1, from Alice:)
(Email 2, from Bob:)
(Email 3, from Alice:)
And so on, with Alice and Bob continuing to expand sub-points like that, managing an increasing number of threads in parallel. (Not strictly increasing! Some will usually be dropped pretty fast as they reach a natural conclusion or don't feel relevant to the interesting parts of the discussion.)
If the conversation is a disagreement, you can see how this structure enables a thorough examination of all the possible argument paths. It's a little like an HCH tree, except in practice the parallel conversations influence each other heavily, because the related ideas in the different threads will bear heavily on one another.
It's a very detail-oriented conversational mode. Sometimes you get the feeling of being stuck in the mud, bogged down by all the different points you're trying to manage.
At times like that, what I usually do is take a step back and consider the big picture. What's really going on with this conversation? Are there any patterns which seem to come up again and again (even if they're very vague/abstract)? This feels like taking a break from trying to understand the explicit content of individual sentences and paragraphs, and instead, focusing on making a mental model of my conversation partner. What are the deep generators of the conversation we're having? Is there some broader picture my conversation partner is trying to convey?
It feels similar to blindsight guesses: I may feel I have no idea what bigger picture the other person is getting at, but I ask my brain for an answer anyway, hypothesizing that my brain must have some mental model of the conversation.
And then I try to articulate that picture (somewhat like Gendlin's focusing), ask where I'm right/wrong, and perhaps give my hypothetical response supposing I'm right.
Sometimes I'm surprised by how correct my guesses are, but I'm also often (very) off the mark. Either way, it usually opens up new threads which have a good chance of being more fruitful than the old ones.
Let's call this conversational move the "holistic response" -- it feels like blurring your eyes a little bit to ignore details and take in a larger shape, or stepping back from a painting to see the composition rather than the brushwork. Often it seems to grant the ability to skip several steps ahead in the arguments, or bridge inferential gaps which were previously blockers. It also helps snap one out of arguments-as-soldiers mentality and exercise empathy, trying on the other person's perspective.
This seems slightly related to double crux. Double crux was invented precisely to focus on the important points rather than going down endless rabbit holes of conversation, and it involves taking the other person's perspective and holding it next to your own
Why wouldn't I always take the holistic approach?
Well, I've been in conversations with someone who seems to "always" (ie, very often) take the holistic approach. It's definitely useful at times, but it often frustratingly drops the more detailed point-by-point analysis which could well go somewhere. Besides, without first taking the detailed approach, there isn't that much fuel for the holistic guesses. (Also, the holistic approach is more cognitively demanding.)