Inclusive Design Strategies for HMI Trust Calibration inRisky Contexts

Abstract

In human-vehicle collaborative environments, drivers’ appropriate level of trust in automated systems is crucial for smooth takeover in risky situations. However, due to the wide variations among drivers—such as gender, cultural background, personality traits, and driving styles—it is challenging to design a Human Machine Interface (HMI) trust baseline that works for everyone, particularly in high-risk takeover scenarios. Excessive trust or dis-trust can lead to takeover delays and potential abandonment of the automation technology. This paper proposes an inclusive design strategy for HMI trust calibration to address this issue. This strategy determines risk metrics through drivers’ subjective perceptions and ob-jective braking distances, mapping various HMI visual elements to individual subjective trust levels, then employs Generative Adversarial Network (GAN) to train and generate datasets of interface design feature combinations. Based on personalised operator data, the strategy enables the inclusive layout of interfaces, aiming to enhance the adaptability of current trust interaction design methods in complex scenarios and personalised operational conditions, thereby ensuring safety in human-machine collaboration.