Scalable ways to associate evidence (pro or con) with claims will be more valuable in elevating accuracy than complex voting and reputation systems

by Andrea Gallagher Jan 8 2017

Discussions on Less Wrong have delved into complex systems of voting and moderation to improve the quality of discourse in a community. I think creating structures to a discussion beyond threaded comments will be more productive and have a higher ROI. For claims, I care less about comments than about whether someone can provide evidence in support or against it, and evidence can be strong or weak and have it's own evidence-children. Summing up across all of that evidence would be interesting.


Andrea Gallagher

So, how do we measure value in elevating accuracy or truth?

Ben Pace

My main intuition against: I think that the value that an arbitrary researcher gives to a field is highly heavy tailed, whereby a few key people move the whole field forward, and so we want clear indicators of who those people are, and to listen to them.

My main intuition for, however, is that the whole premise of arbital is that you can get value from increasing the communication quality in the whole network, which is what increasing the quality of evidence available is (this isn't a product aimed at the few, it's like wikipedia, which gets its value from raising the whole level of discourse).

At a first pass I'm weakly in agreement, but not sure that one is significantly more valuable than the other (or that they necessarily require trading-off).

Eric Bruylant

I foresee good reputation systems being extremely valuable (essentially necessary to scale while maintaining quality), with high credence on that being more important than argument structuring features.

Alexei Andreev

My main point of disagreement is that it's sometimes hard to say if some evidence is "for" or "against." It might also be subjective. Or it might be ___, for which I don't have a word, but an example would be General Relativity as it relates to Newtonian Physics.

Alexei Andreev

Even if the platform had pros and cons, you'd still need to decide for any given claim whether some piece of evidence is a pro or a con and by how much. I'm not even sure what a good solution to that might look like that isn't basically a reputation system.

Paul Crowley

I think that Scott's essays The Control Group Is Out Of Control (particularly the idea of the Ouroboros of Scientific Evidence) and Why I Am Not Rene Descartes are relevant here. In difficult cases, assessing evidence properly is a matter for intricate discussion.

Andrea Gallagher

Thanks for picking apart my claim, folks! Rather than modify this claim, I think I'll work on a Post approach, probably with a few different linked claims in it.

In retrospect, I don't like the value comparison structure of the claim.

I'm used to formulating pseudo-hypotheses in a way that feel testable to me, and relative comparison can be easier than picking and measuring some absolute value. And I do think that in any project each effort is traded off against other efforts. But the claim focused as much on the value of reputation systems, rather than the value of structured discussions, and that muddied the discussion.

Since I'm using this to help get a feel for how good claims, posts, etc are crafted, this was very helpful.