I try in the following to express Laskey's wager for believing in god in a
decision theoretic framework, as well as Pascal's wager. I find this helpful,
as it shows that there are an awful lot of modeling choices you take to come to
an decision.
Laskey's Wager:
States of the world S1
s1: A god of _my own definition_ exists
not-s1: This god does not exist (one can specify this more)
Decisions D1:
d1: I believe in the existence of this god
not-d1: I do not believe in this god
Consequence C1:
Regard of my own life
Utility U1:
(Uschi's one) very high for high regard and minus infinity for no regard
Note: The "believe in god" is a decision and different from a prior probability
of pr(s1).
In this framework, C1 is unresponsive to the states of the world, that is
C1(d1, s1)=C1(d1, not s1) and C1(not-d1, s1)=C1(not-d1, not-s1). So I do not
need pr(s1) to decide. A critical point of this model is that the decision
itself might be actually a chance node responding to S1. That is, may be my
definition of god is such that (s)he is able to influence my decision to
believe in hir, for example by helping in a critical life situation.
Pascal's wager:
S2:
s2: An anthropomorphic, omnipotent parent-surrogate god in the sense of David
Wolpert exists that is so jealous that he (no- not he or she here I'd say)
sends you to hell if you do not believe in him
not-s2: this god does not exist
D2=D1: as before
C2: Amount of suffering after live
U2: High utility for no suffering
Here S2 needs urgently some better definition, as it is totally unclear what
C2(not-s2, d2) is (for example may be some heavy feminist goddess exists that
sends you to hell for believing in such an patriarchic god) as well as
C2(not-s2, not-d2) is unclear. Unfortunately, this patriarchic god might send
you to hell though you believe in him, thus C2(s2, d2) is not defined as well.
We only have for sure that C2(s2, not-d) is not desirable. I find a (slight?)
change of C2 interesting here: C2'=Fear from suffering after live in this live,
because C2' is unresponsive to S2, and it is easy to come to a decision then.
A combination of both wagers leads to
S:
s1: a god of my definition exists
s2: a patriarchic and punishing god exists
s3: not-(s1-or-s2 )
D:
d1: believe in s1
d2: believe in s2
d3: believe in s3
d4: indifference (interestingly I feel some need to add this here but not
above)
C1: Regard of one's own life
C2: Suffering after life
U1 and U2 as above.
U: Some combination of U1 and U2
And here of course my decision depends heavily on the way I combine U1 and U2
to U.
We would have to build some hierarchical framework to combine all ways people
come to a decision of believing in any god of their definition - as there is
some meta decision on what consequences and what utilities you take into
account. And all the modeling process might be influenced by any of these gods.
I prefer to model consequences for this life - as in Laskey's wager.
Uschi Sondhauss
