The emergence of artificial intelligence ethics auditing
Bibliographic record
Abstract
The emerging ecosystem of artificial intelligence (AI) ethics and governance auditing has grown rapidly in recent years in anticipation of impending regulatory efforts that encourage both internal and external auditing. Yet, there is limited understanding of this evolving landscape. We conduct an interview-based study of 34 individuals in the AI ethics auditing ecosystem across seven countries to examine the motivations, key auditing activities, and challenges associated with AI ethics auditing in the private sector. We find that AI ethics audits follow financial auditing stages, but tend to lack robust stakeholder involvement, measurement of success, and external reporting. Audits are hyper-focused on technically oriented AI ethics principles of bias, privacy, and explainability, to the exclusion of other principles and socio-technical approaches, reflecting a regulatory emphasis on technical risk management. Auditors face challenges, including competing demands across interdisciplinary functions, firm resource and staffing constraints, lack of technical and data infrastructure to enable auditing, and significant ambiguity in interpreting regulations and standards given limited (or absent) best practices and tractable regulatory guidance. Despite these roadblocks, AI ethics and governance auditors are playing a critical role in the early ecosystem: building auditing frameworks, interpreting regulations, curating practices, and sharing learnings with auditees, regulators, and other stakeholders.
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 | 0.077 | 0.113 |
| Meta-epidemiology (narrow) | 0.000 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.000 |
| Bibliometrics | 0.004 | 0.004 |
| Science and technology studies | 0.009 | 0.042 |
| Scholarly communication | 0.014 | 0.012 |
| Open science | 0.002 | 0.012 |
| Research integrity | 0.003 | 0.009 |
| Insufficient payload (model declined to judge) | 0.001 | 0.000 |
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".