pharos at gmail.com
Sun May 5 09:46:51 UTC 2013
On Sun, May 5, 2013 at 10:29 AM, Anders Sandberg wrote:
> So my first order guess is that if you want to figure out what group X
> likes, you should try asking group X (or watch revealed preferences in their
> behaviour). If you cannot or dare not, asking group Y that has somewhat
> similar preferences might work, but introduces noise. If you go to group Z
> that is also trying to impress group X, they might have figured out tricks
> you do not know. Of course, it might be that they are actually trying to
> impress an actually different group (like gay men being mostly concerned
> with gay men), so there is some reason to be sceptical there. So my second
> order approach would be to listen to advice or information from all these
> groups and combine it: there are probably robust patterns that would be true
> despite the filtering effects, and these are strong enough to care about.
> Combined expert sets typically outperform individuals on messy pattern
> matching. Inter-individual differences in what people like are anyway pretty
> big, so trying to exactly match advice or what some example person likes is
> bound to be biasing (except if you have a particular person in mind you want
> to impress, of course).
This sounds too academic to me. Rather like the neat efficient market theory.
It assumes that people know what they are doing! :) Immediate fail!
The efficient market doesn't work because of all the human illogical
emotional reactions, herd instinct, etc.
Similarly, fashion victims are trying to dress in the latest fashions,
not particularly to impress any group.
As you say, it also depends on fitting in to the environment you are
going to. People like to 'join in' and be one of the crowd.
(herd instinct, again).
More information about the extropy-chat