A Second Voice
Exploring design opportunities for IVAs to empower care workers and support the labor of caring
Home health aides are a vulnerable group of frontline caregivers who provide personal and medically-oriented care in patient’s homes. Their work is challenging and unpredictable as they adapt to patients’ changing needs. We investigated design opportunities for Interactive Voice Assistants (IVAs) to support aides’ essential care work and published our findings at the 2022 CHI Conference.
OVERVIEW
TIMELINE
February - July 2021
TEAM
6 Design Researcher
MY ROLE
Storyboarding
Conversation design
Video Creation
User Research/Interviews
Qualitative Data Analysis
RESEARCH FOCUS
Voice assistant, AI, intelligent agent, home health care, community health
BACKGROUND
This research project was done in collaboration with CMU CoALA lab and Cornell Tech.
THE CHALLENGE
Home health aides remains chronically undervalued
Home health aides are one of the fastest-growing sectors of the U.S. workforce. As more people age and seek support for aging in place, an increasing number of adults look to aides to support their health needs.
Despite aides’ growing importance in healthcare and patients’ increasing reliance on their services, aides are undervalued by the healthcare team and society at large and do not receive sufficient training or emotional support.
OUR EXPLORATION
How can IVAs support home health aides’ care work?
IVAs, voice-based software systems, enable hands-free interaction and can be embedded in laptops, smartphones, or smart speakers. Many studies have sought to develop IVA applications for healthcare contexts.
We explored a wide range of potential design opportunities for IVAs to support aides’ work, not limited by the current capabilities of commercially-available IVA systems.
OUR PROCESS
Probing possible futures via
Speed Dating
To probe the boundaries of what particular user populations will find acceptable and receive feedback at an early stage of ideation, we conducted 11 speed dating sessions with aides.
We presented our storyboards and asked open-ended questions to elicit participants’ reactions to a range of hypothetical usage scenarios and interactions.
BACKGROUND RESEARCH
We reviewed literature and articles on aides’ work and the struggles they face and identified areas of opportunities where IVA can support their labor of caring.
STORYBOARDING
Based on the background research, we generated multiple hypothetical scenarios in which an IVA might be of use to aides and sketched these scenarios as storyboards. We intentionally left the specific material embodiment of the IVA technology ambiguous to discourage participants from fixating on specific form factors or visual details.
IVA providing emotional support
Eliciting more ground suggestions via
Video Elicitation
To complement our speed dating sessions, we presented 15 participants (in 5 focus groups) with videos illustrating interactions with higher-fidelity IVA prototypes.
These prototypes were implemented using currently available IVA technologies, enabling us to also observe participants’ reactions to particular usage scenarios when the embodiment of the IVA was concrete and specific.
PROTOTYPING
The prototypes were developed iteratively, with conversation flows that were contextualized by real scenarios aides experienced or hypothetical scenarios in which the IVA might prove useful.
VIDEO CREATION
We wrote video scripts that staged the prototypes in their respective contexts alongside aides and filmed a video.
Data Analysis
We audio recorded and transcribed all speed dating and video elicitation sessions. We then analyzed transcripts from aide sessions and began with multiple passes reading these transcripts, allowing codes to emerge.
CODING
Each of the transcripts was then double-coded by two authors, with all coders meeting multiple times to discuss and reconcile their codes.
AFFINITY DIAGRAMMING
We also used affinity diagramming to facilitate code refinement discussions and iteration. Via this process, we converged on a stable codebook of 39 codes.
FINDINGS & IMPLICATIONS
The physical instantiation and location of IVAs in patients’ homes affected perceptions of IVAs’ potential role in aides’ work. IVA is viewed…
as an objective third party, whose presence would enable the IVA to speak up and vocally advocate for aides within the home environment.
as being able to improve transparency, making it more difficult for patients to hide their care plan and manipulate the aide into doing more work.
to be used collaboratively with their patient, strengthening the bond between patients and caregivers.
as a surveillance device (when using it as an emotional support tool) as they were afraid of being overheard by patients or being thought unprofessional for talking about their stress and/or challenges while in the patient’s home.
materiality, form factor, and location of IVAs
data collection and aide responsibilities
Aides see opportunities for IVAs to be used to advocate their work. Yet, patients would likely not want to be surveilled and reported by an IVA. In turn, agency staff suggested that IVAs might monitor aides and report whether or not they demonstrate competence in their work. Although aides discussed instances of how improved information and tracking of their work might be used to advocate for them to secure improved compensation or recognition, they certainly did not suggest that the IVA monitor their competence and report it to their supervisor. Instead, aides wanted control of what information would be tracked as well as when and how it would be used to advocate for them.
privacy, surveillance, and accountability
Without proper care and attention, introducing new technologies like IVAs into complex home healthcare contexts may end up further burdening an already vulnerable and marginalized workforce. Thus, rather than simply not designing technologies for aides, we instead recommend that IVA designers adopt social-justice-oriented design methods that center aides as key stakeholders when building voice-based technologies for deployment in home health care contexts.