When the automated driving system fails: Dynamics of public responses to automated vehicles

• A field study was done to examine the impact of automated driving system (ADS) failures. • A between-subjects, pre-post design was adopted. • ADS failure caused negative influences on various responses to AVs and ADS. • Negative influences of ADS failures should be mitigated in human–machine co-driving. Automated vehicles (AVs) will be at the stage of human–machine cooperative driving probably for a long term and their control is shared by two drivers: the human driver and the automated driving system (ADS). ADS failures are a major contributor to the disengagements of AV testing on public roads and influence human–ADS interactions. The impact of ADS failures (malfunctions or limits) was only examined in the driving simulator environment and mainly on self-reported trust in the ADS or AVs. Utilizing simulators to investigate the influence of ADS failures might lack ecological validity. Participants in this environment might find it hard to imagine the risk of ADS failures and the criticality of taking immediate manual control. We conducted a field study with a real Level 3 AV and a large-scale test track. We invited participants ( N = 261) to experience the AV and an ADS malfunction as passengers in a naturalistic driving environment, exposing almost half the participants to the malfunction. Witnessing the malfunction caused participants’ reductions in six psychological responses to AVs and ADS (trust, positive affect, perceived usefulness, perceived ease of use, attitude, and behavioral intention) and increased negative affect. This finding highlights the importance of mitigating the detrimental influences of ADS malfunctions. Trust repair deserves consideration in maintaining adequate human–machine cooperation in AVs after ADS malfunctions.