Crowdsourcing Design Guidance for Contextual Adaptation of Text Content in Augmented Reality

Title of the Paper

Crowdsourcing Design Guidance for Contextual Adaptation of Text Content in Augmented Reality

Paper Information

• Subject Area: Augmented Reality (AR), Crowdsourcing, Interface Design
• Keywords: Augmented Reality, Crowdsourcing, Privacy, Text Adaptation, Visual Design, Experimental Methods, Dynamic Interfaces, User Experience, Contextual Influence, Head-Mounted Displays

Research Background and Problem

• Identified Problems or Challenges:

  • AR interface design needs to adapt to users' physical environments, but developers cannot anticipate the specific contexts in which applications will be deployed.
  • Existing laboratory studies are highly controlled, have low external validity, and are costly for collecting diverse data.
  • There is a lack of design guidance for dynamic text adaptation, such as how to select text styles and colors based on the background.

• Why This Problem is Important:

  • Augmented reality is increasingly integrated into users' daily environments, and the quality of its design is critical to user experience.
  • Dynamically adjusting text styles in diverse contexts is a key feature for many AR applications.

• Research Motivation and Related Work:

  • Related studies have utilized crowdsourcing for cognitive-behavioral experiments or interface design but often overlook the impact of users' environmental contexts on AR.
  • Previous literature has explored the effects of text color, background, and layout on readability but lacks broad data support based on users' contextual environments.

Solution

• Method or Solution:

  • A novel crowdsourcing experimental method is proposed, capturing user environment data via a mobile application.
  • A privacy-aware design is employed, allowing users to capture images during tasks and dynamically adjust text appearance while providing selective sharing capabilities.
  • The method is applied to the design challenges of text panel color and layout.

• Innovative Aspects:

  • Crowdsourcing is used to collect diverse user context data while ensuring user privacy.
  • The integration of crowdsourced user evaluations with laboratory research results supports AR design.
  • A mechanism is provided to transition from low-fidelity data to high-fidelity AR head-mounted devices.

• Implementation Steps:

  • Users capture images of their surrounding environment using their mobile device cameras.
  • Users adjust properties such as text panel color and position in a low-fidelity AR interface.
  • User feedback is used to optimize text appearance design, with options for users to pixelate shared images to protect privacy.
  • Data analysis combines user preferences with environmental characteristics to provide guidance for AR interface design.

Research Outcomes

• Specific Outcomes:

  • Two crowdsourcing experiments were conducted via Amazon Mechanical Turk, recruiting 400 participants and collecting approximately 2,000 image samples from 22 countries.
  • The influence of background color and texture on text panel color and layout choices was validated.
  • A dynamic text panel adaptation system for Microsoft HoloLens was developed.

• Advantages Compared to Existing Solutions:

  • Overcomes the external validity issues of laboratory studies and enables rapid collection of diverse real-world environment data.
  • Automated analysis tools extract preference patterns from large-scale data more efficiently than traditional manual analysis.
  • Ensures privacy protection, allowing users to refuse or blur shared image data.

• Experimental or Evaluation Results:

  • User preferences for blue and red text panels were validated across various backgrounds.
  • Participants provided positive feedback on the privacy protection features, with over 76% finding the blurring function very useful for safeguarding privacy.
  • Experimental results aligned with existing laboratory studies on panel color and text color preferences.

• Limitations and Future Directions:

  • Display differences on mobile devices may introduce noise in color selection, which needs to be controlled in future studies.
  • The current experiment only allows users to adjust panel color and layout, with limited interaction space; more design options could be explored.
  • Future work could expand to real-time video capture, incorporating temporal and spatial consistency studies to improve the accuracy of dynamic adaptation analysis.

Conclusion

• This paper proposes a privacy-aware crowdsourcing method that efficiently collects user environment data to support context-aware design decisions in AR applications.
• The study validates that design guidance obtained from low-fidelity mobile AR experiments can be translated to high-fidelity head-mounted display devices.
• Overall, the method provides an economical, effective, and scalable research tool for complex AR interface design, significantly advancing the external validity of AR design research.