8127/frame/fm

 
a.
A Bayesian Framework for Inferential Thinking — Principle 
or Practice
The literature written over the last two decades that addresses inference of
contact is almost exclusively Bayesian in outlook. We do not attempt here to
review all of the various arguments in favor of the Bayesian framework as a
tool for inferential thinking, nor do we attempt a rigorous mathematical
treatment. We acknowledge the pivotal role played by Evett in pioneering its
use, and recognize the valuable contributions made by scientists Buckleton,
Taroni, Margot, Champod, Aitken, Stoney, Walsh, Curran, and Weir and by
jurists Robertson, Vignaux, Kaye, and Thompson. We refer the interested
reader to the various publications by these and other authors, some of which
are listed in the references at the end of this chapter. The best one-stop, overall
reference is Aitken and Stoney (1991), whereas a specific application to DNA
evidence is found in Evett and Weir (1998).
Although most of the literature on inference in forensic science is decid-
edly Bayesian-centric, a major dichotomy exists between that literature and
current practice. Common practice in forensic science is to present a report
and testimony regarding source determination. When asked to define the
significance of an examination, practicing criminalists usually respond with
frequencies if they have them, or general word descriptions that amount to
the same thing if they don’t (notwithstanding the criticisms that have rained
 
Association
 
:
An inference of contact between two objects, the source of the
evidence, and the target on which it was found.
8127/frame/ch07 Page 170 Friday, July 21, 2000 11:46 AM

---

Association and Reconstruction — Inference of Contact
171
down on that practice). Even when likelihood ratios (LRs) are used, they are
generally restricted to comparing hypotheses regarding the source of the
evidence. We rarely encounter a report that formally evaluates an inference
of contact between source and target. This has presented a major difficulty
in writing this chapter. While we are convinced that these ideas are both
legitimate and useful, they have not been generally embraced by the practic-
ing community of criminalists,* nor have they undergone the refinement
that only comes with use over time. We present them here as a tool for the
community to consider, while acknowledging that they have not gained
widespread use in practice. The remainder of this discussion will focus on
how a criminalist might contribute to the proposition of contact using a
Bayesian framework. We will assume here that you have read the introduction
to Bayesian thinking offered in Chapter 6.