AI-enhanced Public Policy Feedback Systems for Citizen Trust and Accountability

Thesis Abstract

The study investigates how AI-enhanced public policy feedback systems influence citizen trust and accountability in governance, addressing concerns that traditional feedback mechanisms fail to capture timely, representative, and actionable public input in complex policy environments. The aim is to design, implement, and evaluate an integrated AI-driven feedback platform that synthesizes citizen input from diverse channels, provides transparent policy impact signals, and supports policymakers with timely, interpretable insights. Specific objectives include (1) operationalizing a multi-channel feedback architecture that aggregates data from social media, official portals, town hall transcripts, and mobile surveys; (2) developing an AI-assisted triage and summarization module to extract policy-relevant concerns and prioritize issues for governmental consideration; (3) assessing the platform’s effect on perceived legitimacy, trust, and accountability among citizens through experimental and quasi-experimental designs; (4) evaluating policy-maker adoption, decision quality, and responsiveness facilitated by the system; and (5) formulating governance guidelines for algorithmic transparency, data privacy, and bias mitigation. A mixed-methods approach is employed, combining quantitative surveys, experimental vignette studies, and qualitative interviews. The population comprises adult citizens in three metropolitan regions with varying governance maturity. A stratified random sample of 1,200 respondents for baseline survey and 600 for follow-up will be drawn, alongside 30 policymakers participating in semi-structured interviews and 12 policy teams engaging in real-time platform trials. Data collection instruments include a standardized citizen trust questionnaire, legitimacy-scale items, and perceived accountability measures, complemented by system usage logs and sentiment indicators extracted from the AI platform. The AI components integrate natural language processing for issue extraction, topic modeling (LDA) for thematic clustering, and explainable AI (XAI) modules to generate interpretable policy impact narratives. The triangulated analysis employs regression analysis to test hypotheses linking platform usage and trust outcomes; difference-in-differences (DiD) to identify pre/post effects across regions; and structural equation modeling (SEM) to assess the relationships among trust, perceived accountability, and policy acceptance. Qualitative data will be analyzed via thematic analysis to elucidate mechanisms of trust formation and perceived fairness, with coding reliability established through inter-coder agreement (Cohen’s kappa > 0.8). Key expected findings include (1) higher levels of perceived public legitimacy and trust among citizens who actively engage with the AI-enhanced feedback platform, compared to control groups; (2) improved policy-maker perceptions of input quality and timeliness, evidenced by faster issue triage and more precise policy adjustments; (3) evidence that transparent disclosure of AI-derived narratives and decision rationales enhances perceived accountability and reduces concerns about algorithmic opacity; (4) identification of demographic and socio-political moderators (e.g., education, urban/rural residence, prior trust in institutions) influencing platform effectiveness; (5) practical guidelines for governance, including data governance models, bias mitigation strategies, and user-centered design principles. The study contributes to knowledge by integrating AI-enabled feedback mechanisms with democratic accountability theories, extending the literature on digital governance, deliberative democracy, and technology-mediated citizen engagement. It advances theoretical understanding of how algorithmic transparency and explainability affect trust dynamics and governance legitimacy, while offering empirical evidence on the conditions under which AI-assisted feedback systems enhance policy responsiveness and citizen satisfaction. Policy implications emphasize scalable, ethically sound designs that balance citizen participation with privacy protections and non-discriminatory outcomes. The main conclusion anticipates that well-designed AI-enhanced feedback systems can augment citizen trust and accountability when combined with robust governance practices, transparent disclosure of AI processes, continuous stakeholder involvement, and rigorous evaluation. Recommendations include institutionalizing continuous monitoring of algorithmic impact, adopting open data and audit trails, investing in user education about AI capabilities and limitations, and fostering iterative co-creation between citizens and policymakers to sustain legitimacy and responsive governance.