GuySrinivasan comments on Help Fund Lukeprog at SIAI - Less Wrong

40 Post author: Eliezer_Yudkowsky 24 August 2011 07:16AM

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

Comments (276)

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

Comment author: Thrasymachus 26 August 2011 09:14:49PM 2 points [-]

Saying "the rationality minicamp was highly successful" before you have analyzed the data you have gathered to assess the success of the rationality minicamp is irrational.

If success at the minicamp is important - suggested by it listed first on Eliezer's recommendation - why not wait until you CAN analyze the data, to see whether it really was successful, before you recommend hiring Luke? Doing so means a) you can make a more persuasive case to donors, and b) if the minicamp WASN'T successful, then one can reconsider the hire.

The fact this plug happened before the analysis signals Eliezer is committed to recommending Luke's hire regardless of whether analysis shows the minicamp as successful or not. And if HE doesn't think the minicamp success is relevant to the merits of hiring Luke, why is he using it to persuade us?

Disclaimer: I think Luke has added lots of value to this site, and I would be unsurprised if later transparent analysis showed the minicamp to be highly successful. But the OP (as well comments switching various reasons/excuses for failing to present data, etc. etc.) is suggestive of irrational salesmanship. Perhaps a salutatory lesson that rationality experts still succumb to bias?

Comment author: GuySrinivasan 26 August 2011 09:56:03PM 3 points [-]

I would be unsurprised if later transparent analysis showed the minicamp to be highly successful

What do you mean by unsurprised? Some words don't communicate as well as we feel they do.

Comment author: Thrasymachus 27 August 2011 04:41:17PM 2 points [-]

My sentence was trash, sorry!

What I should have said:

"I have a 0.5ish estimate of the minicamps success, with big error bars. The suspicious behaviour is no more than minute confirmation for it being a failure."

Or something like that.