Notes
1 Thus, as recent scholarship has emphasised (Paavola
2006, McKaughan 2008, Campos 2011), abduction should not be identified
with the so-called Inference to the Best Explanation.
2 The point has been raised by Frankfurt (1958) and
Kapitan (1997). McKaughan (2008) defends the pursuit-worthiness
interpretation against generative interpretations. Schaffner (1993)
similarly criticizes Hanson (1958) for obscuring the distinction between
the generation and preliminary evaluation of hypotheses.
3 Interpretations along these lines have been proposed
by Fann (1970), Curd (1980) and Psillos (2011).
4 Similar objections were raised against Hanson (1958)
and others who defended abduction as a “logic of discovery.” See
Nickles (1980) for several contributions to this debate and Schaffner
(1985) for ones focused on medical diagnosis.
5 Stanley and Campos (2013, 306) call the former of
these “creative abduction” and the latter “habitual abduction.”
6 Sometimes ‘differential diagnosis’ is instead used
to refer to the process or method of considering and distinguishing
different diagnostic hypotheses (e.g. Sox, Higgins and Owens 2013, Ch.
2). We will here use the term only to refer to a set of competing
hypotheses, rather than the process of generating or selecting between
these.
7 A similar point also applies to treatment: even if a
given treatment fails to alleviate a patient’s symptoms, it may still
provide valuable clues for further investigations. The line between
treatment and testing is not always a sharp one.
8 The following case is developed on the basis of the
clinical experience of one of the authors. While we do not make any
claims as to how statistically common scenarios of this type are, we
regard it as sufficiently typical to illustrate the ability of our
framework to explicate patterns of clinical reasoning.
9 ‘Decrescendo diastolic murmur’ means that a sound of
decreasing intensity occurs during diastole, i.e. period of the cardiac
cycle where the heart relaxes and refills with blood.
10 For a recent survey of currently available
programs, see Bond et al (2012). Philosophers in the 1980s debated
whether computer programs could in principle replace all aspects of
diagnostic reasoning (e.g. Schaffner 1985, Wartofsky 1986). We here
focus on how useful currently existing programs are for the task of
generating hypotheses.
11 An aneurysm is an abnormal widening of a blood
vessel. This can cause weakness in the wall of the vessel. A dissection
is a rupture of the blood vessel where blood flows into the layers of
the wall of the vessel, forcing them apart.
12 For instance, threshold models highlight the
importance of weighing initial probability against the potential
benefits and harms of testing or treating. Similarly, Bayes’ Theorem can
highlight importance lessons about probabilistic reasoning. Thus, they
may provide useful analytic frameworks for teaching clinical reasoning
(Sox, Higgins and Owens 2013). We are less optimistic about proposals to
reform clinical practice to conform more closely to probabilistic
models, proposed e.g. by Richardson (2007); see Marewski and
Gigenrenzer’s (2012) critique of information-greedy procedures for
clinical decision-making.
13 Kassirer, Wong and Kopelman, (2013, 46) also
mention aanalogies between expertise in chess and in medical diagnosis.