A responsible artificial intelligence framework for forensic science
Janet Stacey, Rachel Fleming, Dion Sheppard, J. Brigstocke Sheppard, Gillian Dobbie, Deepak Karunakaran
Abstract
As artificial intelligence and automated workflows become embedded in forensic science, there is a need for a comprehensive framework to ensure that they are fit for purpose. A review of existing AI guidelines and policies has identified a commonality of principles but an absence of the level of detail required for an organisation to implement these expectations at an operational level. In response to this, a Responsible AI Framework (RAIF) to support the safe and reliable development and implementation of AI projects within a forensic organisation has been developed. The RAIF consists of three components: a Questionnaire, a Guidelines document, and a Project Register. When used in combination, the RAIF enables organisations to have confidence when undertaking AI projects and to balance the opportunities and risks of this evolving technology. A fully worked example illustrating the application of the RAIF to Lumi Drug Scan, a forensic AI solution to detect illict drugs in real time, is included.