Causation in Personal Injury Law: The Case for a Probabilistic Approach

Abstract

This paper makes the case for a wider acceptance of a probabilistic approach to causation in negligence. This acceptance would help to remove much of the incoherence which has come to afflict the English law of personal injury law. This incoherence can also be found in other common law jurisdictions (notably those of the United States, Canada and Australia). Concentrating upon recent UK case law, the argument opposes the contention that ‘naked statistics’ can play no role in establishing causation. The argument is controversial but it can be reduced to three unremarkable grounds: (1) With its acceptance (albeit in certain carefully prescribed circumstances) of liability for a negligently increased risk which has eventuated, the common law has already embraced a probabilistic conception of causation; (2) The English common law already employs a probabilistic (frequentist) approach to identifying coincidences; and (3) With the ‘balance of probabilities’ as the standard of proof in civil cases, the common law has long had a probabilistic (epistemic) concept at its core. Probabilistic approaches (at both the type and token level) are shown to be consistent with laypersons’ understanding of the concepts such as risk, chance, odds and likelihood. Moreover, a wider acceptance of a probabilistic perspective on causation would entail no major challenge to the fundamental aims of tort, viz. deterrence and corrective justice.

Appendix: Lewis’s Principal Principle

Let C represent some credibility function (epistemic probability) that, at time t, some event A will occur, and that X is the proposition that A had an objective chance of x. E is any proposition compatible with X that is admissible at time t. Then according to Lewis (1986: 87):

$$ {\text{C}}({\text{A}}|{\text{XE}}) $$

For the purposes of this article, suppose X is the proposition that A has a chance x at time t of being the cause of some injury, and E is the overall evidence which is compatible with X, then the above equation relates belief probability to objective chance. Lewis (ibid.) further explains that if the evidence E supports a very high degree of belief in X, that is, that the objective chance of A is causal truly is x, then C(X|E) is very close to one, then it follows that the unconditional epistemic probability that A is causal C(A|E) = x. (By unconditional, we mean conditional only on the existence of that evidence E which persuades us the truth of X.) Only when the fact-finder in a trial has this high degree of confidence in the evidence, can she allow the objective chance to ‘fix’ her epistemic probability in the truth of the causal proposition A, which she then gauges against the BOP standard.

Lewis emphasises that X is ‘the proposition that the chance, at time t, of A’s holding equals x… [and with x, he is] speaking of objective, single-case chance—not credence, not frequency’ (1986: 90). His aim is to relate this token probability to some credence that X will obtain conditional upon the degree of belief in x. In contrast, Sober’s ‘connecting principle’ (1986: 101) assumes the existence of a population, of cause events C and of effect events E, from which a type probability can be measured. The closer this type probability is to 1.0 (that is, the fewer the instances of the effect arising in the absence of the cause), the greater the support for the hypothesis that a single C event at time t₁ token-caused a single E event at t₂. Clearly, under either principle, the lower the estimate of the token probability, the less secure is any claim that one token event was the cause of another.

Philosophy aside, if all the other conditions of the Fairchild exception are satisfied, a negligently increased risk (objective chance) of x < 0.5 should, if it has eventuated in injury, incur liability.