The Safety Paradox of Self-Driving Cars

Abstract

Self-driving cars are undergoing extensive road tests and should enter the market within the decade, but consumers continue to worry about the safety of autonomous vehiclesdeven though most traffic accidents are caused by human errors that are avoidable with automation technology. Four experiments investigated how a vehicle's automation level affects its perceived safety, why excessive safety concerns prevail, and how they can be mitigated. In all experiments, participants read descriptions of full-automation (Level 5) and high-automation (Level 4) self-driving cars: Participants consistently perceived Level 5 (vs. Level 4) vehicles as less safe. The effect persisted when objective safety information was available. A mediation analysis and a mediation-by-moderation approach suggest that the negative effect of the automation level on perceived safety is driven by the perceived lack of control over driving. Finally, the effect disappeared when participants imagined themselves as passengers rather than drivers, offering a practical implication for managers.