Explainable AI policy: It is time to challenge post hoc explanations
Bibliographic record
Abstract
The focus of this paper is on policy guidance around explainable artificial intelligence (AI) or the ability to understand how AI models arrive at their outcomes. Explainability matters in human terms because it facilitates including an individual's "right to explanation" and it also plays a role in enabling technical evaluation of AI systems. The paper begins with an examination of the meaning of explainability, concluding that the constellation of related terms serves to frustrate and confuse policy initiatives. Following a brief review of contemporary policy guidance, it argues that there is a need for greater clarity and context-specific guidance, highlighting the need to distinguish between ante hoc and post hoc explainability, especially in high-risk, high-impact contexts. The question of whether post hoc or ante hoc methods have been employed is a fundamental and often-overlooked question in policy. The paper argues that the question of which method should be employed in a given context, along with the requirement for human-level understanding, is a key challenge that policy makers need to address. A taxonomy for how explainability can be operationalized in AI policy is proposed and a series of recommendations is set forth.
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 | 0.173 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.002 | 0.004 |
| Science and technology studies | 0.004 | 0.022 |
| Scholarly communication | 0.021 | 0.035 |
| Open science | 0.005 | 0.008 |
| Research integrity | 0.016 | 0.024 |
| Insufficient payload (model declined to judge) | 0.018 | 0.004 |
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".