I don’t really buy that argument. It would apply to any measurement scenario. You could say in the two-mirror experiment: “These dots on the screen don’t mean a thing, we just got extremely lucky.” Which is of course always a theoretical possibility.
Of course you can derive that you were extremely lucky, but also that “someone got extremely lucky” [SGEL]. If you start with some arbitrary estimates e.g. P(SWI)=0.5 and P(MWI)=0.5 and try to update P(MWI) by using Bayesian inference, you get:
I don’t really buy that argument. It would apply to any measurement scenario. You could say in the two-mirror experiment: “These dots on the screen don’t mean a thing, we just got extremely lucky.” Which is of course always a theoretical possibility.
Of course you can derive that you were extremely lucky, but also that “someone got extremely lucky” [SGEL]. If you start with some arbitrary estimates e.g. P(SWI)=0.5 and P(MWI)=0.5 and try to update P(MWI) by using Bayesian inference, you get:
By P(SGEL|SWI)=1/2^20
P(SGEL|MWI)=1
You get:
P(MWI|SGEL)=P(SGEL|MWI)P(MWI)/(P(SGEL|SWI)P(SWI)+P(SGEL|MWI)P(MWI))=
0.5/((1/2^20)0.5 + 0.5)=1/(1+1.2^20) ~ 1-1/2^20
Well, yes, but we can’t peek into other Everett branches to check them for lucky people.
I don’t see why you wanted to. You could only increase P(MWI) by finding there any.