Mind The Gap: Designers and Standards on Algorithmic System Transparency for Users

Document Title

Mind The Gap: Designers and Standards on Algorithmic System Transparency for Users

Document Information

• Subject Area: Human-Computer Interaction and Algorithmic System Transparency
• Keywords: Design Practices, Standards, Transparency, Algorithmic Systems, Human-Computer Interaction, Design Guidelines, Artificial Intelligence

Research Background and Issues

• Problems or Challenges Identified by the Authors:

  • Despite the growing demand for algorithmic system transparency, designers often lack clarity on how to achieve this goal in practice.
  • Implementing transparency frequently faces a "gap" between principles and practice, especially due to the lack of clear guidance on designing transparency in algorithmic systems.

• Importance of the Problem:

  • Algorithmic systems have substantial impacts on individuals and communities, necessitating higher levels of transparency to support accountability and reduce information and power asymmetries.
  • Legal frameworks such as the EU AI Act and the U.S. Platform Transparency Act are increasingly mandating greater transparency in algorithmic systems, but their practical implications for designers remain unclear.

• Research Motivation and Related Work:

  • Current transparency standards (e.g., IEEE 7001) propose methods for measuring transparency, but their effectiveness in aiding designers has not been sufficiently explored.
  • The authors chose the IEEE 7001 standard to investigate designers' understanding of transparency and its alignment with the standard's recommendations.

Solution

• Methods or Solutions Proposed by the Authors:

  • A mixed-method study combining online surveys and follow-up interviews to examine designers' understanding and acceptance of transparency standards.
  • Testing designers using specific recommendations from the IEEE 7001 standard, particularly its transparency "levels" (TL0-TL5) designed for end-user applications.

• Innovative Aspects:

  • The first systematic study of the cognitive and practical gap between designers and transparency standards.
  • Proposed mechanisms to enhance designers' awareness of transparency and identified directions for future development.

• Implementation Steps and Key Techniques:

  • Designed two studies: an online survey to understand designers' experiences and challenges, and interviews to further test their comprehension of IEEE 7001 transparency recommendations.
  • Used IEEE 7001 recommendations as the core benchmark, including transparency levels (e.g., TL1 to TL5) to evaluate designers' understanding.

Research Outcomes

• Specific Findings:

  • Most participants acknowledged the importance of transparency but found it difficult to implement. Significant gaps remain between designers' understanding and practice of transparency standards.
  • IEEE 7001 transparency recommendations were perceived by designers as overly abstract and challenging to apply in specific contexts.
  • Designers suggested potential tools and principles to enhance transparency, but these diverged from the standard's recommendations.

• Advantages Compared to Existing Solutions:

  • This study deepened the understanding of the gap between designers and transparency standards, providing actionable feedback.
  • Investigated specific challenges faced by designers and summarized recommendations for promoting transparency.

• Experimental or Evaluation Results:

  • Over 50% of interviewed designers believed that IEEE 7001 transparency recommendations (TL1-TL3) were only conditionally applicable in specific scenarios, while 32% deemed them entirely inapplicable.
  • Surveys and interviews revealed key obstacles such as time constraints, resource limitations, and lack of clear definitions in design processes.

• Limitations and Future Directions:

  • Limited sample size, with most participants based in Europe, lacking global representativeness; future studies could expand to more regions and contexts.
  • Further research is needed on other stakeholders beyond designers (e.g., users, procurement officers).
  • Suggested future standard development to explore specific case studies, tool support, and multi-stakeholder collaboration.
  • Called for globalization and cross-language adaptation in transparency research and standard-setting.

Conclusion

This study reveals the significant gap between transparency standards and design practices, exploring potential pathways to improve transparency design. Findings indicate that transparency standards have limited impact, and designers require more support, collaboration, and awareness to make algorithmic systems more transparent. Future development should integrate designers' practices, transparency standards, and user needs to drive responsible AI system design forward.