Driver or passenger? Use of a Bayesian network for the evaluation of DNA results in a fatal car accident

This article presents a case where the issue was to determine who was the driver and who was the passenger at the time of a fatal car accident involving two persons, one of whom died in the accident. The presence of the two persons in the car was not contested, only the mechanisms that led to the deposition of the DNA (i.e., the activities) were. To our knowledge, few cases are evaluated considering the alleged activities. The reasons for this include the lack of knowledge, and data, as well as the difficulties encountered for the formulation of conclusions. In this case report, we present the architecture of the Bayesian Network (BN) used to evaluate the DNA results of the traces recovered from the steering wheel, driver's and passenger's airbags. The following propositions were considered: "The person of interest (POI) was driving the car and the alternative person (AP) was the passenger at the time of the accident" or vice versa. We discuss the assumptions that were made and how data from the literature was used to parametrize into the BN. A likelihood ratio of the order of 90 was finally assigned. The statement proposed to the mandating authority indicated that, given the information that was made available to us, our observations were of the order of 90 times more probable if the POI was driving the car at the time of the accident rather than if the AP was. A sensitivity analysis was performed (5000 simulations): this shows that our likelihood ratio is robust.