Assessing Drivers’ Mental Model Of Advanced Driver Assistance Systems Using Signal Detection Theory

Abstract

Previous studies evaluated drivers’ knowledge of advanced driver assistance systems (ADAS) using different kinds of percent-correctness-based mental model scores (MMS), which makes cross-study comparisons difficult. To resolve this issue, our study explored the use of sensitivity (i.e., d-prime ( d’)) and response bias (i.e., criterion location ( c)) in signal detection theory (SDT) as a measure of drivers’ ADAS mental models. Based on the data collected from a survey among 287 ADAS users, regression models were fitted, and it was found that d’ and c accounted for a large variance when estimating drivers’ ADAS mental models as measured by MMSs (adjusted R 2 > 0.8). Further, predictors of MMSs were also predictors of d’ and c, but d’ and c include additional information that was not covered in MMSs. These findings support the usage of d’ and c as standard metrics for assessing drivers’ ADAS mental models in future research.