I shall investigate the nature of the probability of a manufactured item's being malfunctioning upon leaving its production plant. My first claim is that the sentence used to express the probability that an artefact will malfunction can be given two quite different readings in which the probability is supposed to refer to chances rooted in reality. In the first reading, the probability concerns the entity qua physical object, in the second qua instance of an artefact type. The probability corresponding to the first reading, the DYSF-PHYS probability, raises no major problem. The one of the second reading, the DYSF-ART probability, however, is a good deal more puzzling. This paper aims to analyse this last sort of probability. I intend to demonstrate that the DYSF-ART probability is often the one referred to when considering the possibility that artefacts might not work properly, and that this probability is as objective and as rooted in reality as is the DYSF-PHYS one. After presenting and explaining both probabilities, I elucidate what problem it raises to have two sorts of probability of dysfunction instead of one and explain why a frequentist interpretation of both does not solve it. I then investigate the causal basis of the chances measured by the DYSFART probability and explain how these chances may show no dependence on the physical make-ups of the artefacts. In the last part of the paper, I ask how should we interpret and qualify probabilities like the DYSF-ART one, and I make some connexions with biological cases to demonstrate how this investigation might prove helpful to the current discussion of probabilities in evolutionary biology.
