OVERVIEW
"A Second Voice" refers to the Interactive Voice Assistant in which we uncovered its opportunities and challenges on home health aides. At a higher level, we created and tested the end-to-end experiences of three features for aides (HHAs) to users and other stakeholders:
-
edema assessment
-
task reminder
-
self reflection
DOMAIN
Healthcare tech, conversational AI, user research
SKILLS
Literature reviews, storyboarding, prototyping (video elicitation), speed dating, focus groups, qualitative analysis
DURATION
February - November 2021 (10 months)
CITATION
[CHI 22] Vince Bartle, Janice Lyu, Freesoul El Shabazz-Thompson, Yunmin Oh, Angela Anqi Chen, Yu-Jan Chang, Kenneth Holstein, and Nicola Dell. “A Second Voice”: Investigating Opportunities and Challenges for Interactive Voice Assistants to Support Home Health Aides. To appear in Proceedings of the 2022 ACM CHI Conference on Human Factors in Computing Systems.
Background
Home health aides are formal, paid caregivers, primarily women of color, who work long hours caring for patients at their home. These caregivers' work is challenging because it involves both physical and emotional labor as they unpredictably adapt to patients’ changing needs and moods. Therefore, we partnered with Weill Cornell Medicine and introduced IVAs, voice-based software agents that allow hands-free and eyes-free interaction, which may make them well-suited to home health care contexts. We aimed to explore and evaluate design opportunities in IVAs to support aides’ life and work.
My responsibilities
Our team consists of 4 user researchers and 2 product designers, supervised by 2 faculty researchers (Cornell + CMU). I was responsible for 1 conversation design workflow and assisted with the research on home health aides. Each teammate contributed equally to every phase of the project.
Process
In this qualitative research study, we explored how IVAs could and might impact home caregivers and their work by designing storyboards, conversational flows and prototypes, and videos to gather qualitative data. We then conducted both speed dating and focus group sessions with aides, our end users, and agency reps to examine the impacts of adopting IVAs.
Literature reviews
We identified four design opportunity spaces in which IVAs could be introduced to aides and improve their work:
-
Assisting communication between aides and stakeholders (agency reps)
-
Protecting aides' labor rights
-
Providing emotional support to aides
-
Managing aides’ workloads
Storyboards
Storyboards were led by my designer teammates and they each contains a story context, problem, and corresponding solution. Prior to brainstorming, we gained guidance from the domain experts who firstly evaluated our proposed opportunity spaces above, and we ultimately iterated some scenarios, and decided to present eight of the storyboards that would run at later speed dating and focus groups sessions. See some examples below:
IVA assisting with communication with other stakeholders
IVA enabling aides to set up patients' preferences and then reminding on time
IVA tracking aides' tasks and recording the details that can later be discussed with agency lead
Prototypes
We transformed storyboard scenarios into conversational prototypes to better explain the workflows. I was responsible for the first prototype: Edema Assessment by creating a video elicitation partnered with a design researcher, and invited two Weill Cornell nurses as an aide and a patient. I also include another two prototypes that were developed by teammates.
Edema Assessment
Task Reminder
Self Reflection
Interviews
We recruited 32 participants in total, this including 6 agency representatives who supervise aides' work and 26 aide, our targeted users, through Weill Cornell Medicine networks and aide labor unions. We received agreement from each participants regarding video recording but due to information confidentiality, I will include contextual feedback below.
Speed dating
Each participant was presented with a number of hypothetical storyboard scenarios. In each round, we paired into two and asked open-ended questions to understand aides' desires and boundaries, how they potentially interact with an IVA in a home setting, as well as how aide agency reps think of the technology.
Focus groups
We organized focus group by inviting aides to form rich group discussions where we were able to collect their response and opinions.
Qualitative analysis
We transcribed all speed dating and focus group recordings into a spreadsheet. We independently reviewed all contexts and assigned two teammates to analyze with double-coded passes on each transcript, thereby generating our codebook of 39. During then, we maintained daily standup meetings to update progress and reconcile our codes with the whole team.
Excerpt of the interview transcripts with defining codes
We also facilitate affinity diagramming method to refine and iterate codes.
Affinity diagramming on Miro and generated 39 codes
Findings
We organized our findings as 4 primary themes from a series of qualitative analysis above:
Filling Gaps
designed to fill gaps in aides’ access to information and care coordination
Advocating
researched if IVAs can help aides in advocating or setting boundaries
Assisting in decision making
interviewed aides on how IVAs could help decide & complete tasks
Providing emotional support
evaluated the significance of IVAs serving as spaces to support aides' emotions
Implications
Our approach enabled us to explore a broad range of possible futures with participants, i.e. key implications for designers and developers interested in deploying IVAs in home health care and related settings.
key findings
01
Physical applications of an IVA in patients' home settings influenced perceptions of its potential responsibility in aides' work and service. Would this arouse concerns of stress or fear for aides if being overheard by patients regarding emotional support contexts?
02
The need to pay even close attention to the potential privacy and surveillance concerns/challenges that may arise due to the always-listening nature of IVAs. and possibly resulting in nuanced tensions and differing stakeholder priorities.
03
The challenge of centering aides in multi-stakeholder home health care contexts. Struggles exist as the focus tend to go back to patients (including their thoughts and reactions) though both stakeholders try to center and improve things on aides.
design implications and next steps
—>
Clarify the materiality (i.e. notion that the physical properties of an artifact have consequences for how the object is used), use cases, form factor and location of IVAs, because there's some ambiguity between them as a strict work tool vs. part of the personal device.
—>
Explore new mechanisms that allow both aides and patients to voice out when IVAs collect information and define the access (requires consent of all parties prior to being enabled). This might include examining the types of information that they can share and reach mutual agreement.
—>
Decide to prioritize on particular groups of stakeholders in a stage. This requires more exploratory research next by stating pros and cons, and pain points that IVAs (and other tech) bring to the complex systematic spaces before them being developed, without negatively affecting patient care.