Several studies have been conducted some decades ago to understand driver drowsiness, its assessment and the ways to mitigate its effects to avoid or reduce fatal road crashes. The driver self-report is a subjective measurement well admitted in the drowsiness context. However, studies that compare the driver’s self-report with observer ratings are very scarce. This study compares driver self-reports using the Karolinska Sleepiness Scale (KSS) and trained observer ratings. 25 subjects were included in this study that took place in a real road environment. During the driving sessions, the drivers were under sleep deprivation or not. 6 independent raters were trained by human factor experts and assessed the driver drowsiness by analyzing the videos. The results showed significant differences between driving with and without sleep deprivation both for observer ratings and driver self-reports. Also, the results showed that the driver’s self-reports tended to underestimate or overestimate the drowsiness state, leading to some inconsistencies between self-reports and observer ratings. This study allows us to see that the observer ratings approach leads to measuring the sleep deprivation effects in drivers. This approach leads to enhancing the assessment accuracy of driver drowsiness state and can be a complementary means in addition to driver self-reports to strengthen the ground truth.

• Driver self-reports
• Observer ratings
• Assessment accuracy challenges
• Complementary means to strengthen assessments

• How can we collect accurate neurophysio-behavioral-subjective data for reliable driver-state monitoring?
• What methods ensure consistent and reliable behavioral inputs for DMS?
• What challenges arise when aligning a system’s algorithm output with actual driver states?
• How do varying conditions and driver demographics affect the accuracy of ground truth data?
• What strategies ensure ground truth data covers diverse driving scenarios and behaviors?