Design for Calibrated Trust For Acceptance of Autonomous Vehicles

Abstract

The advent of a society in which autonomous technology coexists with humans is an inevitability. The project focuses on one such autonomous technology in the form of autonomous vehicles or self driving cars. The benefits of such automation is well documented in academia and is supported by the investment by some of the biggest automobile and technology manufacturers in the research and development of autonomous vehicles. However, there exist certain challenges in realizing the full potential of autonomous vehicles. One such challenge is the attribute of trusting an autonomous vehicle. The project looks at the idea of trust in automation and dives deeper into the concept of calibrated trust as an approach to designing autonomous vehicles for increased acceptance of autonomous vehicles. The project is conducted in association with the Cities of Things Design Lab and People in Transit. Calibrated trust is defined as the balance between the capabilities of autonomous vehicles to the expectations of the end user. In essence it is the creation of an appropriate mental model by the end user. Through literature research and qualitative analysis, prominent challenges in achieving calibration were identified as: approach to designing for socio-technical systems, misalignment in communication between stakeholders, product branding and customer experience. Since, the focus of the project was towards the design and development team, the first two challenges i.e. approach to designing for socio-technical systems and misalignment in communication between stakeholders were selected to define the final design question and direction. The final design intervention is a Calibrated Trust Toolkit that can be used by development teams during the product development process to aid in designing for calibration of trust. It consists of four parts: A sensitizing session package, autonomous function visualization canvas, user decision matrix and trust enhancing communication. Collectively, the four parts allow for addressing the two challenges as selected previously. Each part of the toolkit was tested with designers and engineers and further iterated. The complete toolkit was validated by conducting interviews with experts and triangulating the data with the test data gathered during the testing phase of the design process. The testing and validation of the final outcome shows merit in the use of the toolkit for designing for calibration of trust and at the same time provides the flexibility for further modifications and adjustment. During the testing phase the participants found the use of the toolkit easy and intuitive. The digital method of testing suggested the deployment of the toolkit was possible in a digital setting. However, there were certain limitations to the project, the toolkit was not tested as a whole because of the time required and the unavailability of the necessary stakeholders. These limitations have been detailed out in the recommendation section of the report. Further research directions have also been suggested as a continuation of this project or start of new projects. In conclusion, the project is a step in the right direction when designing for calibrated trust by building on the work of other researchers like Ekman et al.(2016) and Mirning et al.(2016), but requires further research and design in other areas to fully realize the idea of designing for calibration of trust, such as the work of Anika Boelhouwer at TU Twente and David Abbink at TU Delft . In a broader perspective the insights and toolkit designed should not be limited to autonomous vehicles but extrapolated to designing other social robots or autonomous technologies that will coexist in future societies.