Patterns of Defeasible
Reasoning
Various warrants (general defaults) for defeasible reasoning can be
identified. The following ones are discussed by Pollock (1998, 2008):
-
Perceptual inference . If I have a percept with content
P , then I can presumably conclude that P is true. For
instance, if I have an image of a red book at the centre of my field
of vision, I can conclude that there is a red book in front of me.
This conclusion is defeated if I become aware of circumstances that do
not ensure the reliability of my perceptions (I am watching a
hologram).
-
Memory inference . If I remember P , then I can presumably
conclude that P is true. For instance, my recollection that
yesterday I had a faculty meeting lends presumptive support to the
conclusion that there was such a meeting. This inference is defeated
if I come to believe that my supposed recollection was an outcome of
my imagination.
-
Enumerative induction . If I observe a large enough sample of
F s, all of which are G s, then I can presumably conclude
that all F s are G s. For instance, if all crows I have
ever seen are black, then I can presumably conclude that all crows are
black. This inference is defeated if I should see a white crow.
-
Statistical syllogism . If most F s are G s and an
individual a is an F , then I can presumably conclude
that a is a G . For instance, assume that (1) the pages
of most printed books are even-numbered on their verso side and that
(2) the bound pages on my table are a printed book. I can then
conclude that these bound pages are even-numbered on their verso side.
This inference is defeated if I discover that these bound pages were
incorrectly printed with even numbers pages on their recto side.
-
Temporal persistence . If it is the case that P at time
\(t_{1}\), then presumably P is still the case at a later time
\(t_{2}\). For instance, if my computer was on my table yesterday
evening (when I last saw it), then presumably it will still be there.
This inference is defeated if I come to know that the computer was
moved from the table after I last saw it, and more generally if I have
any reason to believe that its location may have changed.
General processes of human cognition, such as abduction and analogy (see
Walton, Reed and Macagno 2008) can support further schemes for
defeasible arguments, such as the following:
-
Abduction of a cause. If \(Q\) is the case, and \(P\) causes
\(Q\), then presumably \(P\) was the case. For instance, if the grass
is wet, and rain causes the grass to be wet, then presumably it has
rained. Arguments based on this warrant can be defeated in different
ways: by indicating alternative, no less probable, causes of the
effect (e.g., somebody has watered the grass), or by showing an
inconsistency between the cause (rain) and other states of affairs
(e.g., the street is not wet, as it should be if it had rained), etc.
-
Basic analogy. If \(P\) is relevantly similar to \(Q\), and
\(P\) has property \(R\), then presumably also \(Q\) has property
\(R\). For instance, if detecting something by just seeing does not
count as search, and detecting something by just seeing is relevantly
similar to detecting drug with a sniffing dog, then also detecting
drug with a sniffing dog does not count as search (for refinements of
the analogy pattern, and for a discussion of the sniffing dog case,
see Walton, Reed and Macagno 2008, Ch. 2). Arguments based on this
warrant arguments can be attacked by questioning or denying that there
is a relevant similarity, or by pointing to relevant differences, by
bringing counterexamples, etc. (I cannot enter here the discussion on
what may count as a relevant similarity or difference, on analogy see
the Chapter by B. Brozek). In many cases, an analogical conclusion can
(or should) also be supported by a more elaborate piece of reasoning,
where the aspects that make the similarity relevant are presented as
the antecedents of a general warrant (e.g. detecting something without
actively interfering does not count as search) which is abducted to
explain the common conclusion, see Walton, Reed and Macagno 2008, Ch.
2, Brewer 1996).
These defeasible warrants are not meant to substitute the logical,
philosophical or psychological theories of the phenomena they address,
such as perception, induction, abduction or analogy (see for instance,
for analogy, Holyoak and Thagard 1996). They should be rather viewed as
rules of thumb that may be supported, explained and constrained by such
theories.
In the previous examples, I have considered further general defaults,
such as those enabling the argument from good or bad consequences or the
argument from expert testimony. I have also observed that more specific
defaults may be used to construct defeasible arguments: empirical
generalisations, as well as legal and moral norms, can be viewed as
defaults. In fact, the set of the defaults that may be used in
individual and social cognition cannot be reduced to an exhaustive list,
since default warrants are justified pragmatically, i.e., because of how
well they serve the needs of different practical or epistemic activity
types (Walton and Sartor 2008). The successful use of a default warrant
in a social activity (such as legal reasoning) critically depends on the
extent to which the scheme enjoys shared acceptance, as providing valid
support to its conclusions (since the default’s acceptance is a crucial
precondition of its successful use in arguments meant to convince other
people, or to converge with them into shared conclusions). Thus, even
abstract legal principles, such as interpretive canons, only justify
their conclusions in those legal systems in which they are in fact
endorsed and deployed, so as to enjoy the status of social and
institutional normative principle.
It is important to stress that defeasible arguments can include multiple
steps. For instance, in an argument culminating in the conclusion of a
rule, the rule may be supported by an interpretive argument, while
rule’s factual antecedent may result from arguments assessing the
available evidence. Consider the liability case illustrated in Figure
16. The argument for the liability of Doctor Mary includes the
following:
-
A norm-based argument that Mary is liable, since she harmed her
patient and doctors are liable for harming their patients unless they
are shown not to be at fault.
-
A teleological interpretive argument (a subspecies of the argument
from good consequences): the law on doctors’ liability must be
interpreted in this way, since this interpretation contributes to
increasing diligence in the medical profession, which is a good thing.
-
An empirical argument based on an expert testimony supporting the
conclusion that there was a causal link between the Mary’s behaviour
and the patient’s harm.