Supporting Informed Self-Disclosure: Design Recommendations for Presenting AI-Estimates of Privacy Risks to Users

Paper Title

Supporting Informed Self-Disclosure: Design Recommendations for Presenting AI-Estimates of Privacy Risks to Users

Publication Info

• Topic area: User-centered design for privacy risk communication in online platforms.
• Keywords: Privacy risks, self-disclosure, population risk estimates, pseudonymity, online anonymity, design recommendations, explainable AI, user empowerment, privacy awareness, re-identification.

Background and Problem

• Problem / challenge: Users on pseudonymous platforms like Reddit often lack tools to assess the privacy risks of their self-disclosures, leading to over-disclosure or unnecessary self-censorship. Existing privacy tools fail to provide actionable, comprehensible, and trustworthy feedback on re-identification risks.
• Significance: Effective tools for privacy risk awareness can empower users to safely disclose sensitive information online, balancing the benefits of self-disclosure with the risks of re-identification.
• Motivation and related work: Prior research has shown the fragility of anonymization techniques and the ease of re-identification using minimal data points. While frameworks like SPAF emphasize awareness, motivation, and ability for privacy tools, there is limited research on how to present quantifiable privacy risks (e.g., population risk estimates or PREs) in a user-friendly manner.

Solution

• Proposed approach: The paper explores five design concepts for presenting Population Risk Estimates (PREs) to users, using speculative design and comic-boarding to evaluate their impact on user awareness, motivation, and ability to address privacy risks.
• Novelty:
  1. Systematic exploration of how different PRE presentations affect user decision-making.
  2. Behavioral insights into the psychological and emotional impacts of quantified privacy feedback.
  3. Evidence-based design recommendations for presenting PREs to balance user engagement and informed decision-making.
  4. Application of comic-boarding to elicit user feedback on abstract privacy technologies.
• Procedure and key techniques:
  1. Developed five PRE designs inspired by prior privacy and security literature.
  2. Embedded these designs in four narrative vignettes depicting privacy threat scenarios.
  3. Conducted an online survey with 44 Reddit users, using comic-boarding and reflective writing to gather qualitative and quantitative feedback.
  4. Analyzed user responses to derive design recommendations for effective PRE presentation.

Results

• Concrete findings:
  • PREs raised risk awareness in 74% of participant reflections, with 79 out of 132 scenarios ending in successful avoidance of re-identification.
  • 63% of participants reported negative emotional reactions (e.g., anxiety) to PREs, while 6% felt empowered.
  • Misinterpretations of PRE outputs occurred in 14 cases, highlighting the need for clearer communication.
• Advantage over baselines:
  • PREs motivated adaptive self-censorship (e.g., editing posts to reduce risk) in 87 cases, compared to maladaptive responses (e.g., avoiding posting entirely) in 25 cases.
  • Participants valued PREs for making abstract risks more concrete, though they emphasized the need for actionable guidance.
• Experiments / evaluation:
  • Participants reviewed three out of 20 comic-boards, each combining a PRE design with a narrative vignette.
  • Feedback was analyzed using inductive coding, yielding 18 themes and 65 sub-themes.
  • Quantitative analysis showed no significant preference for any single PRE design, though Design #5 (Risk by Disclosure) was most favored.
• Limitations and future work:
  • Sample skewed toward younger, U.S.-based, tech-savvy participants, limiting generalizability.
  • Study focused on Reddit and pseudonymous platforms; future work could explore other contexts like journalism or political dissent.
  • PRE designs did not account for risks from cross-platform data aggregation or posting history.

Summary

This study explores how to present Population Risk Estimates (PREs) to users on pseudonymous platforms to support informed self-disclosure. Through comic-boarding and a survey of 44 Reddit users, the authors found that PREs effectively raise risk awareness but can lead to anxiety and misinterpretation without clear guidance. The paper provides four design recommendations: (1) explain how disclosures may be exploited, (2) clarify how PREs are calculated, (3) offer actionable suggestions for reducing risks, and (4) use intuitive, jargon-free language. These findings contribute to the design of user-friendly privacy tools that balance risk awareness with user empowerment.