Artificial intelligence (AI) in the public service: Public administrator perspectives on leveraging AI to support citizen and stakeholder engagement
Bibliographic record
Abstract
The rapid advance of digital technologies offers enhanced methods for engaging citizens and stakeholders in public policy formation. However, one downside of digital engagement is an increase in the volume of external feedback that can overwhelm public servants and political decision makers. One potential solution lies in Natural Language Processing (NLP), a branch of artificial intelligence (AI) that can automatically filter, parse, understand, interpret, and synthesize large amounts of human language. This research assesses the potential of AI for managing the anticipated flood of citizen and stakeholder inputs from digital engagement exercises, and if public administrators find such tools valuable. Twelve Canadian public administrators from across three provinces, all having experience in citizen engagement, participated in semi-structured interviews. The findings reveal that public servants see potential benefits in the careful application of AI to assessing citizen and stakeholder engagement feedback by adjusting the workload of analyzing public input to focus more on meaningful public interaction and service improvement. However, significant concerns were expressed around data privacy, data security, algorithmic bias, and the reliability of outputs from machine analysis. A lack of AI literacy, risk-averse public sector cultures, and dated IT infrastructures were seen as key barriers. Participants emphasized: the need for continued human oversight and judgement in interpreting AI-generated assessments; engaging citizens and stakeholder, and politicians, in shaping AI deployment to strengthen trust and legitimacy; and partnering with technology providers and experts to develop AI solutions that align with public sector values and priorities. Keywords: citizen engagement, public participation, public consultation, artificial intelligence (AI), natural language processing (NLP), generative AI (GenAI), data privacy, algorithmic bias, public trust, human-AI collaboration
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.027 | 0.030 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.003 | 0.005 |
| Science and technology studies | 0.049 | 0.070 |
| Scholarly communication | 0.033 | 0.012 |
| Open science | 0.003 | 0.014 |
| Research integrity | 0.007 | 0.015 |
| Insufficient payload (model declined to judge) | 0.004 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".