Voice Assistant Accessibility for Deaf and Hard of Hearing Users
Given the rise in popularity of smart assistants (SA) and smart home speakers, and voice-based interfaces, more diverse user groups are adopting such technologies for everyday purposes. Previous work has highlighted the promise these devices can offer for users with low vision or mobility concerns to increase their independence in everyday tasks. However, their use by individuals with hearing concerns have been largely overlooked. The current use, opportunities, and challenges faced by deaf or hard of hearing (DHH) users have yet to be addressed, despite evidence that shows a growing interest in these technologies, especially in older adult populations and hearing aid users. To understand the use of SAs, we conducted in-depth interviews with deaf SA users and surveyed 73 individuals who had profound to mild hearing loss. Through this work, we document that despite inherent challenges with the current design of SAs, DHH users are highly motivated to use SAs for a wide range of use cases and believed that SAs held additional opportunities for them, if made more accessible, such as moderating workplace meetings.
Through this we have also developed design recommendations to better address the challenges faced by DHH users, without compromising the “embedded-in-life” experience users seek from voice interaction SAs. A leading challenge of smart assistants was the default characteristics of devices voices, which largely have “female” personas with a higher pitched voice that often fell outside of the hearing range accessible to many. This is especially problematic for individuals with age-related hearing loss, where higher frequencies are lost first. We argue that enabling users to easily tailor their assistant’s voice regarding pitch, volume, and speed, and allow for continued customization as their hearing needs change over time. Additionally, the visual feedback, such as light patterns, paired with assistant’s verbal responses are often too ambiguous to be consistently useful. Therefore, more research is needed in developing more intuitive visual feedback for these types of systems, which increase accessibility beyond the deaf and hard of hearing community. For our participants, receiving a voice response from smart assistants was not always the main goal, but rather using their voice to perform tasks, such as turning off lights or hands-free information search. However, smart assistants were less accurate in completing commands when users spoke without using heading aids or cochlear implants. Therefore, we urge future SA development to incorporate a wider range of voices in training data.
Moving forward, we are interested in exploring additional applications for SAs within this context, such as speech therapy, that may benefit other users and further justify the need for increased accessibility in SA devices.
Papers
- Johnna Blair and Saeed Abdullah. “It Didn’t Sound Good with My Cochlear Implants”: Understanding the Challenges of Using Smart Assistants for Deaf and Hard of Hearing Users. UbiComp ‘20.
- Johnna Blair and Saeed Abdullah. Understanding the needs and challenges of using conversational agents for deaf older adults. CSCW ‘19 Companion. (Poster)
People
Image by vectorjuice on Freepik