Re: Single model or framework for reputation/trust

Ed Gerck <[email protected]> wrote:
> >   Reputation (or past behavior) is a strong indicator of future
> > behavior.  It shouldn't be ignored.
> 
> Of course not. But take a looks at banks: they never lend money
> based on reputation as a strong indicator.

  "never" is perhaps a little strong.

  Look at Venture Capitalists in the computer world.  They loan
millions based on promises & reputation.

> No. What the example says is that reputation is NOT a strong
> indicator of future behavior.

  I'm sorry to disagree with you and Bertrand Russell, but the story
is based on misconceptions, as I pointed out previously.  Was the
chicken wrong for 6 months, when it predicted that "today, the farmer
will feed me?"  No.

  So in your example, for the (say) 6 months that the chicken was
alive, reputatation was correct 99.5% of the time.  The fact that
reputation isn't perfect is something that both you, and Bertrand
Russell, should know.

>  In terms of trust, however, the chicken would know that the farmer
> could not be trusted because self-affirmations, even repeated many
> times, cannot induce trust.  All that the chicken could see were the
> self-affirmations of the farmer.

  Nonsense.  The chicken sees the farmers actions every day, and makes
a conclusion for itself.  The farmer doesn't even have to explain
anything to the chicken.

  Similarly, I trust that the sun will rise tomorrow morning.  Aliens
may take it away before then, but it's a reasonable (albeit imperfect)
trust, based on conclusions I draw for myself.

> Reputation is a component of trust (identified with the 
> backward-looking chain of trust I mentioned here before)
> but there are other mechanisms required to be able to induce
> (and determine) trust.

  That's nice.  But I want to be able to exchange email with people I
don't trust.  So trust appears to be less relevant than reputation.

  Alan DeKok.