I believe that I have demonstrated conclusively that the version of the AP described by A and B is never going to give falsifiable predictions for the parameters of physics and cosmology. For the few times an argument called 'anthropic' has led to a successful prediction, as in the case of Hoyle's argument and Weinberg's first argument, examination shows that it rests entirely on a straightforward deduction from an observed fact about our universe.

Thus, if we are to understand the choices of parameters in the context of a falsifiable theory, we need an alternative approach. One alternative to deriving predictions from a multiverse theory is patterned on the successful model of natural selection in biology. This was also originally motivated by asking the question of how science can explain improbable complexity. To my knowledge, only in biology do we successfully explain why some parameters - in this case the genes of all the species in the biosphere - come to be set to very improbable values, with the consequence that the system is vastly more complex and stable than it would be for random values. The intention is then not to indulge in some mysticism about 'living universes', but merely to borrow a successful methodology from the only area of science that has successfully solved a problem similar to the one we face.15

The methodology of natural selection, applied to multiverse theories, is described by three hypotheses.

(i) A physical process produces a multiverse with long chains of descendants.

(ii) For the space P of dimensionless parameters of the Standard Models of physics, there is a fitness function F(p) on P which is equal to the average number of descendants of a universe with parameters p.

(iii) The dimensionless parameters pnew of each new universe differ, on average, by small random amounts from those of its immediate ancestor

15 Other approaches to cosmology which employ phenomena analogous to biological evolution have been proposed, for example, by Davies [79], Gribbin [80], Kauffman [81] and Nambu [82]. We note that Linde sometimes employs the term 'Darwinian' to describe eternal inflation [83—89]. However, because each universe in eternal inflation has the same ancestor, there is no inheritance and no modification of parameters analogous to the case of biology.

(i.e. small compared with the change that would be required to change F(p) significantly).

Their conjunction leads to a predictive theory, because - using standard arguments from population biology - after many iterations from a large set of random starts, the population of universes, given by a distribution p(p), is peaked around local extrema of F(p). With more detailed assumptions, more can be deduced, but this is sufficient to lead to observational tests of these hypotheses. This implies the following prediction.

(5) If p is changed from its present value in any direction in P, the first significant changes in F(p) encountered must be to decrease F (p).

The point is that the process defined by the three hypotheses drives the probability distribution p(p) to the local maxima of the fitness function and keeps it there. This is much more predictive than the AP, because the resulting probability distribution would then be much more structured and very far from random. If, in addition, the physics that determines the fitness function is well understood, detailed tests of the general prediction S become possible, as we will now see.

Was this article helpful?

## Post a comment