The Many-Worlds interpretation of quantum mechanics can be stated simply thus:
Wavefunctions do not collapse; rather, the observer becomes entangled with the observed.
This clears away the need for an ad-hoc rule about the level at which observation takes place, letting existing rules of quantum mechanics fulfil that role.
One of the principal misconceptions about the many-worlds theory is that it involves the creation of new worlds. It does not. What it does involve is the partition of the world's wavefunction into distinct cases that might as well be different worlds. This partitioning is not a new mechanism -- it is needed to explain many testable quantum phenomena, such as the two-slit experiment. What is different is the recognition that this mechanism is adequate to explain our observations on its own, without the need for any additional rules to be added.
Let's look at a system: a two-slit diffraction experiment. We have an electron shot at a barrier with two slits in it. We can choose to measure at the barrier whether the electron passes through each slit, or we can choose not to measure this. We can also measure the position at which the electron hits a target on the far side of the barrier. The system is supposed to be very 'clean' with the particles not losing coherence quickly, even when passing through the detectors (this is hard to do, but it can be done).
As we start out, there are three basic divisions in the components of the electron's wavefunction:
- The electron passes through the left slit
- The electron passes through the right slit
- The electron hits the barrier
Each of these components proceeds independently
of the others.
If we choose to measure only at the target, then the three components add together, interfering. Options A and B add up to a two-slit pattern on the target; option C does not contribute to wavefunction at the target because it was blocked at the barrier (but it does add some intensity on the other side of the barrier, where the electron would go undetected).
What happens now?
- According to the Copenhagen interpretation, the detector forces the universe to pick from among the possible options. This is a rather mystical effect for a detector to have.
- According to Many Worlds, the detector's wavefunction splits up into independent components just like the electron's did just a moment ago. This time there are many more components, and the detector is bigger than an electron, but that is irrelevant -- it is still the same mechanism.
What happens if we choose to measure at the barrier and
- According to the Copenhagen interpretation, our detector at the barrier forces the universe to pick option A, B, or C, destroying the other components, then our second detector forces the universe to pick again, for the final position of the electron -- since it has already forced it out of all but one component, it will use the one that survived the last round of measurement.
- According to Many Worlds, our detectors all become quantum-entangled with the electron, which proceeds just as before. With the detectors entangled, the various outcomes proceed independently of each other. However, by having our detection in two tiers, we pre-partition ourselves into the A component and B component (and C component), which prevents interference at the later stage.
So, let's compare the two pairs of results. In particular, the fate of the electron.
- In the Copenhagen interpretation, the electron's fate is intimately tied to the on/off switch on a detector which exerts the same force on the electron whether it is on or off (or, if there is a difference, the change in outcome is completely incommensurate to the change in strength of that force).
- In the Many-Worlds theory, the electron is not affected by the on/off setting of the detector. It is the detector which is affected by its own on/off setting.
You decide which makes more sense.
Many Worlds also has the virtue of maintaining locality in the face of the Einstein-Podolsky-Rosen paradox. The instantaneous information transfer is only required when the second wavefunction collapses. However, in many-worlds, wavefunctions do not collapse. Thus, there is no need for instantaneous information transfer at all.
But how, then, do we not observe kooky quantum effects in the macro world, if there's no special rule to keep them out? Well, what kooky quantum effects are we missing?
- Heisenberg Uncertainty: have YOU ever simultaneously measured your position and momentum (or any other appropriate pair of properties) to precision better than Planck's constant? I didn't think so. The restriction here is so remote that we don't notice it.
- Tunnelling across classically prohibited barriers: probability of tunnelling decays exponentially with width and strength of barrier. Any barrier big enough for us to notice is far far too big to tunnel across.
- Wavefunction interference: this may happen all the time everywhere, but it's hard to notice except with multiple measurements on systems which are identical to the point of coherence. The complexity of macroscopic objects makes it impossible to achieve even approximate identicality, and even if we could get the atoms into the right places, they wouldn't be coherent (the difficulties in achieving this are known, unsurprisingly, as decoherence).
- Objects being spread out over distances: It's only before you look at it that its location is undetermined to you. Did your wallet spread out over time while you weren't looking? Well, assuming there was a possibility that some force would come along and displace it, yes, it is spread out. However, the moment you find it, you find it all in one place (unless the force was one that would destroy the wallet). Why? The moment you have enough information to confirm one outcome of the wallet for yourself, you are entangled with that one and you cannot observe the effects of the others. If you do try to perform interference experiments to tease this diffusion out of the system, you will run into the decoherence problems noted above, for Wavefunction Interference.
One critique of Many-Worlds is that one loses justification for the Born Probabilities. This is a bit rich since the Copenhagen Interpretation includes them as a separate axiom. But more to the point, all you need to get out the Born Probabilities is the notion that the wavefunction can be interpreted such that some set of mutually orthogonal subspaces represent probabilities. The Born Probabilities immediately result from the Pythagorean Theorem. All the difficulty is apparently in connecting 'this is real' to 'you can treat it as if it was what reality is made of, in the way that it looks like it actually does', which I find somewhat baffling.
It has been pointed out that a detector does not need to be on in order to 'measure' the system as far as the Copenhagen Interpretation is concerned. This is only going to be true if the detector is disordered enough to cause decoherence. If the detectors are adequately quiet, inactive detectors do not need to cause decoherence (and even if they are on, if they are designed properly). I have only considered this quiet case above.
Concerning the name:
This idea would more properly be called the many worlds interpretation of quantum mechanics, not a hypothesis. If many-worlds is indeed a hypothesis and not an interpretation, then from the point of view of assuming quantum mechanics, it is the null hypothesis (though from the point of view that only observations exist, i.e. the Copenhagen Interpretation, it's not).
Secondly, since there are many partitions, but only one greater universe, I prefer a different name-root than 'Many Worlds', perhaps 'Universal Entanglement' or 'Non-Collapsing Wavefunction'. However, we seem to be pretty solidly stuck with 'Many Worlds' as the recognized name for this idea.