is better to create than not to. As long as our universe is above that cut-off line, God was
justified in creating it, even though superior universes abound, for to create our universe
and the superior ones is better than just creating the superior ones. Thus, multiple
universes can just as much be used in defense of theism as in defense of atheism.
Another kind of teleological argument, which has been promoted by Richard Swinburne
(1968), is based on the fact that the universe displays widespread law
end p.132
like regularity and simplicity. It is argued that there are only two possible explanations
for a fact: either a scientific explanation in terms of boundary conditions and laws of
nature, or a personalistic explanation in terms of the intentional activity of an agent. Now,
because a scientific explanation explains facts by invoking laws of nature, it cannot
explain why there are laws of nature on the pain of circularity. Thus, if there is an
explanation, it must be one that is given in terms of the intentional activity of a designer.
Several replies are available. The first is simply to deny the call for explanation here. The
basic laws of nature are rock-bottom, and they have no explanation. This approach is
particularly attractive if one is willing to bite the bullet and accept the implausible claim
that the laws of nature that in fact actually hold are logically necessary. Once one admits,
however, that the laws are contingent, one faces the following difficulty, at least if one
has the Humean intuition that all possible states of the universe are prima facie equally
likely to happen temporally after any one given state^4 : prima facie, it is vastly improbable
that things should behave in a regular way. Unfortunately, it is very difficult, if not
impossible, to assign precise probabilities to such things as universes. Hence, this
argument may necessarily have to be run on an intuitive level, though aided by simpler
cases. As a toy model, imagine a discrete Humean universe containing only one particle
of a fixed type and whose only degrees of freedom are in the spatial position, and whose
space-time has a temporal series consisting of a hundred instants of time, and the spatial
structure of a 10-by-10 grid. There are 10^200 such universes. A minimal constraint on
regularity is that the particle doesn't fly around to noncontiguous grid locations, but in
each time step is either where it was previously or at one of the up to eight neighboring
grid locations. There are fewer than 100×9^99 universes satisfying this constraint. Thus,
the probability that a randomly chosen toy model universe will satisfy the minimal
regularity constraint is less than the astronomically small value of 10−^100. Moreover, as
the grid becomes finer and finer and the time-series becomes closer and closer to being
continuous, this probability decreases exponentially.
Thus, a fortiori, the initial probability of a regular universe with continuous space and
time is exceedingly small on Humean assumptions, and indeed probably zero. On the
design hypothesis, on the other hand, a regular universe has a probability that is not
astronomically low. For an intelligent agent has good reason to produce order, order
being objectively valuable and necessary for the existence of forms of life capable of
intentional action, and the probability of an agent doing what she has good reason to do is
not astronomically low. After all, prima facie, an agent is not any less likely to be good
than she is to be evil or to be neutral, and so one might assign a probability of 1 3 that the
agent will be good, and then at the least some probability like 0.000001 (which, though
small, is not astronomically so) that if she is good, she will produce a universe exhibiting