At 18:59 16/06/03 +0100, Ted Harding wrote:
>Incorporating circumstantial prior probabilities into probability of
>guilt is a can of worms with all sorts of potential scope for injustice.
Whilst that is obviously the case (as all your examples illustrate), as
I've just written to Allan, I think that, whether we like it or not, human
nature is such that juries and the judiciary undoubtedly do bring a concept
of prior probabilities to any case they consider. That the prior
probabilities are (more-or-less necessarily) 'circumstantial' is not, in
itself, necessarily an issue - after all, people can be (and are) found
guilty 'beyond a reasonable doubt' on the basis of circumstantial evidence
alone.
Many of the 'preconvictions' (which we call prior probabilities) are, of
course, seemingly pretty 'unfair' to individuals, even if statistically
reasonable in 'population' terms. However if we are (in the way I
suggested) accepting that convictions generally have to be probabalistic,
that in itself is implicitly extrapolating (quite possibly 'unfairly') from
a 'population' to an individual (e.g. "95 times out of 100 this court was
told such a story, the defendent would be guilty").
In practical and realistic terms, I don't think one can consider a trial in
court in isolation. It is part of a continuous process, and I think that
most 'reasonable (wo)men' will assume that the fact that the police and/or
CPS have decided to charge a particular individual with a particular
offence greatly increases the 'prior probability' of that person being
guilty of that offence. Again, in probabilistic terms, this surely is the
case - an individual brought to court by police/CPS for a particular
offence being far more likely to be guilty of it than would be a random
member of the public who had been picked up and 'dragged into
court'. However, again, this probabilistic fact is no real consolation to
the innocent person who falls foul of it, and certainly no protection
against misjudgement, or even corruption/ malpractice, on the part of the
police/CPS.
>You can give the non-drunk a chance, if under pressure to do so, by
>being just a bit more discriminating about whom you arrest, allowing
>some of the drunk to escape; and even more of a chance by giving the
>drunk an even bigger chance to escape. If you want complete protection
>of the innocent, then don't arrest anyone and let all the guilty ones
>get away with it. .... Quite where the line is drawn is, in fact, a purely
>political decision. Different regimes will draw their lines in different
>places.
Exactly. The whole thing really is statistical, about the balance between
Type I and Type II errors - and 'society' (in whatever guise) has to define
the desired levels of 'alpha' and 'beta'. What I think is wrong is the
fact that we don't hear of our judicial system being talked about in such
terms, instead taking the 'easy option' of expecting juries to make their
own decisions as to the meaning of 'a reasonable doubt'.
Test cases would be easy enough to engineer. A jury could go back to a
judge and said "We believe that there is just a 1 in 20 (or 1 in 100, or 1
in 1000, or whatever) probability that this person is innocent; is it
acceptable for us to return a guilty verdict?".
>However, I'm very much in favour of using prior probability for
>the purpose of investigation, to identify likely suspects (provided of
>course it is done fairly and without bias). But once suspects have
>been identified, the prior probability which led to them being
>investigated should not contribute to evidence of their guilt;
>otherwise it is being used twice, and comes in as the square in the
>probability of their final conviction.
Statistically fairly sound, but surely often not practical or
realistic? Factors 'leading to people being investigated' may well also be
important circumstantial evidence of their guilt - not enough, on its own,
to secure a 'safe' conviction, but a crucial part of an overall package of
evidence which IS regarded as adequate to convict. To take simple
examples, being found in the vicinity of, or moving away from a crime scene
will both be a 'reason to investigate' and also an important part of the
package of evidence against the person, should that person be subsequently
charged.
Kind Regards
John
----------------------------------------------------------------
Dr John Whittington, Voice: +44 (0) 1296 730225
Mediscience Services Fax: +44 (0) 1296 738893
Twyford Manor, Twyford, E-mail: [log in to unmask]
Buckingham MK18 4EL, UK [log in to unmask]
----------------------------------------------------------------
******************************************************
Please note that if you press the 'Reply' button your
message will go only to the sender of this message.
If you want to reply to the whole list, use your mailer's
'Reply-to-All' button to send your message automatically
to [log in to unmask]
*******************************************************
|