Mechanisms and Laws: Clarifying the Debate - Mechanism and Causality in Biology and Economics - page 131

generalizations are extrapolated in scientific practice. It is unclear why Leuridan
refuses the mechanist's help in addressing the extrapolation problem.

3.2 Do cs-Mechanisms Require Micro-regularities?

Let us move on, then, to the second route (b) by which Leuridan argues that cs-
mechanisms are ontologically dependent on stable regularities (L1). Leuridan
claims: “There can be no cs-mechanism without some lower-level (c)P-regularities
(i.e., the regular behaviors, operations, or activities displayed or engaged in by the
mechanism's parts)” ( 2010 , p. 331). A (c)P-regularity is a causal p-law, a p-law that
is “invariant under some range of interventions” ( 2010 , p. 328). Leuridan argues for
this thesis using a thought experiment. If the behaviors of all of the parts of the
mechanism were to behave completely randomly, by which he means that they do
what they do as the result of a “completely random internal process,” “this would
make it very unlikely to produce a macro-p-regularity, let alone a (c)P regularity”
( 2010 , p. 331). 6 What shall we make of this argument?
Clearly, Leuridan's thought experiment does not support the ontological conclu-
sion that there can be no cs-mechanisms without some p-regularities among the
parts. At most, it supports a probabilistic conclusion that cs-mechanisms are
unlikely without p-regularities, and such an argument cannot support the negated
existential quantifier in Leuridan's second ontological claim (b). The thesis that x is
unlikely to have property F is consistent with the claim that x is F and, for nonzero
probabilities, entails that x is possibly F (directly contradicting Leuridan's stated
thesis). Although randomly behaving components such as those in Leuridan's
example would not form a mechanism (given that the behavior of each is causally
independent of the behaviors of the others), it is still possible that together they
would produce a regularity, even a (c)P-regularity, of some stability and strength.
Just how improbable this would be depends upon the number of variables and the
number of values they might take. In order to make experimental progress in the
discovery of causes and mechanisms, we regularly presume that regularities do not
arise merely from chance. However, as the statistics attached to any causal experi-
ment acknowledge, there is always some nonzero probability that the results of the
experiment did arise strictly from chance. Now if macro-regularities can obtain
even among causally unconnected random events (as in Leuridan's example), then

6

It should be noted that Leuridan defines “irregularity” in such a way as to effectively exclude
discussion of stochastic mechanisms, mechanisms that work only infrequently or whose frequency
of operation and stability in space vary over time. A mechanism that works with probability
0.000001 will count as regular on Leuridan's account because one can write a generalization of the
form P ( X ) 0.000001. This is unfortunate as there are a number of interesting questions that one

might ask about probabilistic mechanisms and mechanisms whose probability of working varies
over time (as one might expect in systems that are regulated). Thanks to Jim Bogen for calling this
to our attention.

¼