[mlthoughts2018]

I think a precondition of the entire discussion is that we’re talking about what’s more effective in resolving disagreement - which would be measured by whether a disagreement process approaches the conclusion that would be reached between two rational agents sharing their information.

Note that I am not saying I believe humans do act that way, rather that is the yardstick of outcome to decide if one disagreement resolution technique is better than another.

[harimau777]

Maybe the issue is different views of what the disagreement is over?

I tend to view the disagreement as being over policy not over rationalism/facts. If at the end of the discussion both people agree on the policy I see that as successful even if they still disagree on the evidence that supports that policy.

[mlthoughts2018]

I don’t think so. In fact there is actual theory about all this.

- https://en.m.wikipedia.org/wiki/Aumann%27s_agreement_theorem

- https://www.scottaaronson.com/papers/agree-econ.pdf

and in particular,

- https://mason.gmu.edu/~rhanson/prior.pdf

If your goal is just to manipulate someone else into doing what you want, regardless of what the facts say nor what their beliefs say about the facts or about your beliefs, then sure, that is pure manipulation and not related to strategies for effective disagreement resolution.

It’s also not relevant or useful to bring up the fact that humans are not rational Bayesian actors, again because we’re talking about how best to achieve unified shared understanding of the true facts and to ensure we only derive beliefs from that (anything else can only be manipulation, not decision resolution, by definition). We may fall short of the best solution due to our flaws and limitations, but that doesn’t render the best solution any less the best, not make some other seemingly more attainable solution better or more worthwhile.