Effect of Reliability, Its Framing and Error Bias on Trust in Human-Vehicle Collaboration

Abstract

System reliability promotes trust, but may also impair human monitoring performance and in turn affects trust. This effect varies across different errors. This study examined the effect of automation reliability (100%, 75%, and 50%) and its framing (negative and positive description of reliability), and error bias (false alarm and miss) on user trust and its related factors in the automated driving system (ADS). Each participant completed 16 trials with human-vehicle collaboration task in a static driving simulator. The results showed that ADS with higher reliability positively impact user trust, but negatively impact situation awareness. Users' trust was higher in false alarm (FA) events than in miss events, but task success and situation awareness were higher in miss events. This study revealed an unusual negative correlation between trust and situational awareness in human-vehicle collaboration and provided possible insights into the internal factors of error bias in automation. Our finding has implications for reliability disclosure strategies and trust calibration.