University Mobility Challenge: Designing Wayfinding Technologies that Blind and Older Individuals Can Trust
Overview
For this multi-year research endeavor, I acted as the head Ph.D. student for the UCI side of the cross-coast, dual-university team. Between the University of Maryland Baltimore County (UMBC) and UCI students, the total team consisted of over a dozen researchers. The entire project, including multiple extensions to address specific issues raised by the research, lasted over two and a half years, during which time I participated in all activities, from design workshops to prototype brainstorming, and took the lead role in project management to make sure that all deliverables, including a multi-modal survey determining MVP and a final prototype demonstration pitch video, were presented to stakeholders such as our advisory board and Toyota's Director of Technology for Human Support.
Problem Statement
What technologies can be designed to help disabled and older individuals get around, particularly when they face challenges with indoor navigation in large locations like hospitals, airports, and hotels?
Methods
My first task, when added to this project, was taking an existing interview study completed by then-undergraduate Maya Gupta and translating the findings into usable personas. This involved the typical process of identifying key similarities across users and grouping real participants under fictionalized amalgams standing for the type of user they were. For this project we took it one step forward and ended up developing four axes or continuums that users’ preferences could be simplified to. Those four preference continuums allowed us to quickly communicate the specifics of those users, as well as theoretically explore potential user preferences that didn't show up in our study but might be represented in the larger community of users. The final personas are presented here, with the continuums labeled either "Preferences" or "Traits".
Following the interview study and the creation of personas, the next step was to gather as a group to complete two design workshops. Before each workshop, every team member was tasked with generating 100 ideas just to get the brainstorming process started. The workshop then consisted of sharing, categorizing, and generating more ideas before spending time creating low-fi prototypes and discussing how specific groups of ideas matched with the needs and preferences of user personas. After the workshop we compiled all the trace data from the experience, including sticky notes, photographs (see below), and digital notes documents into an executive summary of each workshop. These executive summaries allowed us to easily move the ideas from rough sketches and prototypes into more fleshed-out storyboards depicting the interactions we were envisioning.
After both design workshops were completed, our next step was to test prototypes with users. I led the study design and creation of an interview discussion guide for these participatory design sessions, while other members of the team made sure that the remote participants had all the physical supplies needed to try out the form factor of the prototypes. The way I set up and ran these sessions was unique in that, while we didn't have a Wizard-of-Oz style prototype to work with, considering our design was focused on a voice assistant, I had one of my colleagues act as the "ai assistant" during the interviews. This allowed users to actually try out how interactions might go, while my collaborator's knowledge of voice assistants meant that she could predict some of the potential issues or problems that a user might really encounter with a prototype such as this.
After we made form factor decisions based on participatory design, it was time to gather data on the basic features needed for a minimum viable product (MVP). To do this, I designed and ran a multi-modal survey with audio cues for respondents who then recorded their own audio responses. We had to use a survey system that was accessible to screen reader users, since they were a core user group we were targeting. I ended up designing the entire survey in Qualtrics to test the skip logic, question wording, etc., before finding that the plug-in to allow users to record audio had a bug. So, I recreated the survey on the Phonic.ai platform. We ended up getting 80+ respondents to give us examples of voice interactions they needed in this technology, as well as rating what from a set of features was most important to them.
Outcomes:
This project resulted in multiple research outputs, including two CHI papers discussing the methods and findings from different parts of the project (one in 2020 that I was a co-author on and another in 2021) and a Late Breaking Work that I first-authored for CHI 2020 looking at the particular issue of creating personas for people with disabilities and other marginalized identities. Additionally, we created and presented a final demonstration video of the prototype in action (below) to our community members, the Toyota business team, and even Toyota’s Director of Technology for Human Support.