4 weeks, 2019
Many businesses are building and continuing to improve the reliability of their driverless technologies, but without building trust with users, these technologies cannot be adopted. We want to address the dissonance between safer technologies and people’s trust.
We leverage conversational user interfaces to build trust before the passenger even steps into the car. Check out our concept video below. (thank you Maggie for the touchups).
Our project began with two broad areas of exploration: driverless cars and conversational user interfaces. We kicked off with domain research to familiarize ourselves with these territories.
Companies all around are developing AV technologies, and for very different purposes. We looked at services from autonomous ridesharing to food delivery and on-the-road shopping to see where CUIs could provide value.
Our initial domain research led us to brainstorm 14 different scenarios in which a conversational agent could bring value to the area of driverless cars. We took this opportunity to think broadly, from moving conference rooms and personal tour guides to learning how to drive.
According to Eric Montague from Nuance Automotive, "Industry research provides strong anecdotal and scientific evidence that speech-induced anthropomorphism is essential in establishing trust. In other words, when a car talks it becomes more human-like to drivers and passengers."
Through our research, we identified trust as a significant issue in the adoption of driverless vehicles. Given the power of conversation in building trust, we wanted to focus on cultivating trust for a first-time, autonomous ride-sharing experience. Afterall, the first-time user is quite ubiquitous.
We wrote scripts to imagine the experience between a passenger and voice agent. These went along with more technical conversation models made up of utterances, intents, and responses that move across pre-attentive and attentive states. The storyline follows that of a passenger taking a ridesharing car to the airport. Our initial iterations explored greeting sessions, confirmation sessions, and hospitality sessions.
To do a first pass of our experience, we had three people act out the scenario using our script. This prototype was for us to look at the experience from a different perspective and to gain quick feedback on the CUI's role in building trust. We came out of this session with questions in areas such as privacy, personalization, and emotional support.
To dive deeper into forming the conversation and what trust means to people, we interviewed around 10 potential first-time riders, ranging from ages 20 to 63. Through these interviews, we gained an understanding of people's current interactions with driving and ridesharing. We asked them to imagine scenarios with autonomous vehicles in order to understand their attitudes and behaviors in this context.
In synthesizing our interviews, we identified major themes made up of different pain points. Addressing these pain points was an important part in building trust into the conversation.
We further looked into literature surrounding driverless technology and factors influencing adoption. According to the Journal of Engineering and Technology Management, "In understanding key factors in the adoption of driverless cars, two key bodies of literature have been drawn upon, namely technology adoption and driverless cars."
In a talk about how Uber builds trust in self-driving cars, Molly Nix identifies transparency, comfort, and control as important factors, which they work towards through elements like visualizing driving decisions in-vehicle and controlling the type of vehicle. But we wonder how trust can be built through out-of-vehicle contexts as well.
1. An ability to form and grow a relationship
2. An ability to maintain a relationship outside the vehicle
3. A more "integrated" way of providing service
Through our research, we identified the importance of building trust outside of the vehicle itself and the potential of a voice agent in addressing the issue, particularly in the ridesharing space. This helped to focus our problem.
We distill trust into 4 key sentiments: control, awareness, comfort, and safety, which we used to model the dialogue between passenger and voice agent.
For passengers, we are cultivating a sense of trust through control, awareness, comfort, and safety. In addition to the emotional benefits, the voice agent's integration across devices makes transportation more accessible and convenient.
For ridesharing companies, building trust with the technology is just as important as the reliability of the technology itself in order for the technology to be adopted and retained.
For other tech companies, a voice agent operating between and during commutes provides them opportunities to integrate their own services.
Here we began remodeling the conversation and quick sketching for a concept video. Our first exploration involved capturing joining snippets of different user types engaging with the onboarding experience. Through user feedback, we realized how it strayed from the value we proposed and the design principles we wanted to convey right off the bat.
Based on the principles of trust that we had extracted, we iteratively prototyped and tested our storyline and conversation model for AVA, our proposed voice agent.
The final concept video demonstrates interactions between AVA and a new autonomous rideshare passenger.
We recognize that AVA isn't the solution to trust in isolation. It falls into a bigger ecosystem, which includes policy, social factors, and infrastructure.
Ava is still an early-stage concept and in order to validate its success in building trust between passengers and driverless vehicles, we need to investigate more embedded contexts, with more in-depth quantitative and qualitative measurements. Future explorations may include the integration of other sensory cues (e.g. visual and haptic cues), creating greater discoverability in its functionalities, and leveraging VR technology.