Haptic Shared Control Systems And A Neuroergonomic Approach To Measuring System Trust

With sufficient R&D investment the UK is forecasted to experience significant economic growth across a range of sectors by 2035 delivered through automation and robotics. Relevant application areas include the design and safety of future connected autonomous vehicles, remotely monitored teleoperations, and human-robot interactions involved in surgery, nursing, manufacturing, construction, and maintenance. Achieving sufficient user trust in automation is pivotal to realising this ambition in the UK. Measuring trust in automation is critical for determining automation acceptance and its correct usage. In addition, the act of deciding to trust is based on a mixture of analytic and emotional decisions. This complexity means that the widely used questionnaire-based methods of measuring trust are insufficient. Not only do they struggle to measure the full complexity of trust, but they cannot measure changes in trust as they occur in real-time in response to automation experience.

Measuring neural responses during interaction with automation is a potential objective means to measuring trust in real-time. Our own published research has advocated using functional near infrared spectroscopy (fNIRS) where areas of the brain associated with emotional trust judgements were identified. Having a real-time marker of emotional trust is important as it allows for a measure of trust that is uncoupled from analytical decision processes; processes that are involved in a range of cognitive processes, including mental workload. The primary goal of this proposal is to identify a unique neural measurement of automation trust. We will tackle this challenge through the in parallel measurement and analysis of neural correlates of trust (fNIRS) and physiological correlates of mental workload (heart rate variability, pupillometry, galvanic skin response) during experiments where participants interact with automated teammates of varying reliable.

To demonstrate the application of a unique neural marker of automation trust we will examine how trust changes in response to the communication method between humans and automation. Conventionally, responsibility between humans and automation is "traded" from one to another as a lumped whole. For instance, the way adaptive cruise control functions in modern cars. The driver can transfer whole responsibility to the car, typically initiated by a button press. Likewise, transfers are rapid and whole in automatic emergency braking, and initiated by the automation when an imminent collision is sensed. These transitions are often called "bumpy" and are implicated in compromises to safety. A promising alternative communication method is "haptic shared control". It offers greater transparency through the continuous force feedback communication of the automation's behaviour via the system's control input (e.g., steering wheel, accelerator pedal). This means that the user is better kept "in the loop", supporting "smooth" shifts of authority in response to automation-induced faults. However, no studies have been conducted providing a comparison of trust between traded and haptic shared control. Hence, the current proposal aims to provide not an only a demonstration of the application of a neural measure of automation trust, but also addresses the fundamental lack of knowledge surrounding haptic shared control and trust.

To realise this ambitious research, we will focus on initiating a collaborative academic relationship between Coventry University and TU Delft, respective world-experts in operator physiological monitoring and haptic shared control. Together we will establish neural markers of automation trust in a series of laboratory and aviation simulation experiments that involve performing collaborative tasks with automated teammates that will communicate with human participants in various ways - i.e. traded communication versus haptic shared control communication.