Assessing beyond a reasonable doubt' without probability: an info-gap perspective

The concept beyond a reasonable doubt' is fundamental, procedurally and normatively, in legal proceedings. It generates operational criteria for triers' decisions, and reflects the high social value of protecting innocent individuals. Uncertainty in judicial evidence is prevalent, motivating the concept of beyond a reasonable doubt'. Beyond a reasonable doubt' is traditionally operationalized with ethical definitions. Recently, qualitative and quantitative probabilistic concepts have been advocated and disputed for defining beyond a reasonable doubt'. Our first claim is that much judicial uncertainty is consistent with Knight's concept of true uncertainty', and inconsistent with the normalization and additivity axioms of probability. The prevalence of Knightian uncertainty in judicial evidence results from indeterminism in human affairs as developed by Shackle and Popper. Our second claim is that beyond a reasonable doubt' can be operationalized with the concept of robustness as developed qualitatively in info-gap decision theory. A trier's decision is highly robust to uncertainty in the evidence if only very great change in the evidence would undermine the decision. When facing Knightian uncertainty, a decision is acceptable beyond a reasonable doubt if it is highly robust to uncertainty in the evidence. We illustrate the qualitative implementation of this criterion with several examples.