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

Eliezer_Yudkowsky comments on Group rationality diary, 7/23/12 - Less Wrong

4 Post author: cata 24 July 2012 08:49AM

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

Comments (48)

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

Comment author: wsean 24 July 2012 04:48:00PM 22 points [-]

Used Bayes in the wild.

It was really a textbook case. I had a short story under review at Asimov's SF magazine, and they'd held onto it for over two months. Per Duotrope (a writer's market site), Asimov's takes twice as long on average reviewing stories that are ultimately accepted as stories that are ultimately rejected (the likely explanation is that obviously bad stories can be rejected right away, while stories that eventually get bought are handed around to multiple readers). So I sort of casually assumed without really thinking about it that this meant I was much more likely to get my story accepted.

But wait! Duotrope has various response statistics based on reports from users. With a few assumptions, I could use their numbers for a simple Bayes calculation, and figure out the real chances for an acceptance given a two month wait.

I used as my priors the numbers on Duotrope for acceptance rate (P(Acceptance)), mean reply time and std dev of reply time (used for P(Two Month Wait)), and took a guess at P(TMW|A) based on the mean reply time for acceptances. The result: yes, it was more likely that my piece was accepted, but because P(A) was so low to begin with (less than .5%), P(A|TMW) was still really low (less than 1%).

I calibrated my expectations accordingly. Which was just as well, since I got their rejection the next day. Rejections are always disappointing, but I'd have been far more disappointed if my expectations had still been out of joint.

Comment author: Eliezer_Yudkowsky 25 July 2012 05:11:13AM 10 points [-]

Lovely! I intend to add this as a Bayesian sample problem - enough rationality diaries, and we'll be able to make Bayes booklets exclusively out of real-life cases encountered by LessWrongers!