Conditional meta-preferences

I’d just want to make the brief point that many hu­man meta-prefer­ences are con­di­tional.

Sure, we have “I’d want to be more gen­er­ous”, or “I’d want my prefer­ences to be more con­sis­tent”. But there are many vari­a­tions of “I’d want to be­lieve in a philo­soph­i­cal po­si­tion if some­one brings me a very con­vinc­ing ar­gu­ment for it” and, to var­i­ous de­grees of im­plic­it­ness or ex­plic­it­ness, “I’d want to stop be­liev­ing in cause X if im­ple­ment­ing it leads to dis­asters”.

Some are a mix of con­di­tional and anti-con­di­tional: “I’d want to be­lieve in X even if there was strong ev­i­dence against it, but if most of my so­cial group turns against X, then I would want to too”.

The rea­son for this stub of a post is that when I think of meta-prefer­ences, I gen­er­ally think of them as con­di­tional; yet I’ve read some com­ments by peo­ple that im­ply that they think that I think of meta-prefer­ences in an un-con­di­tional way[1]. So I made this post to have a brief refer­ence point.

In­deed, in a sense, ev­ery at­tempt to come up with nor­ma­tive as­sump­tions to bridge the is-ought gap in value learn­ing, is an at­tempt to ex­plic­itly define the con­di­tional de­pen­dence of prefer­ences upon the facts of the phys­i­cal world.

Defin­ing meta-prefer­ences that way is not a prob­lem, and bring­ing the defi­ni­tion into the state­ment of the meta-prefer­ence is not a prob­lem ei­ther. In many cases, whether we la­bel some­thing con­di­tional or non-con­di­tional is a mat­ter of taste, or whether we’d done the up­dat­ing ahead of time or not. Con­trast “I love choco­late”, with “I love deli­cious things” with the ob­ser­va­tion “I find choco­late deli­cious”, with “con­di­tional on it be­ing deli­cious, I would love choco­late” (and “I find choco­late deli­cious”).

  1. This sen­tence does ac­tu­ally make sense. ↩︎

No comments.