Toward Adaptive Trust Calibration for Level 2 Driving Automation

Akash, Kumar; Jain, Neera; Misu, Teruhisa

doi:10.1145/3382507.3418885

Cited by 29 publications

(7 citation statements)

References 44 publications

(46 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…However, these models have linear assumptions of human trust behaviors. Considering these limitations, machine learning models became popularly used to predict driver trust, such as support vector machine (SVM), K neighbor (KNN), and quadratic discriminant analysis (QDA) (32)(33)(34)(35). For instance, Akash et al (33) applied electroencephalography (EEG) and GSR to identify users' trust in an obstacle detection system (as used in AVs) by the QDA.…”

Section: Trust Estimation and Recognition Model In Automated Vehicle ...mentioning

confidence: 99%

Measurement and Real-Time Recognition of Driver Trust in Conditionally Automated Vehicles: Using Multimodal Feature Fusions Network

Cao

Song

et al. 2023

Transportation Research Record: Journal of the Transportation R

View full text Add to dashboard Cite

Trust calibration is essential to prevent misuse and disuse of automated vehicles (AVs). Accurate measurement and real-time identification of driver trust is an important prerequisite for achieving trust calibration. Currently, in conditionally automated driving, most researchers utilize self-reported ratings as the ground truth to evaluate driver trust and explore objective trust indicators. However, inconsistencies between the subjective rating and objective behaviors were reported, indicating that trust measurements cannot rely solely on self-reported ratings. To fill this research gap, a method of subjective and objective combination was proposed to measure and identify driver trust in AVs. Thirty-four drivers were involved in a sequence of takeover events. Monitoring ratio and subjective trusting ratings were collected, and combined to measure driver trust levels (i.e., higher and lower trust). Compared with the subjective measurement, the hybrid measurement can more reliably evaluate driver trust in AVs. More importantly, we established a real-time driver trust recognition model for AVs using label smoothing-based convolutional neural network and long short-term memory network fusing multimodal physiological signals (i.e., galvanic skin response and electrocardiogram) and interactive experiences (i.e., takeover-related lead time, takeover frequencies and system usage time). Four common models were developed to compare with the proposed model: Gaussian naive Bayes, support vector machine, convolutional neural network, and long short-term memory network. The comparison results suggest that the performance of our model outperforms others with an F1-score of 75.3% and an area under curve value of 0.812. These findings could have implications for the development of trust monitoring systems in conditionally automated driving.

show abstract

Section: Trust Estimation and Recognition Model In Automated Vehicle ...mentioning

confidence: 99%

Measurement and Real-Time Recognition of Driver Trust in Conditionally Automated Vehicles: Using Multimodal Feature Fusions Network

Cao

Song

et al. 2023

Transportation Research Record: Journal of the Transportation R

View full text Add to dashboard Cite

show abstract

“…Wang et al (2020) showed how guidance can be displayed using colored lanes on the windshield. Other information that can be presented in AR can include navigation arrows and bounding boxes for the scene objects (Liu et al, 2021;Akash et al, 2020). For route guidance, one can also superimpose a lead vehicle to help guide the human driver on their route.…”

Section: Introductionmentioning

confidence: 99%

Analysis of Driver Behavior in Mixed Autonomous and Non-autonomous Traffic Flows

Liu

Boyle

2022

Proceedings of the Human Factors and Ergonomics Society Annual

View full text Add to dashboard Cite

Autonomous vehicles are expected to improve road safety and efficiency in future transportation systems. A driving simulator study was designed to identify driving styles and the cooperation between human drivers and other AVs. The study captured driver’s following behavior in a fully autonomous driving environment at unsignalized intersections. Participants were asked to make a series of maneuvers (straight through intersection, left turn, and right turn) in two different speed conditions (30, 40 mph) and two different traffic density conditions (with or without other traffic). Analysis of Variance showed that drivers had a significantly larger deviation (defined as the area between two trajectories) during left turn maneuvers when they were traveling at higher speeds. Moreover, the first turning operation had smaller deviation than the second turning operation. The findings have implications for the design of driver-assistance guidance systems in future mixed autonomous and non-autonomous traffic flows.

show abstract

“…Many studies have investigated the possible solutions to calibrate trust. These include paradigms that anticipate human behaviors-such as trust-and inform humans to make optimal choices [7], [8], [9]. However, a primary challenge for such an approach is quantitatively predicting human trust in real-time.…”

Section: Introductionmentioning

confidence: 99%

“…However, it is challenging to obtain self-reports repeatedly without interrupting the task. As an alternative, recent works have developed dynamic models to capture human trust and estimate it in real-time [13], [14], [15], [9]. There are two approaches for developing such models: a general model for the whole population and a personalized model for each individual to account for individual differences [16].…”

Section: Introductionmentioning

confidence: 99%

Clustering Human Trust Dynamics for Customized Real-time Prediction

Liu

Akash

Misu

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

Trust calibration is necessary to ensure appropriate user acceptance in advanced automation technologies. A significant challenge to achieve trust calibration is to quantitatively estimate human trust in real-time. Although multiple trust models exist, these models have limited predictive performance partly due to individual differences in trust dynamics. A personalized model for each person can address this issue, but it requires a significant amount of data for each user. We present a methodology to develop customized model by clustering humans based on their trust dynamics. The clustering-based method addresses the individual differences in trust dynamics while requiring significantly less data than personalized model. We show that our clustering-based customized models not only outperform the general model based on entire population, but also outperform simple demographic factor-based customized models. Specifically, we propose that two models based on "confident" and "skeptical" group of participants, respectively, can represent the trust behavior of the population. The "confident" participants, as compared to the "skeptical" participants, have higher initial trust levels, lose trust slower when they encounter low reliability operations, and have higher trust levels during trust-repair after the low reliability operations. In summary, clustering-based customized models improve trust prediction performance for further trust calibration considerations.

show abstract

Toward Adaptive Trust Calibration for Level 2 Driving Automation

Abstract: Figure 1: Simulation environment of our user study with augmented reality (AR)-based information presentation. The goal of this research is to calibrate a driver's trust in driving automation through AR-based information while avoiding increased driver workload.

Cited by 29 publications

References 44 publications

Measurement and Real-Time Recognition of Driver Trust in Conditionally Automated Vehicles: Using Multimodal Feature Fusions Network

Measurement and Real-Time Recognition of Driver Trust in Conditionally Automated Vehicles: Using Multimodal Feature Fusions Network

Analysis of Driver Behavior in Mixed Autonomous and Non-autonomous Traffic Flows

Clustering Human Trust Dynamics for Customized Real-time Prediction

Contact Info

Product

Resources

About