OVERVIEW

The future of transportation is moving towards the autonomous… Vehicles are incorporating new forms of automation every year. While this technology continues to develop, people still have questions about the viability of autonomous vehicles in the day-to-day. Are they safe? Can passengers trust that the system knows what it’s doing? Are there opportunities for humans to intervene if something goes wrong?

For our Master’s capstone project, my team explored these questions with our clients – Honda’s 99P Labs, a digital proving ground for mobility and energy innovators. After 8 months of foundational research, iterative design, and evaluative testing, we created Orbit, a multi-modal “ecosystem of attendants” to increase a commuter’s sense of safety and confidence in autonomous public transit.

TIMELINE

8 months
Jan – Aug 2022

MY ROLE

Product Designer
Interaction Designer
Visual Design
Motion Designer
UX Researcher

TEAM

Andrea Zhu
Eleanor Hofstedt
Healy Dwyer
Kristian Pham

SKILLS

Rapid Prototyping
Concept Testing
Generative + Evaluative Research
Motion Design

THE PROBLEM

How will passenger behaviors change when a vehicle's attendant is telepresent or a synthetic representation?

As autonomous technology becomes more mainstream, passenger behaviors and mentals models will be forced to shift and adapt. While much of the current research has focused on private automobiles, a driverless future is primed to have an impact on shared transit as well.

So how might autonomous shared transit vehicles incorporate technology that provides passengers with a sense of safety and confidence as they embark on typical journeys in this new, uncharted setting where they are not only without a driver, but also sharing the vehicle with people they may not know?

THE SOLUTION

Orbit addresses the needs of a broad and variable passenger population when there is no operator in a shared vehicle

Meet Orbit: a multimodal system that communicates information across built-in screens, passengers’ personal devices, and a conversation-based user interface, with the additional layer of live agent support available via telepresence.

This system considers the current behaviors and mental models of passengers in shared transit environments – whether those are buses, trains, or rideshare services – and provides consistent, discreet, and accessible support to ensure passenger safety and confidence in autonomous shared transit.

On a passenger’s journey, there are various touch points that they interact with using Orbit…

Hailing a ride with Orbit

Our generative research suggested that, in public transportation, the presence of other passengers has a huge impact on one’s riding experience. We want to ensure everyone is in the best seat.

Receiving discreet information

Our research revealed that no matter where the passenger is trying to go, or how familiar they are with their destination, they want discreet information. The Orbit notification system does just that!

Getting live support help

While our ecosystem of technology was able to meet most passenger needs, we found that there is a threshold where live, human support is desired.

Let’s see how all of these touchpoints play out using the system…

THE PROCESS

Understanding the rules of the road...

99P Labs came to us with the question, “How will passenger behaviors change when an attendant on a public people mover is available via telepresence or a synthetic representation?” The team quickly jumped into research on how people interact with robotics, artificial intelligence, and telepresence to understand how passengers might interact with various attendant types.

However, to get to the root of the problem, we needed to build an understanding of the current public transportation experience and understand the human side of the problem. The team set out to answer a new question: when and why do passengers currently seek assistance on public transportation?

To answer this question and create a solution in the 8 month span of the project, I did project management, contributed to generative and evaluative research throughout, and led the project’s UX and UI designs.

Exploring today's transportation journey with generative research

To understand current mental models of people who take public transportation the team completed contextual inquiries with passengers on city buses, facilitated a five-day diary study with people who regularly take public transit, and led semi-structured interviews of participants with a variety of backgrounds and needs.

Additionally, secondary research encompassing 25 academic papers and 13 expert interviews, helped the team build a collective foundation of how humans' relationship with technology and AI has evolved in recent years, and the principles that lead to positive human-AI and robotic interaction.The key findings from this process highlighted the need for clear expectation setting and personalization in AI systems. Human-like qualities in AI can lead to higher expectations and trust is often broken when a system doesn't live up to them.

Understanding the holistic passenger journey from Point A to Point B

The team rode public transit with 5 Pittsburgh residents as they traveled to routine destinations throughout the week and conducted semi-structured interviews with 11 participants nationwide to describe their experiences with public transportation. From this research, we were able to better understand users' attitudes around transit, uncover user needs, identify scenarios in which users interact with some form of attendant while in transit, observe behaviors and interactions in both routine and atypical moments, and understand the why behind how passengers accomplish their transit goals in the context where they occur.

Observing the authentic commute

To mitigate actor-observer bias of team moderated methods, we conducted diary studies with 14 participants across the U.S. from a variety of backgrounds across geographic location, profession, race and ethnic background, income level, age, and transportation method.

Using the dscout platform, we ran a 5 day diary study in which participants told us about their experiences in public transit. This research helped us identify routine versus unexpected moments and how those impact passenger needs, expectations, and behaviors; mitigate actor-observer bias of team moderated methods.

After the first four months of research, the team arrived at these 4 areas of opportunity...

Familiarity

A passenger 's familiarity with both their route and the system influences their riding experience

Modality

Passengers will access information from our system through many different modalities - whether that’s built-in screens or personal devices

Discretion

Our system needs to give feedback at the appropriate moment, but also needs to be discreet enough so as not to be intrusive during the journey

Inter-Passenger Dynamics

Other passengers have a significant impact on a rider’s journey / experience. Our system should consider the presence of other passengers during a journey and adjust its capabilities accordingly.

EVALUATIVE TESTING

Prototyping! Prototyping! More Prototyping!

With the aforementioned themes in mind, our team built and tested prototypes at increasing levels of fidelity, which ultimately informed today's final version of the product, Orbit. Design explorations started with the conceptual: early concept tests, design thinking activities, and insight-based storyboards. By the final stages of the project, the team was conducting research in moving vehicles with multiple passengers riding along and utilizing Orbit to meet their needs.

Feedback and reactions from users across demographic and geographic user groups helped clarify our vision of the ideal form of a system, along with users’ goals and expectations in transit scenarios – in short to be comfortable, safe, and confident that they'll arrive at their destination.

PROTOTYPE 1

Understanding physical layout and passengers’ preferences for seating arrangements

To what degree should this seating arrangement be movable and customizable, versus fixed? What is the optimal seating arrangement for traveler safety and comfort? How is this preference impacted by the presence of other passengers, both known and unknown? These are some of the questions we first sought to understand as we approached our first round of physical prototyping. For this round, we conducted intercept interviews in a park with 4 chairs and asked participants questions regarding their preferences for seating in the relevant context.

What we found:

In an autonomous vehicle, passengers want to face forward or have a view of any on-board technology or windows
With longer trip length comes an increased desire for customization.
Customization should happen pre-trip to avoid awkward inter-passenger dynamics when moving seats.
Motion sickness plays a critical role in participant’s decisions in seating orientation.

PROTOTYPE 2

The different components of our multi-modal attendant ecosystem

Next, we wanted to understand how participants navigate available modalities of information in unexpected scenarios, especially when there is not a human present to help them. Additionally, we wanted to pinpoint gaps in information that may arise between modalities, such as screen-based information versus audio information.
‍
What we found:

Passengers want multimodal reinforcement of messaging, especially given that they might be using their own personal devices during the ride and might miss a tablet-based notification
The level of detail of information should strike the balance between instilling trust and being confusing or overwhelming.
The level of risk and urgency influences passengers’ modality preferences. Passengers indicated a preference for an audio modality if the situation was higher risk and they needed to pay attention.
Passengers want to provide feedback to the system or a live agent through an effective communication channel.

PROTOTYPE 3

Information clarity and user navigation in a screen-based interface

For this round of usability testing, we sought to understand if the system allows a passenger to successfully complete tasks that may arise specifically in AV situations, such as the vehicle coming to an unexpected stop. Additionally, we aimed to evaluate user preferences and comprehension of in-trip notifications that are designed to be both effective and unobtrusive / discreet, and map out the built-in screen, personal device, and audio-based ecosystem and where certain modalities are leveraged over others
‍
What we found:

Our design should utilize notifications sparingly and don’t require acknowledgment unless critical. Notifications require too many ‘confirmations’ or ‘dismissals’ from users, which adds friction to the experience
Newer interaction styles such as a “long hold” or voice based interactions need to onboard users in order to build familiarity for those with less tech experience
Time is of the essence; user confidence decreases when ETAs are unclear. When prioritizing what information to display, it is more important to show users information on their trip duration and time to destination than what is going on behind the curtain of the vehicle itself

FINAL ITERATION

Simulating the Ultimate AV Experience

To fully test Orbit, we had participants ride in a vehicle and use a built-in screen based prototype to complete certain tasks in a given scenario. Although team members are driving the vehicle, there is a barrier between the driver and the passenger in order to simulate a driverless vehicle. A remote researcher is dialed into the moving vehicle and prompting the passenger throughout their journey.

From this final round of testing, we aimed to learn if the amount and specificity of information that we give passengers via our system instills a sense of safety and confidence during their riding experience and measure the thresholds at which passengers want to interact with a human to get information, rather than an AI-based modality.

Ultimately, passengers trust that the system knows enough information …

Passengers indicated that they trust the vehicle’s system to know enough information to successfully navigate any scenario.
Given that potential scenarios are limited to the car and the road, passengers assume the computer knows what is going on. They also feel confident that the system will let them know if there is information they need in certain situations.

If we were to have more time developing Orbit, some things our team would love to explore would be…

Creating logic around conversation-based agents, and exploring the tone and voice of a CUI and how that ultimately impacts trust. Thinking about conversational interfaces with accessibility in mind – we would also like to develop full functionality through CUI for passengers who want to be hands-free or not rely on sight, while also giving passengers control over their physical environment.

Adapting the experience for frequent and familiar passengers – changing notification frequency and information based on level of familiarity with the system using AI and machine learning to recognize when passengers need more or less support.

REFLECTION

I learned that, ultimately, designing for trust hinges upon designing for value. As I explored designing for autonomous spaces, I found again and again that user feedback revolved around a lack of trust and security towards this emerging technology. What I repeatedly found through research and prototyping was that the first step towards trust was ease of use and accessibility of technology. As I continue designing, I hope to keep in mind how critical convenience is towards tech adoption, and how we can ethically make design streamlined for the user.

Additionally, this capstone project taught me a lot about how to collaborate and work alongside multiple stakeholders and team mates. I learned so much about conducting research, synthesizing insights, rapid prototyping, and articulating and advocating for design decisions. All skills that I hope to take with me in my next professional endeavors!

Orbit

Reimagining the future of transportation with a consistent, discreet, and accessible integrated system.