Explainable AI
Healthcare AI
Hi-Fi Prototyping
Chatbot Interface Design
Black Box Mitigation
KLARIA APP, UX STRATEGY AND PROTOTYPE FOR TRANSPARENT AI
KLARIA APP, UX STRATEGY AND PROTOTYPE FOR TRANSPARENT AI
This overview summarizes a comprehensive product development lifecycle bridging foundational user research and high fidelity UI design. By conducting competitive analysis and developing dynamic interactive prototypes this project translates complex user psychology into a scalable design guide. The end result is a fully transparent digital health product leveraging explainable artificial intelligence, explicit role framing and prominent visual safety signaling to drive genuine user reliance across the entire platform.
Year :
2026
Industry :
Digital Health and HealthTech
Role :
Product Designer and UX Researcher
Project Duration :
9 months
↑0.00%
↑0.00%
Integrating explicit visual safety badges and privacy signaling, drastically elevated user confidence. The experimental group recorded a 5.94 mean trust score compared to just 2.56 for the neutral interface.
Integrating explicit visual safety badges and privacy signaling, drastically elevated user confidence. The experimental group recorded a 5.94 mean trust score compared to just 2.56 for the neutral interface.
Integrating explicit visual safety badges and privacy signaling, drastically elevated user confidence. The experimental group recorded a 5.94 mean trust score compared to just 2.56 for the neutral interface.
↑0%
↑0%
By replacing opaque algorithmic outputs with transparent explainable AI and personalized AR scans the in depth prototype achieved a peak trust score of 6.1 out of 7. This represents a 28% surge in patient confidence compared to the baseline no explanation model.
By replacing opaque algorithmic outputs with transparent explainable AI and personalized AR scans the in depth prototype achieved a peak trust score of 6.1 out of 7. This represents a 28% surge in patient confidence compared to the baseline no explanation model.
By replacing opaque algorithmic outputs with transparent explainable AI and personalized AR scans the in depth prototype achieved a peak trust score of 6.1 out of 7. This represents a 28% surge in patient confidence compared to the baseline no explanation model.
↑0.00%
↑0.00%
By explicitly defining the operational limitations and framing the application strictly as a supportive interim guide the system successfully calibrated patient expectations. This transparent boundary setting completely resolved role ambiguity and elevated the mean trust score to 5.45 out of 7 .
By explicitly defining the operational limitations and framing the application strictly as a supportive interim guide the system successfully calibrated patient expectations. This transparent boundary setting completely resolved role ambiguity and elevated the mean trust score to 5.45 out of 7 .
By explicitly defining the operational limitations and framing the application strictly as a supportive interim guide the system successfully calibrated patient expectations. This transparent boundary setting completely resolved role ambiguity and elevated the mean trust score to 5.45 out of 7 .
My role :
In my dual role as a Product Designer and UX Researcher I directed both the theoretical research and the practical design execution of Klaria. From conducting initial market evaluations to simulating live conversational dynamics through the Wizard of Oz method and running extensive quantitative vignette tests I ensured every interface decision was strictly backed by empirical user data.
CONTEXT :
This comprehensive initiative bridges practical product development with rigorous academic research. Beginning as a foundational competitive analysis of the digital health market it evolved into a fully interactive prototype test and culminated in a large scale quantitative study. The project establishes a complete and scalable design guide for building user reliance in artificial intelligence interfaces.
THE CORE CHALLENGE :
Millions of patients rely on digital symptom checkers to manage their health anxiety but these platforms frequently operate as completely opaque black boxes. They deliver high stakes medical advice without explaining their algorithmic logic or citing verifiable sources. This lack of transparency combined with a failure to explicitly define operational boundaries creates a severe trust deficit causing users to either blindly accept unverified outputs or completely reject the technology out of profound fear.
Methods :
• Competitive and Market Analysis
• User Persona and Job Story Development
• Dynamic Prototyping and Interface Design
• Wizard of Oz Simulation Testing
• Qualitative Thematic Analysis
• Quantitative Vignette Testing
Uncovering the Friction Points :
Through the Wizard of Oz simulation and thematic analysis I identified the exact cognitive barriers preventing everyday users from fully trusting the diagnostic artificial intelligence. The qualitative data revealed that pure algorithmic accuracy is meaningless if the interface fails to immediately communicate transparent reasoning, explicit operational boundaries and structural safety.
Journey 01: THE DIAGNOSTIC ASSESSMENT
Journey 01: THE DIAGNOSTIC ASSESSMENT
When patients initially interact with the symptom checker they are highly vulnerable and actively searching for clear structural assurance. The qualitative interviews demonstrated that this initial triage phase is where trust is most frequently broken because the system operates as an opaque black box.
When patients initially interact with the symptom checker they are highly vulnerable and actively searching for clear structural assurance. The qualitative interviews demonstrated that this initial triage phase is where trust is most frequently broken because the system operates as an opaque black box.
Friction Point A: UNSUBSTANTIATED ALGORITHMIC CLAIMS
Friction Point A: UNSUBSTANTIATED ALGORITHMIC CLAIMS
When moving away from a completely opaque black box system the initial prototype attempted to provide users with direct algorithmic reasoning. However this transition exposed a significant roadblock regarding how patients evaluate the credibility of digital medical claims and diagnostic outputs.
When moving away from a completely opaque black box system the initial prototype attempted to provide users with direct algorithmic reasoning. However this transition exposed a significant roadblock regarding how patients evaluate the credibility of digital medical claims and diagnostic outputs.
• The insight :
• The insight :
Through the Wizard of Oz simulation the data revealed a critical trust anomaly where providing a moderate explanation without tangible proof actually decreased overall user confidence. Providing users with a numerical likelihood score without offering a verifiable external source actively damaged system integrity. Participants explicitly stated that making specific medical claims without linking to authoritative guidelines made the application feel highly suspicious.
Through the Wizard of Oz simulation the data revealed a critical trust anomaly where providing a moderate explanation without tangible proof actually decreased overall user confidence. Providing users with a numerical likelihood score without offering a verifiable external source actively damaged system integrity. Participants explicitly stated that making specific medical claims without linking to authoritative guidelines made the application feel highly suspicious.
• The SOLUTION :
• The SOLUTION :
To resolve this trust deficit I designed the interface to satisfy a strict hierarchy of proof. I integrated explicit external validation by anchoring every diagnostic claim directly to authoritative medical guidelines such as the World Health Organization. This strategic design decision transformed the opaque algorithmic output into a transparent and verifiable claim which successfully increased the average user trust score to 6.1 out of 7.
To resolve this trust deficit I designed the interface to satisfy a strict hierarchy of proof. I integrated explicit external validation by anchoring every diagnostic claim directly to authoritative medical guidelines such as the World Health Organization. This strategic design decision transformed the opaque algorithmic output into a transparent and verifiable claim which successfully increased the average user trust score to 6.1 out of 7.
JOURNEY 02: THE ONBOARDING EXPERIENCE
JOURNEY 02: THE ONBOARDING EXPERIENCE
Before users even begin typing their symptoms they form critical expectations about the capabilities of the application. The qualitative research indicated that this initial entry phase is essential for establishing strict operational boundaries and preventing dangerous misinterpretations of the artificial intelligence.
Before users even begin typing their symptoms they form critical expectations about the capabilities of the application. The qualitative research indicated that this initial entry phase is essential for establishing strict operational boundaries and preventing dangerous misinterpretations of the artificial intelligence.
Friction Point B: SYSTEM ROLE AMBIGUITY
Friction Point B: SYSTEM ROLE AMBIGUITY
When digital symptom checkers fail to define their operational limitations users frequently misinterpret them as definitive medical authorities. This ambiguity leads to either dangerous algorithmic over reliance or complete system rejection disrupting the entire medical triage process.
When digital symptom checkers fail to define their operational limitations users frequently misinterpret them as definitive medical authorities. This ambiguity leads to either dangerous algorithmic over reliance or complete system rejection disrupting the entire medical triage process.
• The insight :
• The insight :
Qualitative interviews revealed that users actually view the application as a crucial tool for managing health concerns during the interim period before a formal doctor visit. However quantitative testing demonstrated that when the interface utilized authoritative framing without explicitly stating its limitations the average user trust score dropped to 2.37 out of 7 during the clinical evaluations.
Qualitative interviews revealed that users actually view the application as a crucial tool for managing health concerns during the interim period before a formal doctor visit. However quantitative testing demonstrated that when the interface utilized authoritative framing without explicitly stating its limitations the average user trust score dropped to 2.37 out of 7 during the clinical evaluations.
• The SOLUTION :
• The SOLUTION :
To properly calibrate expectations I designed clear onboarding disclaimers that explicitly framed the application as a supportive interim guide rather than a definitive diagnostic tool. This transparent boundary setting completely resolved the role ambiguity and successfully elevated the mean trust score to 5.45 out of 7 during the quantitative testing phase.
To properly calibrate expectations I designed clear onboarding disclaimers that explicitly framed the application as a supportive interim guide rather than a definitive diagnostic tool. This transparent boundary setting completely resolved the role ambiguity and successfully elevated the mean trust score to 5.45 out of 7 during the quantitative testing phase.
Journey 03: THE CONTEXTUAL ENTRY PHASE
Journey 03: THE CONTEXTUAL ENTRY PHASE
As patients evaluate a new digital health platform they carry an extreme baseline of privacy concern regarding their highly sensitive medical data. If the application fails to immediately provide absolute proof of structural safety users will instinctively abandon the platform before the diagnostic assessment even begins.
As patients evaluate a new digital health platform they carry an extreme baseline of privacy concern regarding their highly sensitive medical data. If the application fails to immediately provide absolute proof of structural safety users will instinctively abandon the platform before the diagnostic assessment even begins.
Friction Point C: LACK OF STRUCTURAL ASSURANCE
Friction Point C: LACK OF STRUCTURAL ASSURANCE
When symptom checkers operate without explicit visual security measures patients are forced to assume the highest level of personal risk. This lack of environmental safety creates a massive psychological barrier preventing vulnerable users from confidently disclosing their personal health information.
When symptom checkers operate without explicit visual security measures patients are forced to assume the highest level of personal risk. This lack of environmental safety creates a massive psychological barrier preventing vulnerable users from confidently disclosing their personal health information.
• The insight :
• The insight :
Quantitative testing revealed that forcing users to rely on an unverified environment severely damages overall system confidence. When the interface lacked explicit safety signaling the mean user trust score dropped to a critically low 2.56 out of 7 during the clinical evaluations. The research proved that participants require immediate visual proof of regulatory compliance to feel genuinely secure.
Quantitative testing revealed that forcing users to rely on an unverified environment severely damages overall system confidence. When the interface lacked explicit safety signaling the mean user trust score dropped to a critically low 2.56 out of 7 during the clinical evaluations. The research proved that participants require immediate visual proof of regulatory compliance to feel genuinely secure.
• The SOLUTION :
• The SOLUTION :
To resolve this intense perceived risk, I integrated prominent visual safety badges and medical device certifications directly into the user interface. This strategic addition of structural assurance transformed abstract legal jargon into a tangible heuristic for safety which successfully elevated the mean trust score to an outstanding 5.94 out of 7.
To resolve this intense perceived risk, I integrated prominent visual safety badges and medical device certifications directly into the user interface. This strategic addition of structural assurance transformed abstract legal jargon into a tangible heuristic for safety which successfully elevated the mean trust score to an outstanding 5.94 out of 7.
BEYOND STATIC INTERFACES: BUILDING A LIVE AUGMENTED REALITY SCANNER
To accurately measure how patients react to algorithmic diagnoses I needed an environment that felt completely authentic. I developed a fully functional Augmented Reality prototype using MediaPipe to simulate a real medical diagnostic experience. This advanced setup allowed participants to physically scan their hands and receive dynamic visual feedback in real time during the evaluation phase.
THE RESEARCH VALUE :
THE RESEARCH VALUE :
Testing with a live scanner completely elevated the quality of the feedback. Instead of asking users to imagine a feature I observed their genuine reactions to an active tool. This high fidelity environment eliminated hypothetical bias and provided concrete data on how users evaluate technical competence during a diagnostic scan.
Testing with a live scanner completely elevated the quality of the feedback. Instead of asking users to imagine a feature I observed their genuine reactions to an active tool. This high fidelity environment eliminated hypothetical bias and provided concrete data on how users evaluate technical competence during a diagnostic scan.
PROVING THE CONCEPT :
PROVING THE CONCEPT :
This live interaction directly validated two core pillars of user trust. First the physical scan proved the system possessed the technical ability to visually assess actual symptoms. Second linking that visual output directly to World Health Organization guidelines instantly satisfied the need for systemic integrity. The prototype successfully demonstrated that advanced health applications can be both highly capable and deeply transparent.
This live interaction directly validated two core pillars of user trust. First the physical scan proved the system possessed the technical ability to visually assess actual symptoms. Second linking that visual output directly to World Health Organization guidelines instantly satisfied the need for systemic integrity. The prototype successfully demonstrated that advanced health applications can be both highly capable and deeply transparent.
DEFINING SYSTEM BOUNDARIES: THE EMERGENCY REDIRECT
To build genuine trust in an AI symptom checker, the system must know exactly when to stop talking. I designed an emergency hard-stop feature where high-risk inputs, such as "chest pain," instantly halt the diagnostic chat and trigger a redirect to local ambulance services. Instead of trying to retain the user in a continuous loop, the application deliberately hands off control during critical moments.
THE UX STRATEGY :
THE UX STRATEGY :
When an AI tries to handle life-or-death medical situations, it triggers severe algorithm aversion. Users instinctively distrust machines that overstep their boundaries and attempt to act like human doctors during a crisis. By forcing a redirect, the system proves it is self-aware of its own limitations. It trades a short-term engagement metric for long-term user confidence, showing patients that their physical safety always outweighs time-on-screen.
When an AI tries to handle life-or-death medical situations, it triggers severe algorithm aversion. Users instinctively distrust machines that overstep their boundaries and attempt to act like human doctors during a crisis. By forcing a redirect, the system proves it is self-aware of its own limitations. It trades a short-term engagement metric for long-term user confidence, showing patients that their physical safety always outweighs time-on-screen.
PROVING THE CONCEPT :
PROVING THE CONCEPT :
This interaction validated a core principle of product design that honesty builds reliance. By intentionally removing the user from the platform during an emergency, the AI establishes itself as a responsible, supportive guide rather than an infallible oracle. It successfully demonstrated that designing for safety and designing for trust are the exact same thing.
This interaction validated a core principle of product design that honesty builds reliance. By intentionally removing the user from the platform during an emergency, the AI establishes itself as a responsible, supportive guide rather than an infallible oracle. It successfully demonstrated that designing for safety and designing for trust are the exact same thing.
More Projects
Explainable AI
Healthcare AI
Hi-Fi Prototyping
Chatbot Interface Design
Black Box Mitigation
KLARIA APP, UX STRATEGY AND PROTOTYPE FOR TRANSPARENT AI
KLARIA APP, UX STRATEGY AND PROTOTYPE FOR TRANSPARENT AI
This overview summarizes a comprehensive product development lifecycle bridging foundational user research and high fidelity UI design. By conducting competitive analysis and developing dynamic interactive prototypes this project translates complex user psychology into a scalable design guide. The end result is a fully transparent digital health product leveraging explainable artificial intelligence, explicit role framing and prominent visual safety signaling to drive genuine user reliance across the entire platform.
Year :
2026
Industry :
Digital Health and HealthTech
Role :
Product Designer and UX Researcher
Project Duration :
9 months
↑0.00%
↑0.00%
Integrating explicit visual safety badges and privacy signaling, drastically elevated user confidence. The experimental group recorded a 5.94 mean trust score compared to just 2.56 for the neutral interface.
Integrating explicit visual safety badges and privacy signaling, drastically elevated user confidence. The experimental group recorded a 5.94 mean trust score compared to just 2.56 for the neutral interface.
Integrating explicit visual safety badges and privacy signaling, drastically elevated user confidence. The experimental group recorded a 5.94 mean trust score compared to just 2.56 for the neutral interface.
↑0%
↑0%
By replacing opaque algorithmic outputs with transparent explainable AI and personalized AR scans the in depth prototype achieved a peak trust score of 6.1 out of 7. This represents a 28% surge in patient confidence compared to the baseline no explanation model.
By replacing opaque algorithmic outputs with transparent explainable AI and personalized AR scans the in depth prototype achieved a peak trust score of 6.1 out of 7. This represents a 28% surge in patient confidence compared to the baseline no explanation model.
By replacing opaque algorithmic outputs with transparent explainable AI and personalized AR scans the in depth prototype achieved a peak trust score of 6.1 out of 7. This represents a 28% surge in patient confidence compared to the baseline no explanation model.
↑0.00%
↑0.00%
By explicitly defining the operational limitations and framing the application strictly as a supportive interim guide the system successfully calibrated patient expectations. This transparent boundary setting completely resolved role ambiguity and elevated the mean trust score to 5.45 out of 7 .
By explicitly defining the operational limitations and framing the application strictly as a supportive interim guide the system successfully calibrated patient expectations. This transparent boundary setting completely resolved role ambiguity and elevated the mean trust score to 5.45 out of 7 .
By explicitly defining the operational limitations and framing the application strictly as a supportive interim guide the system successfully calibrated patient expectations. This transparent boundary setting completely resolved role ambiguity and elevated the mean trust score to 5.45 out of 7 .
My role :
In my dual role as a Product Designer and UX Researcher I directed both the theoretical research and the practical design execution of Klaria. From conducting initial market evaluations to simulating live conversational dynamics through the Wizard of Oz method and running extensive quantitative vignette tests I ensured every interface decision was strictly backed by empirical user data.
CONTEXT :
This comprehensive initiative bridges practical product development with rigorous academic research. Beginning as a foundational competitive analysis of the digital health market it evolved into a fully interactive prototype test and culminated in a large scale quantitative study. The project establishes a complete and scalable design guide for building user reliance in artificial intelligence interfaces.
THE CORE CHALLENGE :
Millions of patients rely on digital symptom checkers to manage their health anxiety but these platforms frequently operate as completely opaque black boxes. They deliver high stakes medical advice without explaining their algorithmic logic or citing verifiable sources. This lack of transparency combined with a failure to explicitly define operational boundaries creates a severe trust deficit causing users to either blindly accept unverified outputs or completely reject the technology out of profound fear.
Methods :
• Competitive and Market Analysis
• User Persona and Job Story Development
• Dynamic Prototyping and Interface Design
• Wizard of Oz Simulation Testing
• Qualitative Thematic Analysis
• Quantitative Vignette Testing
Uncovering the Friction Points :
Through the Wizard of Oz simulation and thematic analysis I identified the exact cognitive barriers preventing everyday users from fully trusting the diagnostic artificial intelligence. The qualitative data revealed that pure algorithmic accuracy is meaningless if the interface fails to immediately communicate transparent reasoning, explicit operational boundaries and structural safety.
Journey 01: THE DIAGNOSTIC ASSESSMENT
Journey 01: THE DIAGNOSTIC ASSESSMENT
When patients initially interact with the symptom checker they are highly vulnerable and actively searching for clear structural assurance. The qualitative interviews demonstrated that this initial triage phase is where trust is most frequently broken because the system operates as an opaque black box.
When patients initially interact with the symptom checker they are highly vulnerable and actively searching for clear structural assurance. The qualitative interviews demonstrated that this initial triage phase is where trust is most frequently broken because the system operates as an opaque black box.
Friction Point A: UNSUBSTANTIATED ALGORITHMIC CLAIMS
Friction Point A: UNSUBSTANTIATED ALGORITHMIC CLAIMS
When moving away from a completely opaque black box system the initial prototype attempted to provide users with direct algorithmic reasoning. However this transition exposed a significant roadblock regarding how patients evaluate the credibility of digital medical claims and diagnostic outputs.
When moving away from a completely opaque black box system the initial prototype attempted to provide users with direct algorithmic reasoning. However this transition exposed a significant roadblock regarding how patients evaluate the credibility of digital medical claims and diagnostic outputs.
• The insight :
• The insight :
Through the Wizard of Oz simulation the data revealed a critical trust anomaly where providing a moderate explanation without tangible proof actually decreased overall user confidence. Providing users with a numerical likelihood score without offering a verifiable external source actively damaged system integrity. Participants explicitly stated that making specific medical claims without linking to authoritative guidelines made the application feel highly suspicious.
Through the Wizard of Oz simulation the data revealed a critical trust anomaly where providing a moderate explanation without tangible proof actually decreased overall user confidence. Providing users with a numerical likelihood score without offering a verifiable external source actively damaged system integrity. Participants explicitly stated that making specific medical claims without linking to authoritative guidelines made the application feel highly suspicious.
• The SOLUTION :
• The SOLUTION :
To resolve this trust deficit I designed the interface to satisfy a strict hierarchy of proof. I integrated explicit external validation by anchoring every diagnostic claim directly to authoritative medical guidelines such as the World Health Organization. This strategic design decision transformed the opaque algorithmic output into a transparent and verifiable claim which successfully increased the average user trust score to 6.1 out of 7.
To resolve this trust deficit I designed the interface to satisfy a strict hierarchy of proof. I integrated explicit external validation by anchoring every diagnostic claim directly to authoritative medical guidelines such as the World Health Organization. This strategic design decision transformed the opaque algorithmic output into a transparent and verifiable claim which successfully increased the average user trust score to 6.1 out of 7.
JOURNEY 02: THE ONBOARDING EXPERIENCE
JOURNEY 02: THE ONBOARDING EXPERIENCE
Before users even begin typing their symptoms they form critical expectations about the capabilities of the application. The qualitative research indicated that this initial entry phase is essential for establishing strict operational boundaries and preventing dangerous misinterpretations of the artificial intelligence.
Before users even begin typing their symptoms they form critical expectations about the capabilities of the application. The qualitative research indicated that this initial entry phase is essential for establishing strict operational boundaries and preventing dangerous misinterpretations of the artificial intelligence.
Friction Point B: SYSTEM ROLE AMBIGUITY
Friction Point B: SYSTEM ROLE AMBIGUITY
When digital symptom checkers fail to define their operational limitations users frequently misinterpret them as definitive medical authorities. This ambiguity leads to either dangerous algorithmic over reliance or complete system rejection disrupting the entire medical triage process.
When digital symptom checkers fail to define their operational limitations users frequently misinterpret them as definitive medical authorities. This ambiguity leads to either dangerous algorithmic over reliance or complete system rejection disrupting the entire medical triage process.
• The insight :
• The insight :
Qualitative interviews revealed that users actually view the application as a crucial tool for managing health concerns during the interim period before a formal doctor visit. However quantitative testing demonstrated that when the interface utilized authoritative framing without explicitly stating its limitations the average user trust score dropped to 2.37 out of 7 during the clinical evaluations.
Qualitative interviews revealed that users actually view the application as a crucial tool for managing health concerns during the interim period before a formal doctor visit. However quantitative testing demonstrated that when the interface utilized authoritative framing without explicitly stating its limitations the average user trust score dropped to 2.37 out of 7 during the clinical evaluations.
• The SOLUTION :
• The SOLUTION :
To properly calibrate expectations I designed clear onboarding disclaimers that explicitly framed the application as a supportive interim guide rather than a definitive diagnostic tool. This transparent boundary setting completely resolved the role ambiguity and successfully elevated the mean trust score to 5.45 out of 7 during the quantitative testing phase.
To properly calibrate expectations I designed clear onboarding disclaimers that explicitly framed the application as a supportive interim guide rather than a definitive diagnostic tool. This transparent boundary setting completely resolved the role ambiguity and successfully elevated the mean trust score to 5.45 out of 7 during the quantitative testing phase.
Journey 03: THE CONTEXTUAL ENTRY PHASE
Journey 03: THE CONTEXTUAL ENTRY PHASE
As patients evaluate a new digital health platform they carry an extreme baseline of privacy concern regarding their highly sensitive medical data. If the application fails to immediately provide absolute proof of structural safety users will instinctively abandon the platform before the diagnostic assessment even begins.
As patients evaluate a new digital health platform they carry an extreme baseline of privacy concern regarding their highly sensitive medical data. If the application fails to immediately provide absolute proof of structural safety users will instinctively abandon the platform before the diagnostic assessment even begins.
Friction Point C: LACK OF STRUCTURAL ASSURANCE
Friction Point C: LACK OF STRUCTURAL ASSURANCE
When symptom checkers operate without explicit visual security measures patients are forced to assume the highest level of personal risk. This lack of environmental safety creates a massive psychological barrier preventing vulnerable users from confidently disclosing their personal health information.
When symptom checkers operate without explicit visual security measures patients are forced to assume the highest level of personal risk. This lack of environmental safety creates a massive psychological barrier preventing vulnerable users from confidently disclosing their personal health information.
• The insight :
• The insight :
Quantitative testing revealed that forcing users to rely on an unverified environment severely damages overall system confidence. When the interface lacked explicit safety signaling the mean user trust score dropped to a critically low 2.56 out of 7 during the clinical evaluations. The research proved that participants require immediate visual proof of regulatory compliance to feel genuinely secure.
Quantitative testing revealed that forcing users to rely on an unverified environment severely damages overall system confidence. When the interface lacked explicit safety signaling the mean user trust score dropped to a critically low 2.56 out of 7 during the clinical evaluations. The research proved that participants require immediate visual proof of regulatory compliance to feel genuinely secure.
• The SOLUTION :
• The SOLUTION :
To resolve this intense perceived risk, I integrated prominent visual safety badges and medical device certifications directly into the user interface. This strategic addition of structural assurance transformed abstract legal jargon into a tangible heuristic for safety which successfully elevated the mean trust score to an outstanding 5.94 out of 7.
To resolve this intense perceived risk, I integrated prominent visual safety badges and medical device certifications directly into the user interface. This strategic addition of structural assurance transformed abstract legal jargon into a tangible heuristic for safety which successfully elevated the mean trust score to an outstanding 5.94 out of 7.
BEYOND STATIC INTERFACES: BUILDING A LIVE AUGMENTED REALITY SCANNER
To accurately measure how patients react to algorithmic diagnoses I needed an environment that felt completely authentic. I developed a fully functional Augmented Reality prototype using MediaPipe to simulate a real medical diagnostic experience. This advanced setup allowed participants to physically scan their hands and receive dynamic visual feedback in real time during the evaluation phase.
THE RESEARCH VALUE :
THE RESEARCH VALUE :
Testing with a live scanner completely elevated the quality of the feedback. Instead of asking users to imagine a feature I observed their genuine reactions to an active tool. This high fidelity environment eliminated hypothetical bias and provided concrete data on how users evaluate technical competence during a diagnostic scan.
Testing with a live scanner completely elevated the quality of the feedback. Instead of asking users to imagine a feature I observed their genuine reactions to an active tool. This high fidelity environment eliminated hypothetical bias and provided concrete data on how users evaluate technical competence during a diagnostic scan.
PROVING THE CONCEPT :
PROVING THE CONCEPT :
This live interaction directly validated two core pillars of user trust. First the physical scan proved the system possessed the technical ability to visually assess actual symptoms. Second linking that visual output directly to World Health Organization guidelines instantly satisfied the need for systemic integrity. The prototype successfully demonstrated that advanced health applications can be both highly capable and deeply transparent.
This live interaction directly validated two core pillars of user trust. First the physical scan proved the system possessed the technical ability to visually assess actual symptoms. Second linking that visual output directly to World Health Organization guidelines instantly satisfied the need for systemic integrity. The prototype successfully demonstrated that advanced health applications can be both highly capable and deeply transparent.
DEFINING SYSTEM BOUNDARIES: THE EMERGENCY REDIRECT
To build genuine trust in an AI symptom checker, the system must know exactly when to stop talking. I designed an emergency hard-stop feature where high-risk inputs, such as "chest pain," instantly halt the diagnostic chat and trigger a redirect to local ambulance services. Instead of trying to retain the user in a continuous loop, the application deliberately hands off control during critical moments.
THE UX STRATEGY :
THE UX STRATEGY :
When an AI tries to handle life-or-death medical situations, it triggers severe algorithm aversion. Users instinctively distrust machines that overstep their boundaries and attempt to act like human doctors during a crisis. By forcing a redirect, the system proves it is self-aware of its own limitations. It trades a short-term engagement metric for long-term user confidence, showing patients that their physical safety always outweighs time-on-screen.
When an AI tries to handle life-or-death medical situations, it triggers severe algorithm aversion. Users instinctively distrust machines that overstep their boundaries and attempt to act like human doctors during a crisis. By forcing a redirect, the system proves it is self-aware of its own limitations. It trades a short-term engagement metric for long-term user confidence, showing patients that their physical safety always outweighs time-on-screen.
PROVING THE CONCEPT :
PROVING THE CONCEPT :
This interaction validated a core principle of product design that honesty builds reliance. By intentionally removing the user from the platform during an emergency, the AI establishes itself as a responsible, supportive guide rather than an infallible oracle. It successfully demonstrated that designing for safety and designing for trust are the exact same thing.
This interaction validated a core principle of product design that honesty builds reliance. By intentionally removing the user from the platform during an emergency, the AI establishes itself as a responsible, supportive guide rather than an infallible oracle. It successfully demonstrated that designing for safety and designing for trust are the exact same thing.
More Projects
Explainable AI
Healthcare AI
Hi-Fi Prototyping
Chatbot Interface Design
Black Box Mitigation
KLARIA APP, UX STRATEGY AND PROTOTYPE FOR TRANSPARENT AI
KLARIA APP, UX STRATEGY AND PROTOTYPE FOR TRANSPARENT AI
This overview summarizes a comprehensive product development lifecycle bridging foundational user research and high fidelity UI design. By conducting competitive analysis and developing dynamic interactive prototypes this project translates complex user psychology into a scalable design guide. The end result is a fully transparent digital health product leveraging explainable artificial intelligence, explicit role framing and prominent visual safety signaling to drive genuine user reliance across the entire platform.
Year :
2026
Industry :
Digital Health and HealthTech
Role :
Product Designer and UX Researcher
Project Duration :
9 months
↑0.00%
↑0.00%
Integrating explicit visual safety badges and privacy signaling, drastically elevated user confidence. The experimental group recorded a 5.94 mean trust score compared to just 2.56 for the neutral interface.
Integrating explicit visual safety badges and privacy signaling, drastically elevated user confidence. The experimental group recorded a 5.94 mean trust score compared to just 2.56 for the neutral interface.
Integrating explicit visual safety badges and privacy signaling, drastically elevated user confidence. The experimental group recorded a 5.94 mean trust score compared to just 2.56 for the neutral interface.
↑0%
↑0%
By replacing opaque algorithmic outputs with transparent explainable AI and personalized AR scans the in depth prototype achieved a peak trust score of 6.1 out of 7. This represents a 28% surge in patient confidence compared to the baseline no explanation model.
By replacing opaque algorithmic outputs with transparent explainable AI and personalized AR scans the in depth prototype achieved a peak trust score of 6.1 out of 7. This represents a 28% surge in patient confidence compared to the baseline no explanation model.
By replacing opaque algorithmic outputs with transparent explainable AI and personalized AR scans the in depth prototype achieved a peak trust score of 6.1 out of 7. This represents a 28% surge in patient confidence compared to the baseline no explanation model.
↑0.00%
↑0.00%
By explicitly defining the operational limitations and framing the application strictly as a supportive interim guide the system successfully calibrated patient expectations. This transparent boundary setting completely resolved role ambiguity and elevated the mean trust score to 5.45 out of 7 .
By explicitly defining the operational limitations and framing the application strictly as a supportive interim guide the system successfully calibrated patient expectations. This transparent boundary setting completely resolved role ambiguity and elevated the mean trust score to 5.45 out of 7 .
By explicitly defining the operational limitations and framing the application strictly as a supportive interim guide the system successfully calibrated patient expectations. This transparent boundary setting completely resolved role ambiguity and elevated the mean trust score to 5.45 out of 7 .
My role :
In my dual role as a Product Designer and UX Researcher I directed both the theoretical research and the practical design execution of Klaria. From conducting initial market evaluations to simulating live conversational dynamics through the Wizard of Oz method and running extensive quantitative vignette tests I ensured every interface decision was strictly backed by empirical user data.
CONTEXT :
This comprehensive initiative bridges practical product development with rigorous academic research. Beginning as a foundational competitive analysis of the digital health market it evolved into a fully interactive prototype test and culminated in a large scale quantitative study. The project establishes a complete and scalable design guide for building user reliance in artificial intelligence interfaces.
THE CORE CHALLENGE :
Millions of patients rely on digital symptom checkers to manage their health anxiety but these platforms frequently operate as completely opaque black boxes. They deliver high stakes medical advice without explaining their algorithmic logic or citing verifiable sources. This lack of transparency combined with a failure to explicitly define operational boundaries creates a severe trust deficit causing users to either blindly accept unverified outputs or completely reject the technology out of profound fear.
Methods :
• Competitive and Market Analysis
• User Persona and Job Story Development
• Dynamic Prototyping and Interface Design
• Wizard of Oz Simulation Testing
• Qualitative Thematic Analysis
• Quantitative Vignette Testing
Uncovering the Friction Points :
Through the Wizard of Oz simulation and thematic analysis I identified the exact cognitive barriers preventing everyday users from fully trusting the diagnostic artificial intelligence. The qualitative data revealed that pure algorithmic accuracy is meaningless if the interface fails to immediately communicate transparent reasoning, explicit operational boundaries and structural safety.
Journey 01: THE DIAGNOSTIC ASSESSMENT
Journey 01: THE DIAGNOSTIC ASSESSMENT
When patients initially interact with the symptom checker they are highly vulnerable and actively searching for clear structural assurance. The qualitative interviews demonstrated that this initial triage phase is where trust is most frequently broken because the system operates as an opaque black box.
When patients initially interact with the symptom checker they are highly vulnerable and actively searching for clear structural assurance. The qualitative interviews demonstrated that this initial triage phase is where trust is most frequently broken because the system operates as an opaque black box.
Friction Point A: UNSUBSTANTIATED ALGORITHMIC CLAIMS
Friction Point A: UNSUBSTANTIATED ALGORITHMIC CLAIMS
When moving away from a completely opaque black box system the initial prototype attempted to provide users with direct algorithmic reasoning. However this transition exposed a significant roadblock regarding how patients evaluate the credibility of digital medical claims and diagnostic outputs.
When moving away from a completely opaque black box system the initial prototype attempted to provide users with direct algorithmic reasoning. However this transition exposed a significant roadblock regarding how patients evaluate the credibility of digital medical claims and diagnostic outputs.
• The insight :
• The insight :
Through the Wizard of Oz simulation the data revealed a critical trust anomaly where providing a moderate explanation without tangible proof actually decreased overall user confidence. Providing users with a numerical likelihood score without offering a verifiable external source actively damaged system integrity. Participants explicitly stated that making specific medical claims without linking to authoritative guidelines made the application feel highly suspicious.
Through the Wizard of Oz simulation the data revealed a critical trust anomaly where providing a moderate explanation without tangible proof actually decreased overall user confidence. Providing users with a numerical likelihood score without offering a verifiable external source actively damaged system integrity. Participants explicitly stated that making specific medical claims without linking to authoritative guidelines made the application feel highly suspicious.
• The SOLUTION :
• The SOLUTION :
To resolve this trust deficit I designed the interface to satisfy a strict hierarchy of proof. I integrated explicit external validation by anchoring every diagnostic claim directly to authoritative medical guidelines such as the World Health Organization. This strategic design decision transformed the opaque algorithmic output into a transparent and verifiable claim which successfully increased the average user trust score to 6.1 out of 7.
To resolve this trust deficit I designed the interface to satisfy a strict hierarchy of proof. I integrated explicit external validation by anchoring every diagnostic claim directly to authoritative medical guidelines such as the World Health Organization. This strategic design decision transformed the opaque algorithmic output into a transparent and verifiable claim which successfully increased the average user trust score to 6.1 out of 7.
JOURNEY 02: THE ONBOARDING EXPERIENCE
JOURNEY 02: THE ONBOARDING EXPERIENCE
Before users even begin typing their symptoms they form critical expectations about the capabilities of the application. The qualitative research indicated that this initial entry phase is essential for establishing strict operational boundaries and preventing dangerous misinterpretations of the artificial intelligence.
Before users even begin typing their symptoms they form critical expectations about the capabilities of the application. The qualitative research indicated that this initial entry phase is essential for establishing strict operational boundaries and preventing dangerous misinterpretations of the artificial intelligence.
Friction Point B: SYSTEM ROLE AMBIGUITY
Friction Point B: SYSTEM ROLE AMBIGUITY
When digital symptom checkers fail to define their operational limitations users frequently misinterpret them as definitive medical authorities. This ambiguity leads to either dangerous algorithmic over reliance or complete system rejection disrupting the entire medical triage process.
When digital symptom checkers fail to define their operational limitations users frequently misinterpret them as definitive medical authorities. This ambiguity leads to either dangerous algorithmic over reliance or complete system rejection disrupting the entire medical triage process.
• The insight :
• The insight :
Qualitative interviews revealed that users actually view the application as a crucial tool for managing health concerns during the interim period before a formal doctor visit. However quantitative testing demonstrated that when the interface utilized authoritative framing without explicitly stating its limitations the average user trust score dropped to 2.37 out of 7 during the clinical evaluations.
Qualitative interviews revealed that users actually view the application as a crucial tool for managing health concerns during the interim period before a formal doctor visit. However quantitative testing demonstrated that when the interface utilized authoritative framing without explicitly stating its limitations the average user trust score dropped to 2.37 out of 7 during the clinical evaluations.
• The SOLUTION :
• The SOLUTION :
To properly calibrate expectations I designed clear onboarding disclaimers that explicitly framed the application as a supportive interim guide rather than a definitive diagnostic tool. This transparent boundary setting completely resolved the role ambiguity and successfully elevated the mean trust score to 5.45 out of 7 during the quantitative testing phase.
To properly calibrate expectations I designed clear onboarding disclaimers that explicitly framed the application as a supportive interim guide rather than a definitive diagnostic tool. This transparent boundary setting completely resolved the role ambiguity and successfully elevated the mean trust score to 5.45 out of 7 during the quantitative testing phase.
Journey 03: THE CONTEXTUAL ENTRY PHASE
Journey 03: THE CONTEXTUAL ENTRY PHASE
As patients evaluate a new digital health platform they carry an extreme baseline of privacy concern regarding their highly sensitive medical data. If the application fails to immediately provide absolute proof of structural safety users will instinctively abandon the platform before the diagnostic assessment even begins.
As patients evaluate a new digital health platform they carry an extreme baseline of privacy concern regarding their highly sensitive medical data. If the application fails to immediately provide absolute proof of structural safety users will instinctively abandon the platform before the diagnostic assessment even begins.
Friction Point C: LACK OF STRUCTURAL ASSURANCE
Friction Point C: LACK OF STRUCTURAL ASSURANCE
When symptom checkers operate without explicit visual security measures patients are forced to assume the highest level of personal risk. This lack of environmental safety creates a massive psychological barrier preventing vulnerable users from confidently disclosing their personal health information.
When symptom checkers operate without explicit visual security measures patients are forced to assume the highest level of personal risk. This lack of environmental safety creates a massive psychological barrier preventing vulnerable users from confidently disclosing their personal health information.
• The insight :
• The insight :
Quantitative testing revealed that forcing users to rely on an unverified environment severely damages overall system confidence. When the interface lacked explicit safety signaling the mean user trust score dropped to a critically low 2.56 out of 7 during the clinical evaluations. The research proved that participants require immediate visual proof of regulatory compliance to feel genuinely secure.
Quantitative testing revealed that forcing users to rely on an unverified environment severely damages overall system confidence. When the interface lacked explicit safety signaling the mean user trust score dropped to a critically low 2.56 out of 7 during the clinical evaluations. The research proved that participants require immediate visual proof of regulatory compliance to feel genuinely secure.
• The SOLUTION :
• The SOLUTION :
To resolve this intense perceived risk, I integrated prominent visual safety badges and medical device certifications directly into the user interface. This strategic addition of structural assurance transformed abstract legal jargon into a tangible heuristic for safety which successfully elevated the mean trust score to an outstanding 5.94 out of 7.
To resolve this intense perceived risk, I integrated prominent visual safety badges and medical device certifications directly into the user interface. This strategic addition of structural assurance transformed abstract legal jargon into a tangible heuristic for safety which successfully elevated the mean trust score to an outstanding 5.94 out of 7.
BEYOND STATIC INTERFACES: BUILDING A LIVE AUGMENTED REALITY SCANNER
To accurately measure how patients react to algorithmic diagnoses I needed an environment that felt completely authentic. I developed a fully functional Augmented Reality prototype using MediaPipe to simulate a real medical diagnostic experience. This advanced setup allowed participants to physically scan their hands and receive dynamic visual feedback in real time during the evaluation phase.
THE RESEARCH VALUE :
THE RESEARCH VALUE :
Testing with a live scanner completely elevated the quality of the feedback. Instead of asking users to imagine a feature I observed their genuine reactions to an active tool. This high fidelity environment eliminated hypothetical bias and provided concrete data on how users evaluate technical competence during a diagnostic scan.
Testing with a live scanner completely elevated the quality of the feedback. Instead of asking users to imagine a feature I observed their genuine reactions to an active tool. This high fidelity environment eliminated hypothetical bias and provided concrete data on how users evaluate technical competence during a diagnostic scan.
PROVING THE CONCEPT :
PROVING THE CONCEPT :
This live interaction directly validated two core pillars of user trust. First the physical scan proved the system possessed the technical ability to visually assess actual symptoms. Second linking that visual output directly to World Health Organization guidelines instantly satisfied the need for systemic integrity. The prototype successfully demonstrated that advanced health applications can be both highly capable and deeply transparent.
This live interaction directly validated two core pillars of user trust. First the physical scan proved the system possessed the technical ability to visually assess actual symptoms. Second linking that visual output directly to World Health Organization guidelines instantly satisfied the need for systemic integrity. The prototype successfully demonstrated that advanced health applications can be both highly capable and deeply transparent.
DEFINING SYSTEM BOUNDARIES: THE EMERGENCY REDIRECT
To build genuine trust in an AI symptom checker, the system must know exactly when to stop talking. I designed an emergency hard-stop feature where high-risk inputs, such as "chest pain," instantly halt the diagnostic chat and trigger a redirect to local ambulance services. Instead of trying to retain the user in a continuous loop, the application deliberately hands off control during critical moments.
THE UX STRATEGY :
THE UX STRATEGY :
When an AI tries to handle life-or-death medical situations, it triggers severe algorithm aversion. Users instinctively distrust machines that overstep their boundaries and attempt to act like human doctors during a crisis. By forcing a redirect, the system proves it is self-aware of its own limitations. It trades a short-term engagement metric for long-term user confidence, showing patients that their physical safety always outweighs time-on-screen.
When an AI tries to handle life-or-death medical situations, it triggers severe algorithm aversion. Users instinctively distrust machines that overstep their boundaries and attempt to act like human doctors during a crisis. By forcing a redirect, the system proves it is self-aware of its own limitations. It trades a short-term engagement metric for long-term user confidence, showing patients that their physical safety always outweighs time-on-screen.
PROVING THE CONCEPT :
PROVING THE CONCEPT :
This interaction validated a core principle of product design that honesty builds reliance. By intentionally removing the user from the platform during an emergency, the AI establishes itself as a responsible, supportive guide rather than an infallible oracle. It successfully demonstrated that designing for safety and designing for trust are the exact same thing.
This interaction validated a core principle of product design that honesty builds reliance. By intentionally removing the user from the platform during an emergency, the AI establishes itself as a responsible, supportive guide rather than an infallible oracle. It successfully demonstrated that designing for safety and designing for trust are the exact same thing.