Is AI Training Ethical? Complicity, Job Loss and Fair Terms Across Five Disciplines
Description
People considering AI training work ask whether it is ethical, out of two worries: complicity in harms associated with AI, and helping to automate their own field. This normative essay reads both worries through moral philosophy, labour economics, anthropology and human-computer interaction research, the history of technology and political economy, with an exploratory case study of what the 4,353 listings live on one job board on October 5, 2026 tell applicants. The author runs that board, which leans toward expert, hourly roles.
Counted with published keyword rules, 65 of 4,335 listings with text name the lab or model the work trains, a quarter describe the client only as "a leading AI lab", and 24 of 131 red-teaming and harmful-content roles warn about disturbing material. The report argues that AI training work cannot be judged as a single occupation. Taking expert work is easier to defend when its purpose is disclosed, harmful content is optional and supported, and knowledge is handed over knowingly on terms that limit its reuse; it is harder to defend when the work is undisclosed, irreversible, tied to the worker's identity or aimed at automating their profession on poor terms. Fair pay makes the exchange fair to the worker while the hazard to the profession remains, which gives collective remedies a larger role than individual choice. Aggregate counts, the keyword rules and the script are released under CC BY 4.0.
Online edition: Is AI Training Ethical? Complicity, Job Loss and Fair Terms Across Five Disciplines.
More open data on AI training pay and hiring: AI training pay and hiring data.
Files
keyword_rules.json
Files
(382.3 kB)
| Name | Size | |
|---|---|---|
|
md5:bba922e9b76ee1a1b10a7f85de450a2f
|
8.1 kB | Download |
|
md5:37360aee069094b768fbe793fc5de249
|
2.8 kB | Preview Download |
|
md5:f02623aa91d943b65552ba885b0419cf
|
659 Bytes | Preview Download |
|
md5:274c479c1b10c3f09dd8980a7b9fc2a1
|
327 Bytes | Preview Download |
|
md5:909db865e8fa8e42f1be98085a0ed8ae
|
1.1 kB | Preview Download |
|
md5:c8ec4b97ce7398f0ce1d6ab02c84c2a5
|
369.3 kB | Preview Download |
Additional details
Related works
- Is variant form of
- Report: https://aitrainer.work/research/is-ai-training-ethical (URL)
References
- Acemoglu, D., & Restrepo, P. (2019). Automation and new tasks: How technology displaces and reinstates labor. Journal of Economic Perspectives, 33(2), 3–30. https://doi.org/10.1257/jep.33.2.3
- Ashforth, B. E., & Kreiner, G. E. (1999). "How can you do it?": Dirty work and the challenge of constructing a positive identity. Academy of Management Review, 24(3), 413–434. https://doi.org/10.5465/AMR.1999.2202129
- Irani, L. (2015). The cultural work of microwork. New Media & Society, 17(5), 720–739. https://doi.org/10.1177/1461444813511926
- Kagan, S. (2011). Do I make a difference? Philosophy & Public Affairs, 39(2), 105–141. https://doi.org/10.1111/j.1088-4963.2011.01203.x
- Miceli, M., & Posada, J. (2022). The data-production dispositif. Proceedings of the ACM on Human-Computer Interaction, 6(CSCW2). https://doi.org/10.1145/3555561
- Nefsky, J. (2017). How you can help, without making a difference. Philosophical Studies, 174, 2743–2767. https://doi.org/10.1007/s11098-016-0808-y
- Tubaro, P., Casilli, A. A., & Coville, M. (2020). The trainer, the verifier, the imitator: Three ways in which human platform workers support artificial intelligence. Big Data & Society, 7(1). https://doi.org/10.1177/2053951720919776
- Wang, D., Prabhat, S., & Sambasivan, N. (2022). Whose AI dream? In search of the aspiration in data annotation. In CHI '22. https://doi.org/10.1145/3491102.3502121