Democratizing Differential Privacy: A Participatory AI Framework for Public Decision-Making

TL;DR

This work tackles the challenge of implementing differential privacy (DP) in public-sector AI while preserving democratic accountability. It introduces a participatory AI framework that couples adaptive $\epsilon$-DP selection using TOPSIS-based multi-criteria decision analysis with an explainable noise-injection module (featuring real-time Mean Absolute Error visualizations) and GPT-4-powered impact analysis, plus a dynamic legal-compliance mechanism that adjusts privacy budgets as regulations evolve. The approach enables stakeholder-guided privacy tuning, translating user priorities into mathematically valid privacy parameters and providing transparent explanations of privacy-utility trade-offs. Through simulation on the Household Electricity Demand dataset, the authors demonstrate meaningful mappings from preferences to DP configurations, quantified by strong correlations between privacy budgets and utility, and show how adaptive budgeting can better balance privacy with public-service data utility in governance contexts.

Abstract

This paper introduces a conversational interface system that enables participatory design of differentially private AI systems in public sector applications. Addressing the challenge of balancing mathematical privacy guarantees with democratic accountability, we propose three key contributions: (1) an adaptive $ε$-selection protocol leveraging TOPSIS multi-criteria decision analysis to align citizen preferences with differential privacy (DP) parameters, (2) an explainable noise-injection framework featuring real-time Mean Absolute Error (MAE) visualizations and GPT-4-powered impact analysis, and (3) an integrated legal-compliance mechanism that dynamically modulates privacy budgets based on evolving regulatory constraints. Our results advance participatory AI practices by demonstrating how conversational interfaces can enhance public engagement in algorithmic privacy mechanisms, ensuring that privacy-preserving AI in public sector governance remains both mathematically robust and democratically accountable.

Figures (3)

• Figure 1: System architecture showing the participatory DP workflow
• Figure 2: Simulated $\epsilon$ selection under different preference profiles
• Figure 3: GPT-4 powered impact analysis