Dataset Card for ETHICS-Animals

ETHICS-Animals is the subset of ETHICS in which a non-human animal, or something made from or done with animals, is part of the situation. Each row also has ethics_row, its index in the same ETHICS split.

	train	validation	test
commonsense	795	301	184
deontology	1,950	370	347
justice	1,766	206	159
utilitarianism	2,018	839	715
virtue	1,017	170	235

The splits follow the parquet conversion of ETHICS (refs/convert/parquet, revision 6deab74c): validation is the paper's test set and test is its hard test set. Commonsense keeps only items of at most 400 characters, as an approximation of ETHICS's short items.

How the rows were selected

1. Regex. ETHICS rows were matched against a broad regex of animal words and uses.
2. Verification. Claude Sonnet 5.5 read and verified each match, dropping any false positives.

Loading Data

from datasets import load_dataset
load_dataset("drwahl/ethics-animals", "commonsense")

Where commonsense is one of the following sections: commonsense, deontology, justice, utilitarianism, and virtue.

Citation Information

@article{hendrycks2021ethics,
  title={Aligning AI With Shared Human Values},
  author={Dan Hendrycks and Collin Burns and Steven Basart and Andrew Critch and Jerry Li and Dawn Song and Jacob Steinhardt},
  journal={Proceedings of the International Conference on Learning Representations (ICLR)},
  year={2021}
}

@misc{wahl2026ethicsanimals,
  title={ETHICS-Animals: the animal-relevant subset of ETHICS},
  author={Daniel Wahl},
  year={2026},
  howpublished={\url{https://huggingface.co/datasets/drwahl/ethics-animals}}
}