Page 148 - MODES of EXPLANATION

Basic HTML Version

not go far enough. Ockham’s razor is usually regarded as a (or even as
the
) fundamental principle
of scientific
inference
(Harman and Kulkarni 2012). Inference implies belief in what is inferred
and to believe a theory is to believe that it is
true
. So, unless one can, in turn, connect the ersatz
virtues with truth,
presuming
that the ersatz virtues point to the truth is tantamount to
wishful
thinking
, the undeniable epistemic sin of presuming that reality is the way one wishes it were. To
methodologically justify Ockham’s razor as a principle of inference, one must methodologically
connect Ockham’s systematic bias toward simplicity with finding the true theory.
So let us turn, resolutely, to methodological justifications of Ockham’s razor that are
clearly truth directed. For example, Bayesians argue that if you start out with a simplicity bias over
alternative theories, and you continually update by Bayesian conditioning, then your degree of
belief in the true theory will converge to 1 in the limit. Here is what that amounts to. Transport
yourself in imagination to a more bucolic place and time, when a common entertainment was to
shoot rusty cans off fence railings. Suppose that the cans are arranged in ascending order of
complexity (e.g., the labels get fancier). Suppose that one of the cans is nailed to the fence rail.
That is the true theory. Now shoot down the cans, one after another, in order of increasing
complexity, until you shoot the nailed can and it does not fall down (Earman 1992). Ockham’s
razor finds the truth in
that
rote sort of way.
When
the nailed can is hit, it is, indeed, the simplest
can remaining on the fence. But that is as vacuous a truism as saying that the thing you seek is
always in the last place you look! It hardly follows that the things you look for are always
avoiding
you; neither does it follow that simplicity can
smell
the truth. An alternative way to converge to
the truth is to start out with a very complex theory and then to drop it when the anticipated
complexity does not emerge after some long period of time. Historically, that has happened many
times. For example, classical electrodynamics implies that the absolute velocity of the Earth
through space has a causal role in the generation of electrical currents (Staley 2009). Repeated
failure to detect any such effect contributed to the rejection of classical electrodynamics. The
problem is not to show that Ockham’s razor is sufficient for finding the truth. The difficulty is to
show that it is
necessary
for finding the truth, or that it is, at least, better than
alternative
methods
like the one just described. And that is very hard to do without begging the question with a material
assumption that the truth is simple or probably simple.
The Middle Way
Here is the crux of the puzzle surrounding Ockham’s razor. If you demand truth in the short run,
you need to posit a
correlation
between simplicity and truth, and then you can’t explain how it
works without begging the question with a metaphysical story that explains the short-run
correlation. At the other extreme, long-run, methodological convergence is too weak, because
lots of different biases would get you to the truth in the long run. One conception of success is
too strong to be feasible without magic or question-begging assumptions; the other conception of
success is so weak that it does not rule out any conclusions in the short run, however complex.
The obvious, Goldilocks moral is to seek an intermediate success concept. Here is an ancient
hint, from the
Katha Upanishad
, circa 600
BCE
:
2