Meike Kühne, Tim Schrills, Markus Gödker, Patricia Wollstadt, Thomas Franke, "Subjective Information Processing Awareness for Intelligent Charging Agents - Connecting Traceability, Trust & Users’ Ability to Predict", DGPS Kongress 2022, 2022.

While interacting with artificial intelligence (AI), users experience automated information processing, which can remain untraceable to them. This involves evaluating options in the area of intelligent bidirectional charging of electric vehicles (EV). Untraceable information processing can have negative effects on the cooperation between humans and AI, since it will not be recognizable to humans according to which reference values specific charging processes are evaluated and compared. To prevent an unjustified loss of trust, the user must be able to experience the system as traceable. Intelligent charging management preserves buffers that can be used for grid stabilization. However, the use/reservation of EV in e.g. a carsharing fleet limits their use for this purpose. How can the extent of this limitation be communicated in terms of cooperation? And how can the understanding of the interaction with AI be advanced? In an experiment with N=57 laypersons, we modified the amount of information a smart charging algorithm presented in a between-group design. Participants were repeatedly shown the system’s estimated fit of a planned EV reservation and the potential to use EVs for bidirectional charging as well as the information it processed. We also examined to what extent participants were able to predict the estimated fit. While presenting more information partially resulted in an increase in trust and SIPA, no improvement could be seen in the subjects' estimation of fit. Further research should investigate how to avoid effects comparable to a placebic explanation and develop metrics that are easier and faster to learn by users.