Methods
Expert Interviews, Trainee Surveys, Co-design Workshop (Mind Mapping, Crazy 8s, Rapid Prototyping)
02
Empathizing With Coaches And Trainees
Expert Interviews and Trainee Surveys
Knowing that many apps exist out there and that future steps should be user-centered, I conducted expert interviews with 13 professionals who coach (physiotherapists, personal trainers, sports physicians, nutritionists) AND a survey with 70 prior trainees.
Takeaways From Survey and Interviews
Keep things simple
Dropout timeframe is 2-3 weeks
Coaches often act as a bridge for social contact
Positive and uplifting attitude is critical
Tailoring and other persuasive strategies are used by coaches
Trainees want to feel supported, heard, and accountable
Physiotherapists
"I use technology to train my clients, but tailoring workout plans to them can be limiting, which I believe affects rates of adherence and recovery"
Sport Physicians
"Coaching for athletes is easy because they just need fine adjustments, while beginners need the social support of a coach"
Personal Trainers
"Most participants drop out around the 2-3 mark. But what's interesting is that personal trainers also act as mental health consultants more often than not"
Trainees
"A coach is someone there to support me, but it really has to be a good fit otherwise it just makes me hate the activity"
Co-Design Workshop
Research has shown that digital interventions driven through theory are much more effective than through intuition. For this reason, I held a co-design workshop situated in the Health Belief Model to understand different perceptions of physical activity or inactivity.
Workshop Takeaways
Reduce barriers to physical activity
Ability to exercise from anywhere while still having the 'connection' to a coach - ie., a mobile web app
Creation of positive feedback loops for cues-to-action and self-efficacy
Tailored to beginner exercisers
Goals and progress should be visualized and achievable
Metaphor usage (given the proper cultural context)
Synthesized voice as a modality for future projections
Mind Mapping
For exploring different perceptions about barriers, benefits, severities, susceptibilities, demographics, cues to actions and self-efficacy to being physically active
Crazy 8s
8 minutes to quickly sketch out 8 ideas or 'solutions' to the problem at hand
Rapid Prototyping
Participants were split into 2 groups to create a digital prototype or physical prototype to address the problem. What was interesting was not the prototype itself, but rather their rationale for creating this prototype.
Methods
Summary of Qualitative Methods, Voice Synthesis Requirements, Persuasive Strategies
03
Formulation of Design Requirements
From the generative and exploratory research that was completed above, a set of design strategies were curated prior to the design of low and high fidelity prototypes. This ensured that the system effectively employed persuasive strategies, the AI-generated voice was humanistic, the exercises were beginner-friendly, and the system was usable and had a positive user experience.
1
Persuasive
Strategies
Tailoring
Tunneling
Positive Praise
2
Beginner
Exercises
Bodyweight
Min. 2x a week for 1 hour each session
Learnable
Stretches and strengthening movements
3
Interface &
Architecture
System tested for usability prior to 5-week field deployment
Audio tested for understandability, legibility, intelligibility, and likeability
Condition selection, onboarding, goal-setting dashboard, exercise library, main workout page, FAQs
4
Self
Reporting
UX, cognitive load, and intrinsic motivation questionnaires
Diary studies for understanding the 'why' beyond the 'what'
Methods
Visual Design, Usability Testing, Questionnaires (UXUX-lite, NASA-TLX, SUPR-Q, SRQ-E), Diary Studies
04
Prototyping, UI Curation & 5 Week Deployment for A/B Testing
An initial usability test was done to test the usability of the system and the synthesized voice. The screens were designed to educate participants while using the synthesized audio to keep them engaged and motivate them throughout the exercise journey.
As part of a Masters thesis, this also involved an experiment and therefore a questionnaire and diary study was included. While I created the UI from design to development, I also ensured the UX research was being done along the entire journey.
Methods
Thematic Analysis of Diary Entries, Statistical Analysis of Quantitative Metrics, Evaluation of KPIs
05
Results of Synthesized Voice Digital Coaching - Quantified
Rates of Exercise Adherence Over 5 Weeks
The adherence rate can be seen to be higher for the audio group compared to the visual group. What's more interesting is that this adherence rate pattern aligns with that seen for regular physiotherapy or personal training practice. This means that a digital audio coach was shown to be just as effective in getting people to stay to their exercise activity as an in-person coach.
AI-Voiced Training Results In Higher Adherence Rates, Okay Cool… But Why Is That? Here Are Some Individual Perceptions.
“The lack of communication through visual coaching caused me to lose motivation in the digital coach in this condition"
Visual Coaching Condition Participant
"Same as last week - I don’t think there’s enough ’coaching’. There are infinite workout routine videos on the internet, with more depth/detail."
Visual Coaching Condition Participant
"I really liked that it felt like I was exercising alongside someone the entire time. I know the voice was AI, but it was nice. I didn't feel so alone.
Audio Coaching Condition Participant
Other KPIs
Synthesized Speech As A Coach
Higher Usability of Synthesized Speech
Significantly Higher Trust
Exercise Perceived As Way Less Difficult
Behaviour Change KPIs
Participants noted they did not realize how out of shape they were
5 people made changes to their diets
1 person went for a routine medical checkup
06
Reflections & Business Applications
Modern voice user interfaces (VUIs) can be used to ease interactions with users or customers in a more social manner. Consider a conversational agent on your e-commerce site. Imagine how much more likely a customer is able to complete their goal, thereby increasing rates of trust and conversions with the added benefit of increased accessibility for visually impaired customers.
Business Applications
Most people enjoyed the exercise experience with the synthesized voice agent
Voices provide a sense of social support even without a trainer on site
The participants found the visuals very appealing and UI very engaging for the exercise context
Speech coaching showed a higher rate of adherence
What Went Well
Testing different voices across cultures
Exploring the application with individuals with ADHD
Tailoring alongside a physio or personal trainer within a full training setting
Weekly randomizations of the tunneling strategy
Multimodal application of synthesized voice and visuals for coaching (is a limitation since this was done for science and not a commercial approach)
What Could Be Done In The Future
Articoach
Articoach
Articoach
An end-to-end mobile exercise app for facilitating heart-healthy exercise as part of my Master's thesis. Using voice synthesis, this app was found to increase adherence to exercise by 26.7% through rigorous A/B testing, diary studies, and questionnaires.
Healthcare
Healthcare
UX / UI
UX / UI
Redesign
Conversational AI
UX Research
UX Research
Role
UX / UI Designer
UX Researcher
Frontend Developer
Team
Corey Tran
Tools
Figma
HTML
CSS
JavaScript
jQuery
Speech Synthesis
Research Methods
Interviews
Surveys
Competitor Analysis
Co-design Workshop
Usability Testing
UX Questionnaires
Diary Studies
Duration
May 2023 - Feb 2024 (9 mo.)
?!
Problem Space Within Cardiovascular Disease
We live in an age where health and exercise knowledge is accessible from the palm of one's hands. Yet, physical inactivity is one of the biggest causes of death in modern times. In 2019 alone, more than 32% of deaths were from cardiovascular disease according to the WHO - which begs the question:
Ludwig Boltzman Institute of Technology
Ludwig Boltzman Institute of Technology
"How can we get physically inactive individuals to perform heart-healthy physical activity to prevent the onset of cardiovascular diseases?"
00
Project Outcomes and Key Learnings In A Nutshell
Outcomes
Mobile web app that utilizes AI-synthesized audio as a form of coaching
End-to-end design and front-end development (from ideation, research, to interface and interaction design and testing)
AI voices were effective to facilitating exercise adherence and behaviour change
Points of Improvement
As the sole developer, I lacked knowledge in backend development and therefore did not design a login page (which would've made the personalization experience better)
Utilizing different AI voices as a design decision would present interesting opportunities to explore coaching voice effectiveness
:)
Data-Driven KPIs Achieved Through A/B Testing
Technical Metrics
26.7% Higher Adherence Rate
26.7% Higher Adherence Rate
Higher Usability of Synthesized Speech
Higher Usability of Synthesized Speech
Significantly Higher Trust
Significantly Higher Trust
Exercise Perceived As Way Less Difficult
Exercise Perceived As Way Less Difficult
Behaviour Change Metrics
Awareness of physical inactivity
Awareness of physical inactivity
5 people made changes to their diets
5 people made changes to their diets
1 person went for a routine medical checkup
1 person went for a routine medical checkup
Team
Corey Tran
Richard Parayno
Mats Reckszügel
Role
UX / UI Designer
UX Researcher
Frontend Developer
Tools
Figma
HTML
CSS
JavaScript
jQuery
Speech Synthesis
Duration
May 2023 - Feb 2024 (9 mo.)
Research Methods
Interviews
Surveys
Competitor Analysis
Co-design Workshop
Usability Testing
UX Questionnaires
Diary Studies
01
Market Research
Methods
Competitor Analysis
One glance in the app store and one will notice that there are many apps that provide exercise coaching, but most people don't stick to them. Why is that? As a first step to understanding this phenomenon, I looked at some current exercise and medication apps on the market to explore their core features and weaknesses. Meditation apps were explored due to their use of voice for coaching.
Takeaways From Market Research
1
Audio (real or synthesized) only as a form of feedback or short interactions such as telling you to "go" or "start"
2
Most apps relied on visuals for exercise interaction facilitation
3
Meditation apps employ social strategies in verbal coaching and cues (an opportunity for exercise coaching)
02
Empathizing With Coaches and Trainees
Methods
Expert Interviews, Trainee Surveys, Co-design Workshop (Mind Mapping, Crazy 8s, Rapid Prototyping)
Expert Interviews and Trainee Surveys
Knowing that many apps exist out there and that future steps should be user-centered, I conducted expert interviews with 13 professionals who coach (physiotherapists, personal trainers, sports physicians, nutritionists) AND a survey with 70 prior trainees.
Takeaways From Survey and Interviews
Takeaways From Survey and Interviews
Keep things simple
Dropout timeframe is 2-3 weeks
Coaches often act as a bridge for social contact
Positive and uplifting attitude is critical
Tailoring and other persuasive strategies are used by coaches
Trainees want to feel supported, heard, and accountable
Controls are important, but not always obvious
Specific use-case scenarios
Importance of sterility is context-dependent
Training require for device use
Smaller form factor beneficial
Mobile-first
Physician not necessarily the primary controller of device
Spatial arrangement of operating room is important
Speed of image production and retrieval is very critical
Physiotherapists
"I use technology to train my clients, but tailoring workout plans to them can be limiting, which I believe affects rates of adherence and recovery"
Sport Physicians
"Coaching for athletes is easy because they just need fine adjustments, while beginners need the social support of a coach"
Personal Trainers
"Most participants drop out around the 2-3 mark. But what's interesting is that personal trainers also act as mental health consultants more often than not"
Trainees
"A coach is someone there to support me, but it really has to be a good fit otherwise it just makes me hate the activity"
Co-Design Workshop
Research has shown that digital interventions driven through theory are much more effective than through intuition. For this reason, I held a co-design workshop situated in the Health Belief Model to understand different perceptions of physical activity or inactivity.
Workshop Takeaways
Reduce barriers to physical activity
Ability to exercise from anywhere while still having the 'connection' to a coach - ie., a mobile web app
Creation of positive feedback loops for cues-to-action and self-efficacy
Tailored to beginner exercisers
Goals and progress should be visualized and achievable
Metaphor usage (given the proper cultural context)
Synthesized voice as a modality for future projections
Mind Mapping
For exploring different perceptions about barriers, benefits, severities, susceptibilities, demographics, cues to actions and self-efficacy to being physically active
Crazy 8s
8 minutes to quickly sketch out 8 ideas or 'solutions' to the problem at hand
Rapid Prototyping
Participants were split into 2 groups to create a digital prototype or physical prototype to address the problem. What was interesting was not the prototype itself, but rather their rationale for creating this prototype.
03
Design Requirements, Deployment Strategy, Persuasive Strategies
Methods
Summary of Qualitative Methods, Voice Synthesis Requirements, Persuasive Strategies
From the generative and exploratory research that was completed above, a set of design strategies were curated prior to the design of low and high fidelity prototypes. This ensured that the system effectively employed persuasive strategies, the AI-generated voice was humanistic, the exercises were beginner-friendly, and the system was usable and had a positive user experience.
From the generative and exploratory research that was completed above, a set of design strategies were curated prior to the design of low and high fidelity prototypes. This ensured that the system effectively employed persuasive strategies, the AI-generated voice was humanistic, the exercises were beginner-friendly, and the system was usable and had a positive user experience.
1
Persuasive
Strategies
Persuasive
Strategies
Tailoring (curated for beginners)
Tunneling (exercises follow a set path)
Positive Praise (use of synthesize voice to praise participant after each exercise)
2
Beginner
Exercises
Exercise for
Beginners
Bodyweight
Min. 2x a week for 1 hour each session
Learnable
Stretches and strengthening movements
3
Interface &
Architecture
System tested for usability prior to 5-week field deployment
Audio tested for understandability, legibility, intelligibility, and likeability
Condition selection, onboarding, goal-setting dashboard, exercise library, main workout page, FAQs
4
Self
Reporting
UX, cognitive load, and intrinsic motivation questionnaires
Diary studies for understanding the 'why' beyond the 'what'
4
Self Reporting
UX, cognitive load, and intrinsic motivation questionnaires
Diary studies for understanding the 'why' beyond the 'what'
3
Interface and Architecture
System tested for usability prior to 5-week field deployment
Audio tested for understandability, legibility, intelligibility, and likeability
Condition selection, onboarding, goal-setting dashboard, exercise library, main workout page, FAQs
From the design research that was completed above, a set of design strategies were curated prior to the design of low and high fidelity prototypes. This ensured that the system effectively employed persuasive strategies, the AI-generated voice was humanistic, the exercises were beginner-friendly, and the system was usable and had a positive user experience.
1
Persuasive
Strategies
Tailoring (curated for beginners)
Tunneling (exercises follow a set path)
Positive Praise (use of synthesize voice to praise participant after each exercise)
2
Beginner
Exercises
Bodyweight
Min. 2x a week for 1 hour each session
Learnable
Stretches and strengthening movements
3
Interface &
Architecture
System tested for usability prior to 5-week field deployment
Audio tested for understandability, legibility, intelligibility, and likeability
Condition selection, onboarding, goal-setting dashboard, exercise library, main workout page, FAQs
4
Self
Reporting
UX, cognitive load, and intrinsic motivation questionnaires
Diary studies for understanding the 'why' beyond the 'what'
04
Prototyping, UI Curation & 5 Week Deployment for A/B Testing
Methods
Methods
Low-High Fidelity Prototype Iterations, Visual Design, Usability Testing, Questionnaires (UXUX-lite, NASA-TLX, SUPR-Q, SRQ-E), Diary Studies
As part of a Masters thesis, this also involved an experiment and therefore a questionnaire and diary study was included. While I created the UI from design to development, I also ensured the UX research was being done along the entire journey.
An initial usability test was done to test the usability of the system and the synthesized voice. The screens were designed to educate participants while using the synthesized audio to keep them engaged and motivate them throughout the exercise journey.
05
Results of Synthesized Voice Digital Coaching - Quantified
Methods
Thematic Analysis of Diary Entries, Statistical Analysis of Quantitative Metrics, Evaluation of KPIs
Rates of Exercise Adherence Over 5 Weeks
The adherence rate can be seen to be higher for the audio group compared to the visual group. What's more interesting is that this adherence rate pattern aligns with that seen for regular physiotherapy or personal training practice. This means that a digital audio coach was shown to be just as effective in getting people to stay to their exercise activity as an in-person coach.
AI-Voiced Training Results In Higher Adherence Rates, Okay Cool… But Why Is That? Here Are Some Individual Perceptions.
“The lack of communication through visual coaching caused me to lose motivation in the digital coach in this condition"
Visual Coaching Condition Participant
"Same as last week - I don’t think there’s enough ’coaching’. There are infinite workout routine videos on the internet, with more depth/detail."
Visual Coaching Condition Participant
"I really liked that it felt like I was exercising alongside someone the entire time. I know the voice was AI, but it was nice. I didn't feel so alone.
Audio Coaching Condition Participant
Other KPIs
Synthesized Speech As A Coach
Synthesized Speech As A Coach
Higher Usability of Synthesized Speech
Significantly Higher Trust
Exercise Perceived As Way Less Difficult
Behaviour Change KPIs
Behaviour Change KPIs
Participants noted they did not realize how out of shape they were
5 people made changes to their diets
1 person went for a routine medical checkup
01
Market Research
Methods
Competitor Analysis
One glance in the app store and one will notice that there are many apps that provide exercise coaching, but most people don't stick to them. Why is that? As a first step to understanding this phenomenon, I looked at some current exercise and medication apps on the market to explore their core features and weaknesses. Meditation apps were explored due to their use of voice for coaching.
Takeaways From Market Research
Audio (real or synthesized) only as a form of feedback or short interactions such as telling you to "go" or "start"
1
Most apps relied on visuals for exercise interaction facilitation
2
Meditation apps employ social strategies in verbal coaching and cues (an opportunity for exercise coaching)
3
06
Reflections and Business Applications
What Went Well
Most people enjoyed the exercise experience with the synthesized voice agent
Voices provide a sense of social support even without a trainer on site
The participants found the visuals very appealing and UI very engaging
Speech coaching caused a higher rate of adherence
What Would Be More Ideal
Testing different voices across cultures
Exploring the application with individuals with ADHD
Tailoring alongside a physio or personal trainer within a full training setting
Business Applications
Modern voice user interfaces (VUIs) can be used to ease interactions with users or customers in a more social manner. Consider that a conversational agent on your e-commerce site. Imagine how much more likely a customer is able to complete their goal, thereby increasing rates of trust and conversions with the added benefit of increased accessibility for visually impaired customers.
Articoach
An end-to-end mobile exercise app for facilitating heart-healthy exercise as part of my Master's thesis. Using voice synthesis, this app was found to increase adherence to exercise by 26.7% through rigorous A/B testing, diary studies, and questionnaires.
Healthcare
UX / UI
Redesign
UX Research
Role
UX / UI Designer
UX Researcher
Frontend Developer
Team
Corey Tran
Tools
Figma
HTML
CSS
JavaScript
jQuery
Speech Synthesis
Research Methods
Interviews
Surveys
Competitor Analysis
Co-design Workshop
Usability Testing
UX Questionnaires
Diary Studies
Duration
May 2023 - Feb 2024 (9 mo.)
?!
Problem Space Within Cardiovascular Disease
Ludwig Boltzman Institute of Technology
"How can we get physically inactive individuals to perform heart-healthy physical activity to prevent the onset of cardiovascular diseases?"
We live in an age where health and exercise knowledge is accessible from the palm of one's hands. Yet, physical inactivity is one of the biggest causes of death in modern times. In 2019 alone, more than 32% of deaths were from cardiovascular disease according to the WHO - which begs the question:
00
Project Outcomes and Key Learnings
Outcomes
mobile web app that utilizes AI-synthesized audio as a form of coaching
end-to-end design and front-end development (from ideation, research, to interface and interaction design and testing)
AI voices were effective to facilitating exercise adherence and behaviour change
Points of Improvement
As the sole developer, I lacked knowledge in backend development and therefore did not design a login page (which would've made the personalization experience better)
Utilizing different AI voices as a design decision would present interesting opportunities to explore coaching voice effectiveness
:)
Core KPIs Achieved Through A/B Testing
Technical Metrics
26.7% Higher Adherence Rate
Higher Usability of Synthesized Speech
Significantly Higher Trust
Exercise Perceived As Way Less Difficult
Behaviour Change Metrics
Awareness of physical inactivity
5 people made changes to their diets
1 person went for a routine medical checkup
More Case Studies
>
medPhoton
medPhoton
An x-ray imaging device controller (HMI) for improved usability, ergonomics, and mobility.
An x-ray imaging device controller (HMI) for improved usability, ergonomics, and mobility.
Read Case Study
Schnitzelberg
Schnitzelberg
A VR culinary experience 2nd place winner for the Tourism Technology Austria 2023 hackathon.
An x-ray imaging controller for improved usability, ergonomics, and mobility.
A VR culinary experience 2nd place winner for the Tourism Technology Austria 2023 hackathon.
Read Case Study