0 cumulative citations
View corpus contextAI replies to citizens raise reported satisfaction, but benefits are content-dependent: consultations see the biggest gains while suggestions fare worst; perceived fairness explains the boost and prior participation shapes who benefits.
The Right Key for the Right Lock: How Government Response Methods and Demand Types Affect Citizen Satisfaction
openalex
rct
medium evidence
7/10 relevance
Summary only
summary available; pdf_status=paywall
DOI
Source
PDF
In a randomized survey experiment, AI-enabled government responses increase citizen satisfaction overall, with the largest gains for consultation requests and smaller or negligible gains for suggestions, and these effects operate via perceived fairness and vary by citizens' prior participation experience.
Citation observations
Cumulative provider counts captured on specific dates; providers are never combined.
ABSTRACT Faced with diverse and complex demands from citizens, the government has consistently sought ways to enhance its ability to respond. Metaphorically, citizens' needs can be viewed as locks, and the government requires keys to unlock them. Artificial intelligence (AI) is an innovative new key offering a fresh paradigm for unlocking the door to high citizen satisfaction. Drawing on government responsiveness theory, this study applies a 2 × 3 between‐subjects survey experiment to explore the impact of government response methods (traditional responses, AI‐enabled responses) and citizen demand types (consultation, complaint/help‐seeking, suggestions) on citizen satisfaction. Findings show that AI‐enabled government responses significantly increase citizen satisfaction compared to traditional approaches. Among demand types, consultation yields the highest satisfaction, while suggestion results in the lowest. Further, the satisfaction level is significantly influenced by the interaction effect between response method and demand type. Perceived fairness mediates the relationship between government responses and citizen satisfaction, with citizen participation experience having a moderating effect. This study offers new insights into increasing citizen satisfaction from the perspective of government responses. It reveals the content‐dependent nature of AI‐enabled government responses, helping the government create unique AI response strategies and enhance digital interaction with citizens.
Summary
Main Finding
AI-enabled government responses produce significantly higher citizen satisfaction than traditional responses. However, this effect depends on the type of citizen demand: consultation requests evoke the highest satisfaction, suggestions the lowest, and there is a significant interaction between response method and demand type. Perceived fairness mediates the effect of response method on satisfaction, and citizens’ prior participation experience moderates these relationships.
Key Points
- Experimental design: 2 (response method: traditional vs AI-enabled) × 3 (demand type: consultation, complaint/help-seeking, suggestion) between-subjects survey experiment.
- Overall, AI-enabled responses outperform traditional responses on citizen satisfaction.
- Demand-type heterogeneity:
- Consultation demands → highest satisfaction.
- Complaint/help-seeking → intermediate satisfaction.
- Suggestions → lowest satisfaction.
- Interaction effect: the advantage of AI responses varies by demand type (i.e., AI is not uniformly superior across all content).
- Mechanism: perceived fairness mediates the link between government response method and satisfaction.
- Moderator: citizens’ prior participation experience alters the strength/direction of effects (i.e., the impact of AI responses and perceived fairness depends on participation history).
Data & Methods
- Methodology: Between-subjects 2×3 survey experiment manipulating (a) government response method and (b) type of citizen demand.
- Outcomes measured: citizen satisfaction (primary), perceived fairness (mediator), and citizen participation experience (moderator).
- Statistical approach: tests for main effects, interaction effects, mediation (perceived fairness), and moderation (participation experience).
- Notes on scope: results are based on experimental survey data (internal validity strong for causal claims within the experiment; external validity to field settings may require further corroboration).
Implications for AI Economics
- Heterogeneous returns to AI deployment: value of AI in public service is content-dependent—investing in AI will yield higher welfare/satisfaction returns for some request types (e.g., consultations) than others (e.g., suggestions).
- Resource allocation and targeting: governments should prioritize AI tools for demand types where marginal gains in satisfaction and perceived fairness are largest, rather than universal rollout.
- Design priorities: because perceived fairness mediates effects, economic models and procurement should incorporate fairness-enhancing design (transparency, explainability, appeals processes) to maximize social returns.
- Role of user experience and prior engagement: citizens’ prior participation experience moderates impact—targeted outreach or training may increase the effectiveness of AI interventions and increase realized returns.
- Measurement for cost–benefit analysis: satisfaction and perceived fairness are key outcome metrics to include when evaluating public-sector AI investments; heterogeneous effects imply benefit estimates should be disaggregated by demand type.
- Policy/regulatory considerations: since AI has content-dependent effects and fairness is central, regulation should focus on governance standards that ensure equitable and perceived-fair AI-mediated interactions to avoid uneven welfare outcomes.
Assessment
Paper Typerct
Evidence Strengthmedium — Internal causal identification is strong due to random assignment and explicit tests for main, interaction, mediation, and moderation effects, but evidence is limited by being a survey experiment with hypothetical or vignette-style interactions, unspecified sample details, and unknown external/long-run validity.
Methods Rigormedium — Design is appropriate for causal tests and includes mediation and moderation analyses, but the supplied text omits key methodological details (sample size, sampling frame, pre-registration, manipulation checks, balance tests, effect sizes and robustness checks), and external validity to real-world public-service deployments is uncertain.
SampleBetween-subjects online survey experiment with participants randomized to response method (AI-enabled vs traditional) and to one of three citizen demand vignettes (consultation, complaint/help-seeking, suggestion); outcomes measured: citizen satisfaction (primary), perceived fairness (mediator), and respondents' prior participation experience (moderator). (Sample size, sampling frame, and country/context not specified in supplied text.)
Themesgovernance human_ai_collab
IdentificationRandomized between-subjects 2×3 survey experiment: participants randomly assigned to response method (traditional vs AI-enabled) and demand type (consultation, complaint/help-seeking, suggestion), enabling causal inference about the effect of response method and its interaction with demand type within the experimental setting.
GeneralizabilitySurvey experiment (vignette/hypothetical) — may not reflect actual behavior or reactions in real-world interactions, Sampling frame unspecified (likely online panel) — may not be representative of target population or diverse civic contexts, Context- and country-specific institutional factors not described — results may not generalize across governance systems, Short-term reported satisfaction and perceived fairness measured; longer-run outcomes (trust, compliance, repeated interactions) not observed, AI system specifics (capabilities, transparency, error rates) not detailed — effects may vary with real-world AI design and performance
Claims (6)
| Claim | Direction | Outcome | Confidence & Evidence | Details |
|---|---|---|---|---|
| AI-enabled government responses produce higher citizen satisfaction than traditional government responses. Consumer Welfare | positive | Citizen satisfaction with the government response |
Reading fidelity
high
Study strength
low
|
not reported
|
| Citizen satisfaction differs by demand type: consultation requests generate the highest satisfaction, complaint or help-seeking requests generate intermediate satisfaction, and suggestions generate the lowest satisfaction. Consumer Welfare | mixed | Citizen satisfaction by type of citizen demand |
Reading fidelity
high
Study strength
low
|
not reported
|
| There is a significant interaction between government response method and demand type, meaning that the satisfaction advantage of AI-enabled responses varies across types of citizen demand. Consumer Welfare | mixed | Citizen satisfaction as a function of response method and demand type |
Reading fidelity
high
Study strength
low
|
not reported
|
| Perceived fairness mediates the relationship between government response method and citizen satisfaction. Ai Safety And Ethics | positive | Citizen satisfaction through perceived fairness of the government response |
Reading fidelity
high
Study strength
low
|
not reported
|
| Citizens’ prior participation experience moderates the effects of response method and perceived fairness on citizen satisfaction. Consumer Welfare | mixed | Citizen satisfaction conditional on prior participation experience |
Reading fidelity
high
Study strength
low
|
not reported
|
| The benefits of deploying AI in public-service responses are heterogeneous across demand types rather than uniformly positive. Consumer Welfare | mixed | Variation in citizen-satisfaction gains from AI-enabled responses |
Reading fidelity
high
Study strength
low
|
not reported
|