Causation is Hard: Multiverse Edition
While we wait to critique Steve Gold's upcoming defense of Milward v. Acuity (the paper is available at SSRN but, alas, it's flagged as a work in progress, neither to be quoted nor cited) we'll have a go at his new paper on causation titled "When Certainty Dissolves Into Probability: A Legal Vision of Toxic Tort Causation for the Post-Genomic Era".
It takes twenty eight pages to get to the meat of the paper and upon arrival, despite some serious disputes with his take on causality, we were pleased when we finally made it (for awhile at least). That's because it read as though Gold seemingly endorsed the same sort of approach to causal apportionment that we've been advocating here (and elsewhere) for years. Consider his suggestion:
I propose that courts should adopt an expressly probabilistic view of causation when the dominating evidence comprises population-based data of toxic effect. To frame the standard, an exposure should be considered a cause of disease if it was a contributing factor to the disease's occurrence. To be a contributing factor, an exposure would be shown by a preponderance of the evidence - not limited to any single favored type of evidence - to have added incremental risk that the plaintiff would develop a disease that the plaintiff has, in fact, developed. Damages should be apportioned to that contributing factor in proportion to its contribution to the plaintiff's risk.
Thereafter he even goes on to state that de minimis contributions to risk ought not be actionable. So what's not to like? Well, lots, actually. And it begins with a fundamental misunderstanding about "but for" causation.
Gold begins with the claim that "but for" (i.e. counterfactual) causal reasoning somehow doesn't work in toxic tort cases. He writes, "[p]roof of toxic tort claims conform poorly to the traditional deterministic legal model of but-for causation, because toxic injuries almost never involve an easily observed chain of physical events connecting a particular defendant's conduct with a particular plaintiff's harm." Gold's worry is that biomarkers, sequenced genes and the data from molecular epidemiology have delivered only more causal complexity rather than the allegedly promised "deterministic" certainty necessary for old-fashioned legal reasoning. So if "but for" causal thinking doesn't lead to the right legal solution, what does?
After pondering the conclusion that necessarily follows from the realization that e.g. cancer is the consequence of genes + epigenetic regulation + socioeconomic status + gender + microbiota + environmental exposures + unknown unknowns + bad luck (i.e. stochastic processes) Gold thinks he's fallen down the rabbit hole and met Schrodinger's cat and it's both alive and dead all at once. Confusing the uncertainty (which is to say ignorance) described by the statistics used in population studies with quantum indeterminancy, Gold declares that the real problem is one of "scientific indeterminacy". By this he seems to think that because the precise chain of events leading to disease is not only unknown but ultimately unknowable, and because randomness may lie literally at the root of the universe, it is therefore impossible to say that "but for" a particular exposure plaintiff would not have developed her illness.
That's bad enough because we know it's not the wave function that killed Schrodinger's cat . It was the decaying atom; and every time the atom decays the cat must surely die. But it gets worse.
Gold then suggests that the answer is not to live with the uncertainty - i.e. to make our best judgment about whether a gene or exposure or whatever was more likely than not a cause such that the disease would otherwise not have arisen, determine the risk it posed ex ante, and then apportion liability among those non-de minimis tortious risks ex post. Instead he suggests we throw every black box risk factor that passes his lowered Milward v. Acuity test of so-called general causation out for the jury's consideration, bad genes included and then let the them apportion liability among all risk factors, tortious and otherwise. He's thus changed the formula from Liability = Duty + Breach + Causation + Harm, to Liability = Duty + Breach + Risk + Any-disease-that-has-ever-been-associated-with-that-risk-even-if-it-probably-wasn't-causative-in-this-case.
In part II tomorrow we'll explain why such a formulation would vastly expand the number of defendants swept up into toxic tort litigation and why his belief that even with a tiny risk "someone somewhere" is ultimately harmed is true only if A) risk factors mean something beyond how likely it is we're wrong about our causal inferences; and, B) there's a whole lot more universes out there besides our own.
