Requirements for trustworthy AI-enabled automated decision-making in the public sector: A systematic review
Bibliographic record
Abstract
With AI adoption for decision-making in the public sector projected to rise with profound socio-ethical impacts, the need to ensure its trustworthy use continues to attract research attention. We analyze the existing body of evidence and establish trustworthiness requirements for AI-enabled automated decision-making (ADM) in the public sector, identifying eighteen aggregate facets. We link these facets to dimensions of trust in automation and institution-based trust to develop a theory-oriented research framework. We further map them to the OECD AI system lifecycle, creating a practice-focused framework. Our study has theoretical, practical and policy implications. First, we extend the theory on technological trust. We also contribute to trustworthy AI literature, shedding light on relatively well-known requirements like accountability and transparency and revealing novel ones like context sensitivity, feedback and policy learning. Second, we provide a roadmap for public managers and developers to improve ADM governance practices along the AI lifecycle. Third, we offer policymakers a basis for evaluating possible gaps in current AI policies. Overall, our findings present opportunities for further research and offer some guidance on how to navigate the multi-dimensional challenges of designing, developing and implementing ADM for improved trustworthiness and greater public trust. • We determined 18 aggregate trustworthiness requirements for AI-enabled ADM. • These requirements were mapped to six trust dimensions to create a theory-oriented research framework. • For practice, we mapped the requirements to the OECD AI lifecycle. • Context sensitivity was found to be key to the trustworthiness of AI-enabled ADM. • AI algorithms require multiple forms of accountability to be trustworthy
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 | 0.057 | 0.240 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.005 | 0.007 |
| Bibliometrics | 0.022 | 0.021 |
| Science and technology studies | 0.002 | 0.003 |
| Scholarly communication | 0.007 | 0.008 |
| Open science | 0.003 | 0.004 |
| Research integrity | 0.002 | 0.003 |
| Insufficient payload (model declined to judge) | 0.003 | 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".