Less Wrong is a community blog devoted to refining the art of human rationality. Please visit our About page for more information.

SaidAchmiz comments on LW 2.0 Strategic Overview - Less Wrong

47 Post author: Habryka 15 September 2017 03:00AM

You are viewing a comment permalink. View the original post to see all comments and the full post content.

Comments (294)

You are viewing a single comment's thread. Show more comments above.

Comment author: SaidAchmiz 16 September 2017 04:01:04AM 3 points [-]

If you have a more-legible quality signal (in the James C. Scott sense of "legibility"), and a less-legible quality signal, you will inevitably end up using the more-legible quality signal more, and the less-legible one will be ignored—even if the less-legible one is tremendously more accurate and valuable.

Your suggestion is not implausible on its face, but the devil is in the details. No doubt you know this, as you say "this sketch has many problems of its own". But these details and problems conspire to make such a formalized version of the "expert's vote" either substantially decoupled from what it's supposed to represent, or not nearly as legible as the simple "people's vote". In the former case, what's the point? In the latter case, the result is that the "people's vote" will remain much more influential on visibility, ranking, inclusion in canon, contribution to a member's influence in various ways, and everything else you might care to use such formalized rating numbers for.

The question of reputation, and of whose opinion to trust and value, is a deep and fundamental one. I don't say it's impossible to algorithmize, but if possible, it is surely quite difficult. And simple karma (based on unweighted votes) is, I think, a step in the wrong direction.

Comment author: ingres 16 September 2017 04:17:28AM 1 point [-]

As far as an algorithm for reputation goes, academia seems to have something that sort of scales in the form of citations and co-authors:


It's certainly a difficult problem however.