Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

When Feasibility of Fairness Audits Relies on Willingness to Share Data: Examining User Acceptance of Multi-Party Computation Protocols for Fairness Monitoring

Changyang He, Parnian Jahangirirad, Lin Kyi, Asia J. Biega

TL;DR

This paper investigates whether fairness audits enabled by multi-party computation (MPC) are feasible given users' willingness to share sensitive data under GDPR and AI Act constraints. It conducts an online survey of 833 European job seekers, employing scenario-based conjoint analysis and direct attribute ranking to study MPC protocol designs across seven attributes (benefits, risks, and dual aspects). Key findings reveal a divergence between direct and simulated judgments: direct responses emphasize privacy-risk features, while simulated decisions shift toward benefit-related factors such as fairness objectives and monetary incentives; moreover, fairness and privacy orientations, plus personal context, shape acceptance. The study offers design and communication guidance for privacy-preserving fairness monitoring, including trust-building through distributed storage with credible third parties and tailored consent experiences to support informed participation and representative data collection.

Abstract

Fairness monitoring is critical for detecting algorithmic bias, as mandated by the EU AI Act. Since such monitoring requires sensitive user data (e.g., ethnicity), the AI Act permits its processing only with strict privacy measures, such as multi-party computation (MPC), in compliance with the GDPR. However, the effectiveness of such secure monitoring protocols ultimately depends on people's willingness to share their data. Little is known about how different MPC protocol designs shape user acceptance. To address this, we conducted an online survey with 833 participants in Europe, examining user acceptance of various MPC protocol designs for fairness monitoring. Findings suggest that users prioritized risk-related attributes (e.g., privacy protection mechanism) in direct evaluation but benefit-related attributes (e.g., fairness objective) in simulated choices, with acceptance shaped by their fairness and privacy orientations. We derive implications for deploying and communicating privacy-preserving protocols in ways that foster informed consent and align with user expectations.

When Feasibility of Fairness Audits Relies on Willingness to Share Data: Examining User Acceptance of Multi-Party Computation Protocols for Fairness Monitoring

TL;DR

This paper investigates whether fairness audits enabled by multi-party computation (MPC) are feasible given users' willingness to share sensitive data under GDPR and AI Act constraints. It conducts an online survey of 833 European job seekers, employing scenario-based conjoint analysis and direct attribute ranking to study MPC protocol designs across seven attributes (benefits, risks, and dual aspects). Key findings reveal a divergence between direct and simulated judgments: direct responses emphasize privacy-risk features, while simulated decisions shift toward benefit-related factors such as fairness objectives and monetary incentives; moreover, fairness and privacy orientations, plus personal context, shape acceptance. The study offers design and communication guidance for privacy-preserving fairness monitoring, including trust-building through distributed storage with credible third parties and tailored consent experiences to support informed participation and representative data collection.

Abstract

Fairness monitoring is critical for detecting algorithmic bias, as mandated by the EU AI Act. Since such monitoring requires sensitive user data (e.g., ethnicity), the AI Act permits its processing only with strict privacy measures, such as multi-party computation (MPC), in compliance with the GDPR. However, the effectiveness of such secure monitoring protocols ultimately depends on people's willingness to share their data. Little is known about how different MPC protocol designs shape user acceptance. To address this, we conducted an online survey with 833 participants in Europe, examining user acceptance of various MPC protocol designs for fairness monitoring. Findings suggest that users prioritized risk-related attributes (e.g., privacy protection mechanism) in direct evaluation but benefit-related attributes (e.g., fairness objective) in simulated choices, with acceptance shaped by their fairness and privacy orientations. We derive implications for deploying and communicating privacy-preserving protocols in ways that foster informed consent and align with user expectations.
Paper Structure (72 sections, 3 equations, 3 figures, 6 tables)

This paper contains 72 sections, 3 equations, 3 figures, 6 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Privacy-Preserving Fairness Monitoring
  4. Differential Privacy (DP)
  5. Synthetic Datasets
  6. Multi-Party Computation (MPC)
  7. Our Contributions.
  8. User Acceptance of Data Sharing under Privacy Constraints
  9. Our Contributions.
  10. Situating the Study: MPC Protocols for Fairness Monitoring
  11. From Protocol Designs to User Acceptance: A Privacy Calculus Perspective
  12. Benefit-related Attributes
  13. Fairness Objective: Which fairness objective does the data donation contribute to?
  14. Monetary Incentive: How is the data donor compensated?
  15. Risk-related Attributes
  16. ...and 57 more sections

Figures (3)

  • Figure 1: Example of a Conjoint Task Designed to Examine Users' Trade-offs Among Protocol Design Attributes. Respondents choose between two MPC fairness monitoring protocol designs or "none of these". Conjoint analysis is applied to analyze their selections to reveal attribute importance.
  • Figure 2: Means and Standard Deviations of Users' Fairness and Privacy Orientations, Prior Experience, and System Literacy
  • Figure 3: Attribute Importance in MPC Fairness Monitoring Protocol Design, Including (a) Revealed Importance Based on Choice-Based Conjoint Analysis and (b) Stated Importance Based on Attribute Ranking