Sensitivity Auditing of Simulation Work for Autonomous Vehicles: Interviewing the Modelers. Results from i4Driving European Project (Project no.101076165).

Aims: The main objective of this study, part of i4Driving project, is the Sensitivity Auditing (Saltelli et al., 2013; 2020; 2023), of Autonomous Vehicles (AVs) modelling. Sensitivity auditing is an approach recommended by existing EU guidelines (2021) which consists of a wider consideration of the effect of all types of uncertainty, including structural assumptions embedded in the model and subjective decisions taken in the framing of the problem that is critically monitored by the extended multi-stakeholder community that is responsible for validating the model. The use of iterative consultations among modelers during the sensitivity audit ensures that perceptions and ‘techno-scientific imaginaries’ about AVs are thoroughly analysed and that possible blind spots produced by accepted norms and beliefs are identified. The ultimate objective of sensitivity auditing is to ensure the plausibility and credibility of developed models. Methods: We zoom on an aspect of sensitivity auditing that suggests exploring the worldviews and sociotechnical imaginaries of the actors involved in the modelling. The method used for multi-stakeholder consultation is the structured Delphi for the consortium i4Driving through the interview, to use information acquired from modelers to improve the realism of AVs to various factors and reduce risks and uncertainties incorporated in the model. Combined with this, interviews were conducted with sociologists who are experts in autonomous driving in the relevant areas of social defense and public investment. Results: Content analysis has been developed on the main subject areas and isotopes (constant in variations), including social factors that enable AV penetration and its benefits; the interaction between AVs, and vulnerable road users; the wider and longer-term social, economic and environmental impacts of AVs (public transport, changes in energy consumption, traffic flow, insurance market, role of the public sector). Sentiment analysis tools are used for some themes to register the perceptions of modelers.