Bodies of Evidence

If one piece of evidence looks just plain wrong (or is overly influential in the analysis) it stands out and can be replicated. But this will not torpedo the entire line of inquiry, which relies on the many pieces of evidence within a Body of Evidence. Bodies of Evidence+ overcome many of the difficulties with causality as discussed elsewhere. We are looking for many pieces of evidence all pointing in the same direction. Our conclusions can be asserted beyond a reasonable doubt. We deal with exceptions on an exceptional basis.

A large body of poorly constructed evidence doesn't magically get better. The body of evidence approach works best if each single item of evidence is independently constructed, and constructed within rigorous controls to ensure its effectiveness. We include evidence based on cheap heuristics+ but only in moderation, and labeled with the appropriate warning messages. For example, in the Life Sciences industry statistical evidence is admissible, but only to the extent empirical evidence supports the purported causality of the statistics. No assumption of causality supported merely by a veil of numbers is permitted. We seek a body of evidence, each one of which can potentially withstand the assault of a determined adversary.

One caveat: we do not require a body of evidence to take action on issues of product safety. Although our product may ultimately be found innocent, it's prudent to limit possible damages should that not be the case. See topics on Expect the Unexpected+ or Product Safety for more.

For approaches on combining heterogeneous evidence (i.e., circumstantial vs. direct) into a complete whole, see:

Further Reading