The Impacts of Transparency and Personalization on Feelings of Agency and Connection in Democratic Decision Making

Paper Title

The Impacts of Transparency and Personalization on Feelings of Agency and Connection in Democratic Decision Making

Publication Info

• Topic area: Effects of transparency and personalization in civic technology on democratic participation.
• Keywords: transparency, personalization, civic technology, democratic decision-making, agency, community connection, vertical transparency, horizontal transparency, social curiosity, AI-generated feedback.

Background and Problem

• Problem / challenge: Limited empirical understanding of how transparency and personalization in civic technologies affect participants' perceptions of legitimacy, agency, and connection, particularly in scenarios with winners and losers.
• Significance: Enhancing transparency and responsiveness in civic decision-making could improve trust, legitimacy, and community engagement, addressing challenges in democratic governance.
• Motivation and related work: Prior research has explored transparency and responsiveness in civic technologies but has not sufficiently examined the effects of AI-enabled personalization or how transparency mechanisms impact participants differently in favorable versus unfavorable outcomes.

Solution

• Proposed approach: A controlled experiment testing three levels of transparency (control, generic, personalized) in a civic decision-making scenario, with a focus on vertical and horizontal transparency, agency, and social curiosity.
• Novelty:
  1. Empirical evaluation of AI-generated personalized transparency in civic decision-making.
  2. Analysis of transparency effects on both winners and losers in democratic outcomes.
  3. Exploration of horizontal transparency's role in fostering community connection.
  4. Integration of qualitative and quantitative methods to assess participant experiences.
• Procedure and key techniques:
  • Participants (N=266) advocated for a skate park or tennis court in a hypothetical scenario.
  • Participants were exposed to one of three transparency conditions: control (minimal justification), generic (diverse voices cited), or personalized (participant's contribution explicitly cited).
  • Surveys measured vertical transparency, horizontal transparency, agency, and social curiosity using Likert scales.
  • AI-generated content was used for personalized responses and community voice data.
  • Qualitative coding of open-ended responses and usage analytics complemented quantitative findings.

Results

• Concrete findings:
  • Generic transparency improved vertical transparency (β = 0.56, p = .012) and agency (β = 0.56, p = .006) compared to control.
  • Personalized transparency had mixed effects, with no significant advantage over generic transparency for agency or vertical transparency.
  • Horizontal transparency was enhanced by generic transparency (β = 0.42, p = .015) but not by personalization.
  • Winners reported higher vertical transparency (β = 0.84, p < .001) and agency (β = 1.27, p < .001) than losers.
• Advantage over baselines:
  • Generic transparency consistently outperformed control and sometimes personalized transparency in enhancing perceptions of legitimacy and agency.
  • Personalized transparency occasionally backfired, creating feelings of tokenism for some participants.
• Experiments / evaluation:
  • 2×3 factorial design with six conditions (control/generic/personalized × winner/loser).
  • Measures included Likert-scale surveys, qualitative coding of 1064 open-ended responses, and usage analytics.
  • Participants engaged with transparency features at an 85.5% interaction rate, with significant scrolling and citation-hovering activity.
• Limitations and future work:
  • Hypothetical scenario limits generalizability to real-world contexts.
  • AI-generated voices may not fully capture community diversity.
  • Future work should explore higher-stakes, complex civic contexts and refine personalization strategies to mitigate negative reactions.

Summary

This study evaluated the impacts of transparency and personalization in civic decision-making, finding that generic transparency consistently enhanced perceptions of legitimacy, agency, and horizontal transparency, while personalized transparency had mixed effects and sometimes backfired. Winners reported higher outcomes than losers, but transparency buffered negative reactions to losing. The findings suggest that civic technologies should prioritize interactive, explorable voice banks and shared-value framing over AI-generated personalization. Future research should investigate personalization's nuanced effects and explore transparency mechanisms in more complex, real-world scenarios.