Sleeping Beauty Problem
Lubos wrote polemically about Sleeping Beauty Problem. The procedure is as follows.
Sleeping Beauty is put to sleep and coin is tossed. If the coin comes up heads, Beauty will be awakened and interviewed only on Monday. If the coin comes up tails, she will be awakened and interviewed on both Monday and Tuesday. On Monday she will be put into sleep by amnesia inducing drug. In either case, she will be awakened on Wednesday without interview and the experiment ends. Any time Sleeping Beauty is awakened and interviewed, she is asked, "What is your belief now for the proposition that the coin landed heads?" No other communications are allowed so that the Beauty does not know whether it is Monday or Tuesday.
The question is about the belief of the Sleeping Beauty on basis of the information she has, not about the actual probability that the coined landed heads. If one wants to debate one imagine oneself to the position of Sleeping Beauty. There are two basic debating camps, halfers and thirders.
- Halfers argue that the outcome of coin tossing cannot in any manner depend on future events and one has have P(Heads)= P(Tails)=1/2 just from the fact that that the coin is fair. To me this view is obvious. Lubos has also this view. I however vaguely remember that years ago, when first encountering this problem, I was ready to take the thirder view seriously.
- Thirders argue in the following manner using conditional probabilities. The conditional probability P(Tails|Monday) =P(Head|Monday) (P(X/Y) denotes probability for X assuming Y) and from the basic formula for the conditional probabilities stating P(X|Y)= P(X and Y)P(Y) and from P(Monday)= P(Tuesday)=1/2 (this actually follows from P(Heads)= P(Tail)=1/2 in the experiment considered!) , one obtains P(Tails and Tuesday)= P(Tails and Monday).
Furthermore, one also has P(Tails and Monday)= P(Heads and Monday) (again from P(Heads)= P(Tails)=1/2!) giving
P(Tails and Tuesday)= P(Tails and Monday)=P(Heads and Monday). Since these events are independent for one trial and one of them must occur, each probability must equal to 1/3. Since "Heads" implies that the day is Monday, one has P(Heads and Monday)= P(Heads)=1/3 in conflict with P(Heads)=1/2 used in the argument. To me this looks like a paradox telling that some implicit assumption about probabilities in relation to time is wrong.
When one speaks about independent events and their probabilities in physics they are must be causally independent and occur at the same moment of time. This is crucial in the application of probability theory in quantum theory and also classical theory. If time would not matter, one should be able to replace time-line with space-like line - say x-axis. The counterparts of Monday, Tuesday, and Wednesday can be located to x-axis with a mutual distance of say one meter. One cannot however realize the experimental situation since the notion of space-like amnesia does not make sense! Or crystallizing it: independent events must have space-like separation. The arrow of time is also essential. For the conditional probabilitys P(X|Y) used above X occurs before Y and this breaks the standard arrow of time.
This clearly demonstrates that philosophy and mathematics cannot be separated from physics and that the notion of time should be fundamental issued both in philosophy, mathematics and physics!