Gold in "Certainty Into Probability" proposes to do away with the requirement of "but for" causation in toxic tort litigation and to replace it with an apportionment of liability among all potentially causative factors, whether tortious or not, so long as each factor passes the Milward v. Acuity holistic-weight-of-the-evidence-via-subjective-judgment-performed-by-an-adequately-credentialed-person "general causation" test. The impetus for the proposed change is despair. Despairing that scientific discovery seems only to propagate more and more causal associations he concludes that they are each ultimately "population-based", and thus probabilistic. The reasoning goes as follows: not every predisposing gene generates a cancer, not every environmental encounter produces a biomarker record of it, and not every toxic insult produces a damaged chromosome to confirm it. Therefore causation is more like a probability density function and less like the ordinary cause and effect relationship of hammer-missed-the-nail and sore thumb.

To illustrate he gives an example: "The plaintiff worked for many years at a job that daily entailed the inhalation of vapors of some solvent. After the appropriate latency period, the plaintiff's doctors diagnosed a particular cancer. Undisputed, peer-reviewed scientific research shows that (1) cells of this cancer almost always have mutations in a particular set of genes; (2) DNA treated with the solvent in vitro is statistically significantly more likely than untreated DNA to contain mutations in at least some of those genes; (3) the frequency of this cancer among exposed individuals with a specific susceptible genotype is slightly, but statistically significantly, greater than among unexposed individuals of the same genotype; and (4) no other specific risk factors have yet been identified for this cancer, so the vast majority of cases are considered unexplained or idiopathic." (The foregoing, by the way, "should be considered sufficient to support a factual finding of "general causation", according to Gold.  What that means is despite the fact that (1) genome-wide association studies attempting to tie certain genes or groups of genes to certain cancers are notoriously prone to spurious findings, (2) biological associations resting on simple statistical significance are usually false, and (3) conclusions derived from epidemiological studies finding "slightly" increased risk are almost always false, if you add enough zeros together a jury can somehow eventually conclude that the likelihood the solvent is a human carcinogen is greater than 50%. Admittedly, that is indeed the central teaching of Milward v. Acuity ... but we digress.) The hypothetical plaintiff also has a "susceptible genotype" thereby further adding to risk.

So how to do justice? Under Gold's approach the fact-finder would be asked to determine whether the solvent was responsible for "a non-trivial increase in plaintiff's risk of developing [the (presumably)] cancer. Upon an affirmative answer, the fact-finder would estimate, as a percentage, the extent to which the solvent exposure contributed to the plaintiff's risk of cancer." That percentage would then be multiplied by the damages to determine the award. All other non-tortious factors go into the mix of "background" risk. So how might it work?

Let's start with an example. Think of cancer risk as a six-sided die. If you roll a 1 you get cancer and die of it. Now imagine that the die has been weighted and the odds of getting a 1 have gone from 1 in 6, or 16.67%, to 1 in 5, or 20% (which conveniently also works out to a 20% increase in risk). Under the "more likely than not" standard of "but for" causation the plaintiff must lose as, out of 1,000 people exposed to the solvent who also got cancer, 167/200 would have gotten it anyway. Per Gold however, the entity that weighted the die may be sued nonetheless, and the 20% increase in risk, obviously beyond de minimis, entitles the plaintiff to 20% of his damages. We can't figure out which 33 of the 200 victims developed cancer as a result of the solvent but collectively justice is done. Cosmic justice. What's not to like?

At least three things.  First, there's the problem of trying to say something about an individual based on population frequencies. Indeed such an approach is so grossly prone to error that it's got it's own name: the ecological fallacy. Beware the tyranny of the central tendency - failure to do so results in anticipating your being perfectly content with your feet in the freezer and your head in the oven. Remember also the case of the Pliofilm workers' relative risk. Whether working with benzene significantly increased their risk of acute myelogenous leukemia depends more on the year each observer measured their outcomes than the benzene to which they were exposed.

Second there's the frequentist issue; which is to say the interpretation of the frequencies calculated in the hypothetical solvent study two paragraph above means that somewhere, necessarily elsewhere and in an infinite space, the number of people developing cancer as a result of the solvent exposure is exactly 20% greater than would otherwise have been the case. Extrapolating from a typical epidemiological sample size of a few hundred to maybe a few thousand, to infinity, is quite a leap. It's like pronouncing what nature is like based upon your examination of an impatiens in a pot on your patio.

Third, assuming arguendo that something like "a 20% increase in risk" is both accurate and has some concrete meaning beyond games of chance where rules (seem to) encompass and control for all known variables, Gold's formulation violates his own edict of "risk as cause, not as harm". To demonstrate let's go back to our hypothetical above of 1,000 workers exposed to a solvent that increases their risk of cancer by 20%.

Normally we'd expect, through no fault of anyone's save Mother Nature's, 167 of the workers to develop cancer. Instead, taking as true and absolute the 20% increase in risk, 200 workers develop cancer. However, all 200 (per Gold) have a viable claim to 20% of the lost wages, medical bills, pain, suffering and mental anguish even though the cancers of 167 of them had nothing to do with exposure to the solvent. For what then are they being compensated? Not a harm that manifested from the risk. In them it manifestly did not. Rather, the vast majority of plaintiffs wind up being compensated solely for a risk that never produced a harm. Tell us that won't increase the number of claimants in toxic tort litigation.

In the end, Gold's formulation only does justice to an individual if you believe that (1) epidemiological studies generate risk measurements nearly as accurate as "what are the odds of drawing a face card from a standard deck of cards" and (2) the individual's circumstance is best represented not by what actually happened to him but rather by the sum of all his possible outcomes. Unfortunately, the evidence for both (1) the ability to accurately predict the fate of an individual based on population studies and (2) the multiverse, ranks no higher than that for unicorns and elves.