Mechanisms and Laws: Clarifying the Debate - Mechanism and Causality in Biology and Economics - page 130

the extrapolation. Crucially, one need not compare all of the entities, activities, and
organizational features of a mechanism to those in the target organism in order to
assess the likelihood that one's extrapolation will work: one might, for example,
compare downstream (rather than upstream) portions of a mechanism, given that
crucial differences downstream will indicate crucial differences earlier. Con-
versely, similarity at a key bottleneck point in the mechanism might allow one to
neglect any differences upstream in the mechanism to focus on what comes later
(see Steel 2008 , p. 90). Furthermore, if one is interested simply in gross qualitative
differences, such as whether a given drug is positively relevant for a side effect or
negatively relevant for a side effect, certain minute and highly specific differences
in the mechanisms might be less germane than the simple matter of whether there is
a positive (excitatory) or negative (inhibitory) causal or correlational relationship in
the model organism. Thinking about underlying mechanisms, in short, provides
new tools for assessing when our knowledge is likely to extrapolate and when
extrapolation is more precarious.
Steel's strategies rely primarily on considering the mechanisms that underlie a
regularity, but one might also justify extrapolation on the basis of antecedent
mechanisms, such as the mechanism of natural selection. That is, one might
claim that the hereditary mechanisms in Drosophila can be expected to apply
outside of the laboratory and in other species because hereditary mechanisms are
evolutionarily ancient and therefore widely conserved across the tree of life. As
Bechtel ( 2009 ) argues, this mechanistic fact about the history of life warrants
tentative (heuristic) extrapolation about closely related species: they might use
the same mechanism, or a mechanism composed of similar entities and activities,
or mechanisms with similar organizational structures. And one might expect evo-
lutionarily ancient mechanisms to be more widely conserved, and so more fitting
for extrapolation, than are relatively recent adaptations. This kind of heuristic is
especially interesting in the present context given that, according to this heuristic, a
singular mechanism (the one-off mechanism that produced the tree of life as we
now know it) warrants extrapolation of p-laws in extant species.
While we admit that these mechanistic contributions to our understanding of
extrapolation solve neither the problem of induction nor Goodman's new riddle of
induction ( 1955 ), we insist they nonetheless have considerably more content than
the bare tautology that p-laws warrant extrapolation because they are stable and
strong. Indeed, extrapolation is at least often justified by appeal to knowledge of
mechanisms. In sum, it appears that our epistemic reformulation of Leuridan's
argument runs into a dilemma. Either his claim is a tautology to the effect that
mechanisms must be general if one is to form true generalizations about them, or it
is a substantive epistemological thesis that extrapolation is possible only if there are
p-laws. If the latter, then we have shown how Leuridan begs the question by
presuming, rather than showing, that p-laws solve the extrapolation problem and
by asserting, rather than defending, the disputed thesis that p-laws are required for
extrapolation. Most importantly, however, we have reviewed some of the progress
mechanists have made in thinking about the problem of extrapolation. Focusing on
mechanisms provides fruitful and substantive ways of

thinking about how