That metric is for measuring the overall utility of the system, not the individual utility to Dave, which would still be proportional to the number of speakers.
Well, actually, potential utility is wierd because it's not additive; it behaves a lot like probability in that I can derive the same utility from several potential sources. So, Dave might speak to Bar in Orc and learn something, or he might talk to Baz in Orc and learn the same thing, and the potential utility doesn't accumulate. Properly, when deciding whether to engage in some action, I have to derive the possible outcomes, evaluate the utility of each, calculate the probability distribution, and calculate the expected utility — hooray. And of course, human interaction is enormously unpredictable unless you intentionally limit it, which is why Dan's model is somewhat silly.
Hopefully, we can simplify that through assumptions, but that's difficult here: 1. Learning a language may provide intrinsic utility, independent of communication. Orc, for example, allows humans to digest cellulose. 2. Learning a language affects both the subject and the object of one's conversations in unpredictable ways (q.v. human interaction), which in turn affects primarily non-language utility. 3. Learning a language affects whether other people will learn that language.
All in all, modelling utility is hard, let's go shopping.
![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png)
timeasmymeasure
no subject
Date: 2006-07-27 04:06 am (UTC)Well, actually, potential utility is wierd because it's not additive; it behaves a lot like probability in that I can derive the same utility from several potential sources. So, Dave might speak to Bar in Orc and learn something, or he might talk to Baz in Orc and learn the same thing, and the potential utility doesn't accumulate. Properly, when deciding whether to engage in some action, I have to derive the possible outcomes, evaluate the utility of each, calculate the probability distribution, and calculate the expected utility — hooray. And of course, human interaction is enormously unpredictable unless you intentionally limit it, which is why Dan's model is somewhat silly.
Hopefully, we can simplify that through assumptions, but that's difficult here:
1. Learning a language may provide intrinsic utility, independent of communication. Orc, for example, allows humans to digest cellulose.
2. Learning a language affects both the subject and the object of one's conversations in unpredictable ways (q.v. human interaction), which in turn affects primarily non-language utility.
3. Learning a language affects whether other people will learn that language.
All in all, modelling utility is hard, let's go shopping.