Ethics dumping in artificial intelligence
Bibliographic record
Abstract
Artificial Intelligence (AI) systems encode not just statistical models and complex algorithms designed to process and analyze data, but also significant normative baggage. This ethical dimension, derived from the underlying code and training data, shapes the recommendations given, behaviors exhibited, and perceptions had by AI. These factors influence how AI is regulated, used, misused, and impacts end-users. The multifaceted nature of AI's influence has sparked extensive discussions across disciplines like Science and Technology Studies (STS), Ethical, Legal and Social Implications (ELSI) studies, public policy analysis, and responsible innovation-underscoring the need to examine AI's ethical ramifications. While the initial wave of AI ethics focused on articulating principles and guidelines, recent scholarship increasingly emphasizes the practical implementation of ethical principles, regulatory oversight, and mitigating unforeseen negative consequences. Drawing from the concept of "ethics dumping" in research ethics, this paper argues that practices surrounding AI development and deployment can, unduly and in a very concerning way, offload ethical responsibilities from developers and regulators to ill-equipped users and host environments. Four key trends illustrating such ethics dumping are identified: (1) AI developers embedding ethics through coded value assumptions, (2) AI ethics guidelines promoting broad or unactionable principles disconnected from local contexts, (3) institutions implementing AI systems without evaluating ethical implications, and (4) decision-makers enacting ethical governance frameworks disconnected from practice. Mitigating AI ethics dumping requires empowering users, fostering stakeholder engagement in norm-setting, harmonizing ethical guidelines while allowing flexibility for local variation, and establishing clear accountability mechanisms across the AI ecosystem.
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 | 0.078 | 0.107 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.003 | 0.003 |
| Science and technology studies | 0.012 | 0.121 |
| Scholarly communication | 0.017 | 0.017 |
| Open science | 0.003 | 0.014 |
| Research integrity | 0.007 | 0.015 |
| Insufficient payload (model declined to judge) | 0.002 | 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".