Lyft - OTTO

Researching for the Future of Autonomous:

Our research began with establishing a clear idea of what the future autonomous would look like and what problems currently existed with this future. Familiarizing ourselves with a variety of sources and reviewing each with a S.T.E.E.P. framework, the team established 3 major insights that would likely face Lyft’s autonomous ridesharing service.

1. Nervous and Scared

   • A public opinion of nervousness and fear towards autonomous vehicles

2. Poor Communication

   • Technology that doesn’t yet communicate driving style or intent

3. Lack of Trust

   • No trust for A.I. driver’s decision-making capability

Along with these problems our research allowed us to created a profile for our potential users. Because Lyft - OTTO would be a service for everyone, we delved into the broad category of urban commuters between 25-35. This large group along with their key needs helped us Identifying what our users would need most.

Design Principles

Initial Voice Interface Designs

We felt an immediate connection to the rounded angles across Lyft’s brand. In many ways, this shape represents the spirit of Lyft, symbolizing, harmonious, openness, and structure all at once. Modeling our VUI after them was a perfect choice for both brand and function.

While the form meshed with Lyft’s brand, we had concerns that its simplicity was short of the “person” feeling of OTTO’s brand. However, before we could fully address this, we needed to know how users would interact with the interface. Setting aside this potential issue, the team reviewed the most common way a user would need to with OTTO. This analysis showed smartphones could do the job, but testing also showed us that users didn’t want to be stuck staring at their phones. The solution was an update to Lyft’s iconic Amp.

Reinventing the Amp to display our virtual assistant allowed passengers to focus on a “face” during the ride and meant they could close their phones. This design showed promise, but we would need further iteration as the customer’s journey revealed itself.

Journey Map

To understand the potential journey of a customer, we generated feature lists from our research. From this collection organized and developed 7 categories of interactions, we would need to manage.

Based on comments from Professors Paul Pangaro and others, we took these features and into simple user flows. Focusing only on the elements that would be highlighted in the concept final video. To ensure to test and validate some of our early assumptions about these interactions, we relied on Wizard of Oz and other simulations created by our team. Through envisioning and acting out the experience with ourselves and others, we quickly generated a speculative journey.

This exercise surfaced many, as-yet-to-be-designed elements of OTTO, but the completed picture gave us a bird’s eye view of the comments working together, what states would be needed, and how where we still needed to work.

OTTO’s VUI Motion Design

With OTTO’s Amp as the display for our virtual assistant and a journey map, we had clarity around users would interact with OTTO. To manage all these behaviors we began to explore animations that would support the VUI’s need for clarity and sincerity. To begin, we established definitions and roles for each UI state would.

Once each state was well defined, we began expanding on our original circular shape.

From our initial work, the ring forms aligned best with our needs, allowing identity, personality, and clarity throughout the design of each state. However, even this one shape offered thousands of options for motion and color. Without more guidelines, we realized our animation could become disconnected or inconsistent with our communication needs. In order to iterate the design further without creating divergent patterns, we established universal rules for animations and established color references based on Lyft's brand.

With these rules applied we went through several iterations, eventually removing shadows and shifting in the animation’s subtly and speed.

The completion of these essential interactions allowed out voice assistants’ inputs and outputs direct links to the specific reactions and stimuli, giving OTTO a foundational non-verbal communication system. Our final consideration was the display’s visibility. Testing revealed back seat passengers would have a hard time seeing the OTTO AMP. Our solution was to move the Amp to the ceiling (replacing the mirror) and to design OTTO’s accompanying app to also display each reaction. With these final adjustments, we ensured accessibility communication between drivers and riders.

Designing Adaptive, Personal, and Authentic A.I.

To create each driver’s personality, we established baseline profiles from the BIG 5 personality matrix, from these each A.I. would be developed further from rider feedback and star ratings. Functioning similarly to Lyft’s system today, the least successful personalities would eventually improve or be removed based on the community preference.

While the A.I.’s personality would have a baseline and method of growth, we also wanted each driver to adapt for their familiar passengers. Just as a person might speak differently around familiar faces, OTTO would become familiar and more personable over time. So each personality would be a combination of the A.I.’s baseline personality, the context of the ride, and the user’s stated/unstated preferences. With such a mechanism in place, each A.I. driver could provide each customer with a bespoke experience.

The consideration for the UI display was influenced by today’s public trans. When we enter a ride-share or climb aboard a bus, there is a moment of connection and trust when you see the driver’s face. For OTTO to be trusted, we knew it needed to create a similar moment of trust. To recreate this trust, we relied on the phenomenon of people recognizing a smile, even though every face is different. We called this concept a Facet, a single of many, all belonging to the same object, and the design required each A.I. to display a unique icon but to always follow the same familiar pattern.

This shift to many facets also created a business advantage, by isolating the negative impact of a bad ride to a single driver rather than the entire Lyft brand. Most importantly, this update would keep the service anchored to a deeply authentic, adaptable, and personal ride experience, mirroring human ride-sharing today. OTTO’s Facets allows Lyft to maintain a trustworthy experience, even with fewer humans driving.

Finalize Mobile Demo

Throughout our building process, we knew that a mobile design would be key. Most notably, our mobile app would be the primary controller when voice commands weren’t practical, such as the start and end of rides. Keeping within Lyft’s brand, we added a new primary color and established “Lyft Pink” as the primary association with the A.I. driver.

StoryBoard & Script

Now that we had our largest pieces developed and built, we need to create a world in which all of them would interact. This meant further developing our user journey into scripted dialogs, interactions, storyboards. To begin with, each of us sketched out the most critical flows when using OTTO. The goal of these many flows was to evaluate the many ways a person might interact with OTTO even within our set parameters.

Our initial sketches helped get us in the right direction, but we needed a framework to analyze the interactions between the A.I. and human. We found this structure in an Input-Output map, allow us to ensure consistent interaction flows between the user, the interface, and even backend systems.

With this structure to reference, we wrote and developed a complete script of what each scene would need. Surfacing many of the low-fidelity sketches and dialogs to test and confirm.

Now that we had the raw material for what would become our concept video, we started iterating and gathering feedback around the best way to execute this story.

Prototyping & Testing the Experience

While we had tested our designs throughout the process, our biggest output would but a concept video. To create a video that would sufficiently communicate our ideas and outcomes we tested several prototypes around how to display our content.

Our first attempt was working within a very flat and comic-stip style narrative. The focus of this concept was showing the conversation happening between the users and the A.I., but still allowing the viewer to see the VUI’s language of motion.

Final Production

As we came closer to a finished concept, organization become a key component of our success. With testing finished and feedback gathered, we finalized scripts, voice recordings, images, and scene descriptions. To surface all this and ensure that the entire team member was on the same page, we collected and updated a large spreadsheet almost hourly.

As the story finalized, so did the characters involved. For the 6 scenes, we wanted to show we had 7 humans and 6 A.I. and to ensure that each scene felt real, A.I. and person had a different voice, personality, and mood.

5 Key Moments of Concept Video:

Full Concept Video:

Outcomes

Throughout the building of our concept video and the concepts of OTTO our team was able to push into the future and ask questions not about how the technology might make things easier, but most importantly how it could make A.I. more human. Our goal was to develop the speculative world where A.I. is fully present where trust is still held at the highest value.

While our team was fully committed to this idea, our logic arose not from our own fancies, but from observing the studies on the subject and by looking at the importance of trust in driverless ride-sharing. In this case, business and design outcomes both led us to the same solution, A.I. with a human touch.

Reflections

One of our biggest hurdles was balancing rendering a 3D city and the time it takes to output and edit such large files. While we made this concession due to COVID, the final quality of the video is not quite where we could want it to be. As a team, we learned the immense complications in balancing a 3D workflow alongside a 2D workflow.

However, as the primary animator, director, writer, editor, and product manager, I’m quite pleased with the novel idea we developed. As with any strong outcome, none of it would have been possible without a collaborative approach and effective communication throughout the process.