A Framework for Proactive Interaction in Automated Vehicles

As AI advances, and can draw upon previously disparate information sources, automated vehicles are gaining the ability to proactively collaborate with humans to accomplish everyday tasks. This collaboration will be based on increased sensing and information and will feature greater proactive behavior by the vehicle. However, there is little guidance on how vehicles should behave proactively, or even a shared understanding of what that means. Using prior work, we develop a framework of proactive interaction comprising three dimensions: initiation of action, alignment with users’ goals, and communication of context awareness. We apply the framework in a video-based online study (N=351) of a vehicle navigation system that exhibits proactive behavior. The system advises or acts, supports or counters users’ goals in two scenarios (to reach a work meeting quickly or carry groceries without walking too far), and provides a more or less detailed rationale. We find that when goals are misaligned, communicating contextual awareness decreases user experience and acceptance. As a result, if alignment of (immediate) goals is not possible—such as when drivers want to reduce walking distance but the closest parking lot is full—systems should exercise care in communicating their situation awareness, as this can reinforce users’ negative perceptions. Dispositional traits such as propensity to trust and locus of control do predict users’ trust, distrust, and experience in our scenarios. We also find that participants’ primary rationales for accepting the system’s decision are largely similar (saving time, trusting, and making sense) but rationales overall, especially for rejecting the system’s decision, are more varied than expected and often based on experience. Therefore, systems that look to personalize decision-making—such as whether to proactively execute versus merely suggest an action—could take into account users’ dispositional traits and other preferences, such as those as exemplified by these rationales.