Researchers argue for technical precision in AI classifications within public administration. A proposed typology could enhance clarity and accountability.
Artificial intelligence permeates the public sector. Yet, research often lumps AI systems into a single, oversimplified category. The paper's key contribution: it highlights the need for nuanced technical distinctions when evaluating AI's impact on public values like accountability and justice.
Five Categories of AI Systems #
The authors introduce a compelling typology, splitting AI into five distinct categories: hand-coded, glass-box, black-box, general-purpose, and agentic systems. Crucially, each category has unique implications for public values. This builds on prior work from the field, urging a move away from one-size-fits-all approaches.
Why does this matter? Because different AI systems impact core public values differently. A hand-coded system might enhance accountability, while a black-box system could obscure it. The authors argue that ignoring these distinctions leads to imprecision and, ultimately, flawed policies.
Evaluating Current Research #
The study evaluates 91 highly-cited papers from 2019 to 2025. Shockingly, 55% of these papers don't specify the AI system type they're discussing. Even more concerning, 31% use a different system for motivation than the one they study. And 41% make conclusions that their studied system can't support.
This imprecision isn't just academic nitpicking. It's a real barrier to effective public administration. Without clarity, how can policymakers make informed decisions about AI's role in governance?
Practical Recommendations #
To address these issues, the authors provide practical recommendations. They suggest that researchers should always specify their AI system type in their studies. They've even developed a guide with diagnostic questions that don't require specialist knowledge.
But will researchers heed this advice? It's a question that looms large over the field. Until there's a shift towards greater technical precision, the risk of misunderstanding AI's impact remains significant.
Get AI news in your inbox
Daily digest of what matters in AI.