Master's Thesis Case Study
The Adoption Gap: Why Older Adults Don't Keep Training?
ROLE
Lead UX Researcher
START / END
Sep 2022 – Jun 2024
SECTOR
Preventive Health · Aging · Behavior Change
Overview
OUTCOME
Preventive training device for dysphagia (swallowing difficulty), designed for adults 50+, targeting the window before symptom onset. The core challenge: what drives someone to keep doing a preventive exercise that doesn't yet feel necessary?
APPROACH
Used the system itself as a research instrument: a gamified, biofeedback-enabled CTAR device to study engagement, motivation, and long-term use intent. PLS-SEM analysis (n=30) combined with qualitative interviews.
PRESENTED AT
IEEE-ICASI 2025 conference; Journal of Design 2024 conference, Taiwan
The Challenge
Dysphagia affects nutrition, health, and independence in older adults. By the time symptoms appear, significant muscle decline has often already occurred. Preventive training exists. The problem is that it rarely sustains consistent engagement over time.
That raised the research question I kept coming back to: what actually drives an older adult to keep doing a preventive exercise that doesn't yet feel necessary?
I didn't want to answer that question from the outside. So I used the system itself as a research instrument: building a gamified, biofeedback-enabled CTAR device to create the conditions where I could study engagement, motivation, and long-term use intent in a realistic context.
My Contributions
I led the research and design end to end, across five phases:
Framing & Research Design
Defined the research question and theoretical framework (Flow Theory + the Person-Artifact-Task model). Conducted a literature review and speech therapist interviews to ground the design in clinical reality.
Measurement & Instrument Design
Built a theory-driven measurement model and developed a 27-item, 5-point Likert questionnaire to test it.
System & Interaction Design
Defined the training parameters and game design criteria: feedback logic, challenge progression, session structure. Set the performance thresholds the sensor needed to detect.
Data Collection
Recruited 30 participants independently through community centers. Administered the questionnaire post-use and conducted qualitative interviews to contextualize the findings.
Analysis & Synthesis
Ran PLS-SEM analysis in SmartPLS and synthesized quantitative and qualitative findings into design recommendations.
Team
I collaborated with a game design studio, which built the game to my specifications, information architecture, and logic flow across two rounds of iteration. I also worked with a mechanical engineering professor and two student engineers, who defined sensor specs and built the physical prototype to my system requirements. Together, we integrated the game with the sensor-equipped device into one working system and iterated on data calibration.
What I Built
The system has two integrated components.
Physical device
A modified Neckline Slimmer, a chin-resistance product originally designed for cosmetic use, which speech therapists later adopted as a CTAR training instrument for its feasibility and low cost. I worked with the engineering team to embed a custom sensor that captures chin tuck depth and resistance in real time, routing it through an Arduino to the game interface. I defined the performance thresholds through literature review and therapist interviews; the engineering team translated those into sensor specs, and together we calibrated the raw distance readings into four discrete ranges to make the data usable by the game logic.
Game interface
A desktop game where the training movement directly controls the character and the game provides immediate biofeedback. I defined the feedback logic, challenge progression, and session structure (grounded in Flow Theory and therapist interviews); the game studio built it to spec. The training exercise and the gameplay are the same action.
My Approach
I took a research-driven approach: ground the design in theory first, then test whether that rationale held with real users.
I chose Flow Theory because it offers a structural way to understand sustained engagement: challenge-skill balance, clear feedback, a sense of control. For preventive training, where the core challenge is sustaining motivation in the absence of symptoms, this felt like the right lens.
I paired it with the Person-Artifact-Task (PAT) model, a framework built specifically for digital interactive environments that accounts for how user characteristics, device properties, and task demands interact. That made it more precise than a general technology acceptance model for a system like this.
Gamification and serious games also have an established evidence base in rehabilitation, which gave me a grounded rationale for the approach rather than treating it as an untested assumption.
Key research activities
- Built a theory-driven measurement model proposing how motivation, self-efficacy, feedback quality, and challenge-skill balance shape attitude toward and continuance intention (the likelihood someone keeps using the system long-term).
- Developed and administered a 27-item questionnaire post-use to 30 participants. This captures self-reported perception and attitude, not observed behavior.
- Conducted qualitative interviews alongside the questionnaire to surface patterns the model alone couldn't capture.
- Applied PLS-SEM (structural equation modeling, via SmartPLS) rather than covariance-based SEM. PLS-SEM is built for smaller sample sizes like this study's n=30, while still letting me test how multiple latent variables jointly shape an outcome, rather than examining predictors in isolation.
- Recruited 30 participants independently through neighborhood community centers. Inclusion criteria: age 50+, no diagnosed swallowing difficulty (screened via the EAT-10, a standard 10-item clinical tool for swallowing-difficulty risk; score of 3+ excluded), no history of head or neck injury or surgery.
Key Insights
Going in, I expected gamification and engagement quality to be the primary drivers of adoption. The data confirmed experiential factors mattered, but what surfaced was a more complicated picture about whose model of motivation was actually embedded in the design.
1.
Enjoyment was necessary but not sufficient for adoption.
Participants found the biofeedback and gamification genuinely engaging, but engagement alone didn't translate to adoption intent. A recurring pattern: "I can see myself using it, but I'll only use it when I really need it." The system was competing not just on usability, but on a person's readiness to act preventively at all.
2.
Intrinsic motivation and health identity were the real entry points.
Participants who already maintained proactive health habits showed meaningfully higher interest: "I've been massaging my chin muscles before eating nowadays. Where can I get this system?" Adoption intent was shaped less by in-the-moment motivation and more by whether someone already saw themselves as a person who takes preventive action.
3.
Medical professionals are the trust gatekeepers.
Users consistently said they'd adopt the system if a doctor or therapist recommended it. This wasn't skepticism toward the device — it was how this population makes health decisions. The adoption pathway needs to run through clinical relationships, not just consumer appeal.
4.
Setup complexity competed with the actual goal.
I had assumed engagement was the goal. For most participants, the goal was effortless training that fit an existing routine. Even participants who enjoyed the experience flagged setup as a barrier: "I might need a tech person or my kids to help me." One suggested a simpler mechanical device used while watching TV; another said content matched to personal interests, like classical music, would feel more sustainable. Setup became its own cognitive task, competing with the training goal instead of supporting it.
5.
The design encoded my values, not the user's.
The system was informed by speech therapist interviews, but validated with healthy adults 50+, not people currently experiencing swallowing difficulty. The clinician perspective shaped what I built; user testing revealed what it was actually like to use. The motivational logic in the design reflected my assumptions more than users' realities.
Reflections & Next Steps
The gap between design intent and user reality wasn't a failure — it was diagnostic. It clarified the actual design problem: not building a better training device, but designing for the full adoption journey and lifestyle fit.
If I continued this work, I would:
- Run longitudinal testing over weeks or months, and triangulate the questionnaire's attitudinal data against behavioral data — the gameplay and biofeedback logs the system already captures — rather than relying on self-reported intent alone. Intent and long-term adherence are different things, and behavioral data would close that gap.
- Include people currently experiencing early-stage swallowing difficulty alongside healthy preventive users, to close the gap between the consulted population (clinicians) and the tested population (healthy adults) that limited this study.
- Bring medical professionals into the design process earlier, treating referral and trust as design requirements — not external factors — and building the clinical handoff into the system architecture from the start.
- Explore a business model shift: distribute through hospitals and therapy clinics on a rental basis instead of direct-to-consumer, with structured onboarding and peer or caregiver support built into that channel rather than left to the user. This would lower cost and setup burden, route the system through the medical-authority trust this population said they needed, and shift onboarding and tech support onto clinical staff already equipped to provide it.
Impact
- Presented at IEEE-ICASI 2025 and at the Journal of Design 2024 conference, Taiwan
- Received an “Excellent” rating in a government-funded research project
- Found that device confidence mattered less for adoption than existing health identity — an explanation the SEM data alone couldn't have provided
Why This Matters
This project is where I learned what it actually means to mix methods. The SEM told me which variables mattered, and which didn't: intrinsic motivation was the strongest predictor of long-term use, while self-efficacy — which I expected to be a strong driver — came out weak and negative. On its own, that's just a number. The interviews told me why: for this population, self-efficacy wasn't about confidence in the device. It was about whether someone already saw themselves as a person who takes preventive action at all. That gap between what the model predicted and what the interviews explained is what shaped the design guidelines I propose in Reflections & Next Steps, away from interface polish and toward the trust infrastructure and service context surrounding the device.
A well-researched, theoretically grounded design can still encode the researcher's assumptions rather than the user's reality. Rigor isn't a guarantee of fit. Getting to fit means staying genuinely curious about the gap between the model and the person, and treating that gap as the most important data.