Deconstructing Milward (part II)
Plaintiff's expert reached his opinion that benzene is capable of causing acute promyelocytic leukemia (APL) in humans by employing a "weight of the evidence methodology. The U.S. Court of Appeals for the First Circuit decided that's ok. Consequently, since no researcher had previously reported an increased incidence of the illness following exposure, since the conjecture that benzene causes APL has never been tested, and since "even delusions are permitted as hypotheses" (see p. 19) in science it's fair to say that the First Circuit has adopted an "anything goes" approach to science in the courtroom (at least so long as your expert is bright enough or at least prepared enough to intone a few magic words). For today let's start with this "weight of the evidence methodology" business.
The court was impressed that plaintiff's expert reached his opinion using a process of elimination approach polished into the more formal-sounding "inference to the best explanation". Essentially the expert picked the evidence he thought was important, weighed it in the scales of his "professional judgment" and found that the theory advanced by his employer was the correct one. The court should have been shocked! Shocked! Instead it was satisfied that an expert weigher paid (as they all are in our advocacy-centric system) to have a thumb on the scale had fairly come to an admissible opinion thanks to his sound methodology. But is basic decision theory the sort of "method" the Supremes had in mind when they decided Daubert?
When you think of methodology, whether in the context of Bendectin or skid marks, you think about whether the expert used sound epidemiological principles or had a sound grasp of coefficients of friction and the relationship between force, mass and acceleration. If she got the methods right and from there opined as to why a justifiable conclusion is the correct one using a "weight of the evidence" approach then sure, let the jury decide. But if the expert chooses what to weigh you know that selection bias is at work; and when the data he decides to weigh is unsupported you know another sort of bias is at work; and when the only defensible items left in his scale are temporality and plausibility then you know that all he has is an untested hypothesis. What the court fails to understand is that the weighing of the evidence doesn't even begin until there is a "validly observed association". Otherwise, all plaintiff's expert has done is to propose a theory that ought to be tested.
When all you have is an untested hypothesis doesn't defendant win? Doesn't plaintiff yet carry the burden of proof? Not in the First Circuit, if Milward stands.
