I think the Vassarian–Taylorist conflict–mistake synthesis moral is that in order to perform its function, the English court system needs to be able to punish Raleigh for "fraud" on the basis of his actions relative what he knew or could have reasonably been expected to know, even while Raleigh is subjectively the hero of his own story and a sympathetic psychologist could eloquently and truthfully explain how easy it was for him to talk himself into a biased narrative.

Where mistake theorists treat politics as "science, engineering, or medicine" and conflict theorists treat politics as war, this view treats politics as evolutionary game theory: the unfolding over time of a population of many dumb, small agents executing strategies, forming coalitions, occasionally switching strategies to imitate those that are more successful in the local environment, &c. The synthesis view is mistake-theoretic insofar as the little agents are understood to be playing far from optimally and could do much better if they were smarter, but conflict-theoretic insofar as the games being played have large zero-sum components and you mostly can't take the things the little agents say literally. The "mistakes" aren't random and not easily fixable with more information (in contrast to how if I said 57 was prime and you said "But 3 × 19", I would immediately say "Oops"), but rather arise from the strategies being executed: it's not a coincidence that Raleigh talked himself into a narrative where he was lone angel who would discover limitless gold.

Agents select beliefs on the basis of either their being true (and therefore useful for navigating the world) or because they successfully deceive other agents into mis-navigating the world in a way that benefits the belief-holder. "Be more charitable to other people" isn't necessarily great advice in general, because while sometimes other agents have useful true information to offer (Raleigh's The Discovery of Guiana "includes some material of a factual nature"), it's hard to distinguish from misinformation that was optimized to benefit the agents who propogate it (Discovery of Guiana also says you should invest in Raleigh's second expedition).

Mistake theorists think conflict theorists are making a mistake; conflict theorists think mistake theorists are the enemy. Evolutionary game theorists think that conflict theorists are executing strategies adapted to an environment predominated by zero-sum games, and that mistake theorists are executing strategies adapted to an environment containing cooperative games (where the existence of a mechanism for externally enforcing agreements, like a court system, aligns incentives and thereby makes it easier to propogate true infromation).

I don't understand how the second paragraph follows from the first at all.

This account of Walter Raleigh’s life seems… misleading, at best (and in parts just plain inaccurate)—assuming, that is, that we can trust the Wikipedia page. There seems to be quite a bit of conflict (of interpretation, at least) between that page and this one about Raleigh’s book.

I don’t think we should draw any moral from this story, without first thoroughly verifying it from reliable sources. As it stands, we have several Wikipedia pages, which paint a murky and contradictory picture (and some of which are inconsistent with Scott’s summary).

Would your views on speaking truth to power change if the truth were 2x as offensive as you currently think it is? 10x? 100x?

For some multiplier, yes. (I don't know what the multiplier is.) If potentates would murder me on the spot unless I deny that that they live acting by others' action, and affirm that they are loved even if they don't give and are strong independently of a faction, then I will say those things in order to not be murdered on the spot.

I guess I need to clarify something: I tend to talk about this stuff in the language of virtues and principles rather than the language of consequentialism, not because I think the language of virtues and principles is literally true as AI theory, but because humans can't use consequentialism for this kind of thing. Some part of your brain is performing some computation that, if it works, to the extent that it works, is mirroring Bayesian decision theory. But that doesn't help the part of you can that talk, that can be reached by the part of me that can talk.

"Speak the truth, even if your voice trembles" isn't a literal executable decision procedure—if you programmed your AI that way, it might get stabbed. But a culture that has "Speak the truth, even if your voice trembles" as a slogan might—just might be able to do science or better—to get the goddamned right answer even when the local analogue of the Pope doesn't like it. I falsifiably predict that a culture that has "Use Bayesian decision theory to decide whether or not to speak the truth" as its slogan won't be able to do science—Platonically, the math has to exist, but letting humans appeal to Platonic math whenever they want is just too convenient of an excuse.

Would your views on speaking truth to power change if the truth were 2x less expensive as you currently think it is? 10x? 100x? I falsifiably predict that your answer is "Yes." Followup question: have you considered performing an experiment to test whether the consequences of speech are as dire as you currently think? I think I have more data than you! (We probably mostly read the same blogs, but I've done field work.)

(If so, are you sure that's not why you don't think the truth is more offensive than you currently think it is?)

Great question! No, I'm not sure. But if my current view is less wrong than the mainstream, I expect to do good by talking about it, even if there exists an even better theory that I wouldn't be brave enough to talk about.

Immaterial souls are stabbed all the time in the sense that their opinions are discredited.

Can you be a little more specific? "Discredited" is a two-place function (discredited to whom).