On Wed, 26 Jun 2013, omd wrote:
> Although there are no appeals, I do strenuously object to this judgement.
> The default assumption is the default because it is
> usually accurate. Unless you believe that my brother and I are not in
> control of the theagoranundead Gmail account, something
> which I could verify if necessary but which I thought would be fairly
> obvious, you should agree that it is clear (>95% chance,
> since we're talking Bayesian) I made a random-ish mental determination of
> whether to send the message myself or ask my brother
> to do it. Although the chance of each option was probably not exactly 50%,
> as it depended on my mental state, the availability
> of my brother, etc., and I admit I did not actually flip a coin, still you
> should agree that there is a substantial chance of
> each option which, based on the information you have, is not more than 10 or
> 20 percent away from 50%. Therefore, from a
> Bayesian perspective, there is not enough information for you to establish a
> reasonably high likelihood of either option.
>
> From a legal perspective - well, that's different, but although it is within
> reason that a court could distort a determination
> of the truth (who sent the message) into a hard rule to be applied in lieu of
> absolute proof to the contrary, designed to
> prioritize having some answer over correctness, I question whether we should
> do so (or have done so) in Agora. I suppose we do
> make such assumptions about mail sent from a particular email address, but
> that's a much narrower case and involves the
> definition of the sender of a message, and hasn't been tested in any case.
Email identity is a special case in every email nomic. In order to play, even
start the game,
a player must assume the independence (as thinking entities) of different email
sources
(Well, I assume that everyone who registered since 2003 is an avatar of Steve,
but that's
jus me) . If a question on identify is raised, it necessarily involves
examining what is known
about the real facts and adjust those assumptions. Basically, since we start
out making a
strong assumption about each email address, and this assumption is necessary
just to
play the game, it takes genuine evidence (not just "maybe it was me, maybe
not") to
shift the assumption.
This is actually fully compatible with a Bayesian worldview, at least in the
sub discipline
of decision theory. At the end of the day it's the preponderance of the
evidence, which
Is strongly informed by the priors.
