As governments around the world rapidly experiment with integrating artificial intelligence into public administration, a critical question remains: how do everyday citizens feel about public agencies using these powerful tools?
This central governance challenge took center stage at the 27th Annual International Conference on Digital Government Research (dg.o 2026). Hosted by the University of Nebraska at Omaha from June 1-4, the forum brought together leading global scholars to explore the intersection of technology innovation, civic engagement, and digital government.
Among the highlighted research was a multinational study presented by Dr. Sunny Nillasithanukroh, Assistant Professor at the Clinton School of Public Service. The paper, titled “Commercial Generative AI Exposure and Public Acceptance of Government AI Use,” explores how citizens’ personal experiences with everyday AI tools, like ChatGPT, shape their willingness to support AI deployment in the public sector.
Bridging Private Experience and Public Value
The study leverages data from the Global Public Opinion on Artificial Intelligence (GPO-AI) survey, analyzing responses from over 23,000 participants across 21 countries. Dr. Nillasithanukroh co-authored the paper alongside Dr. Robert Richards Jr., Associate Professor of Communication, Dr. Andreas Sihotang, Assistant Professor, and Dr. Chul Hyun Park, Senior Research Fellow.
The core findings reveal a powerful trend: more frequent personal use of commercial generative AI is consistently associated with higher public acceptance of government AI utilization.
According to Dr. Nillasithanukroh, interacting with tools like ChatGPT provides tangible reference points that demystify the technology.
“Through repeated interaction, citizens directly observe how AI systems perform tasks, respond to prompts, and handle ambiguity,” Nillasithanukroh said. “Everyday AI use may help citizens imagine efficiency gains in routine public services. This increases public acceptance by making the benefits of government AI feel more concrete and credible.”
Where Citizens Draw the Line: Routine vs. High-Discretion
A key contribution of the study is its distinction between different types of government AI applications. The researchers categorized government AI tasks into two primary buckets:
- Administrative and Support-Oriented Functions: Low-stakes tasks focused on operational support, such as enrolling people in social welfare programs, matching job seekers with available jobs, and notifying citizens to provide documents.
- High-Discretion and Enforcement-Oriented Functions: High-stakes context applications tied to authoritative determinations, such as screening people at borders, monitoring social media for public safety, or determining eligibility for visas.
The analysis found that while personal AI experience boosts acceptance across the board, the connection is stronger for routine administrative tasks than for high-discretion, outcome-determinative enforcement functions.
Perceived Benefits Drive Support, But Risks Remain Institutional
Interestingly, the study’s mediation analysis showed that increased public acceptance is driven primarily by an understanding of AI’s functional benefits—namely, its speed and efficiency—rather than a change in how citizens perceive risk.
While users may be aware of technical limitations like “hallucinations,” individual interaction with commercial tools does not naturally teach citizens about broader public sector risks, such as algorithmic bias, surveillance, or institutional accountability failures.
Because of this, Nillasithanukroh stresses that governments cannot rely on the public to guard themselves or assume that familiarity equals a comprehensive understanding of potential societal harms.
“Governments should not assume that public familiarity with ChatGPT or other commercial AI tools will automatically produce informed concern about bias, surveillance, accountability, or unequal impacts,” Nillasithanukroh cautioned. “Risk awareness must be supported through explicit communication and governance.”
Implications for Public Service Leaders
The Clinton School-led research offers actionable blueprints for public administrators navigating the digital transition.
First, to build authentic public trust, governments should pivot toward low-stakes pilot programs and sandbox environments. Allowing citizens to directly interact with public sector AI tools in controlled settings can make the use of AI feel transparent and practical.
Second, communication must be tailored to the stakes of the technology. For routine operations, highlighting performance gains may suffice. However, for enforcement and security operations, public agencies must move beyond standard efficiency arguments to clearly demonstrate how human oversight is maintained, how errors are checked, and how fairness is guaranteed.
Ultimately, the research demonstrates that building long-term, citizen-centric digital governments requires a commitment to technical transparency and robust regulatory safeguards.