Designing Trustworthy Smartwatch Health Data Collection
Project Description:
Health data, inherently sensitive, plays an important role in shaping our understanding of population health, predicting diseases, and driving innovation. However, handling such data necessitates a delicate balance between technological innovation and ethical responsibility to safeguard user trust, security, and dignity. This quick little case study delves into the realm where technology and healthcare intersect, with a focus on smartwatch data collection, model design, ethical practices, and human-computer interaction (HCI) principles. As this case study explores, ethical and design challenges are abundant when utilizing smartwatch data for health purposes. While such data holds immense potential for population health insights and disease prediction, ethical considerations and responsible design are crucial.
Challenges:
Privacy and Transparency: Smartwatches often lack transparency regarding data collection and usage. Users deserve clear information on what data is collected, how it’s utilized, and with whom it’s shared.
Third-Party Access: Users often lack control over how their data is shared with third parties and this app update can help them control that, sharing to third parties who will use this data ethically with the user’s permissions
Explainability: Limited knowledge about the internal workings of algorithms can hinder user trust and understanding. So an in depth information and visualization helps the user to see the use of prediction when their data is used.
User Control: Users should have granular control over data storage duration and the option to opt-out completely.
Bias and Fairness: Algorithmic bias risks inaccurate insights and unfair treatment based on cultural or demographic factors.
Project Goals:
Literature Review: Analyze current trends in smartwatch data collection, utilization, and existing model designs.
Design Analysis: Evaluate existing models and propose improvements focused on transparency, explainability, and user control.
Ethical Evaluation: Scrutinize ethical practices of leading companies and recommend enhancements.
Accomplishments:
Conducted a comprehensive literature review and research on existing models.
Identified risks and shortcomings in current practices.
Analyzed sampled data for a deeper understanding.
Proposed Solutions:
Enhanced Transparency: Provide easily accessible information sheets outlining data collection, usage, and storage.
Informed Consent: Clearly explain third-party access scenarios and collaboration details.
Improved Explainability: Utilize interactive visualizations and clear explanations to communicate the workings of algorithms.
Granular User Control: Allow users to control storage duration and opt-out options.
Inclusive Design: Promote diverse cultural representation in data collection and model development to minimize bias.
Visual Demo:
UI Design: Utilize clear language, visual hierarchy, and intuitive controls to enhance user understanding and control.
Interactive Visualizations: Provide dynamic visuals to explain algorithm functionality. This will help the users to understand how their data is being used giving them better understanding of how it can predict information as well.
User Control Interface: Offer simple opt-out options and granular data retention controls.
Cultural Preferences: Integrate a “cultural preferences” option to cater to diverse populations. Especially with a health app where people of different cultures can set preferences based on their cultural needs.
AI + HCI Heuristics:
This project aligns with several HCI principles-
Visibility of the System: Real-time progress updates and clear explanations improve user trust and control.
Match between System and Real World: Utilizing user-familiar language for health terms fosters better understanding.
Design for Different Stakeholders: Tailoring information presentation based on user needs caters to diverse audiences.
Unmet Heuristics:
Predicting User Emotional State: This remains an ongoing research area with ethical and technical complexities.
Error Recovery Guidance in Learning Modules: Designing error-proof interfaces and providing clear guidance remain essential areas for future improvement.
Future Work:
Federated Learning: Explore utilizing data on user devices to minimize data sharing risks.
Dynamic Privacy Settings: Develop settings that adapt to user context and preferences.
Enhanced Cultural Customization: Implement features that cater to diverse cultural and social contexts.
Co-author:
Tanmay Sameer Shravge
