Dataset Card for MFRC

Dataset Summary

Moral framing and sentiment can affect a variety of online and offline behaviors, including donation, environmental action, political engagement, and protest. Various computational methods in Natural Language Processing (NLP) have been used to detect moral sentiment from textual data, but achieving strong performance in such subjective tasks requires large, hand-annotated datasets. Previous corpora annotated for moral sentiment have proven valuable, and have generated new insights both within NLP and across the social sciences, but have been limited to Twitter. To facilitate improving our understanding of the role of moral rhetoric, we present the Moral Foundations Reddit Corpus, a collection of 16,123 English Reddit comments that have been curated from 12 distinct subreddits, hand-annotated by at least three trained annotators for 8 categories of moral sentiment (i.e., Care, Proportionality, Equality, Purity, Authority, Loyalty, Thin Morality, Implicit/Explicit Morality) based on the updated Moral Foundations Theory (MFT) framework.

Number of Reddit Posts Annotated by N Annotators for Each Subreddit:

subreddit	2	3	4	5
AmItheAsshole	3	1336	0	0
Conservative	22	1911	1	0
antiwork	31	1757	0	0
confession	6	1325	0	0
europe	15	2641	1	0
geopolitics	2	112	0	0
neoliberal	10	1660	3	2
nostalgia	6	1336	0	0
politics	17	1758	0	0
relationship_advice	4	1350	0	0
worldnews	19	2555	3	0

Table 2: Frequency of Reddit posts per Foundation Calculated Based on Annotators’ Majority Vote.

subreddit	Care	Equality	Proportionality	Loyalty	Authority	Purity	Thin Morality	Non-Moral
AmItheAsshole	371	138	79	78	98	49	389	676
Conservative	209	231	123	62	249	47	459	1362
antiwork	308	156	194	48	129	53	328	1264
confession	320	57	83	28	66	63	400	772
europe	106	194	88	137	330	34	590	2106
geopolitics	0	3	2	2	8	1	11	108
neoliberal	34	86	46	56	191	18	321	1422
nostalgia	29	13	13	10	15	24	130	1216
politics	153	132	91	60	249	37	409	1309
relationship_advice	424	102	52	112	51	75	369	700
worldnews	154	264	80	105	345	60	626	1929
Overall	2108	1376	851	698	1731	461	4032	12864

Languages

English

Data Fields

• text
• subreddit
• bucket
• annotator
• annotation
• confidence

Licensing Information

cc-by-4.0

Citation Information

@inproceedings{trager-etal-2026-moral,
  title = {The Moral Foundations Reddit Corpus},
  author = {Trager, Jackson P. and Ziabari, Alireza S. and Rahmati, Elnaz and Davani, Aida Mostafazadeh and Golazizian, Preni and Karimi-Malekabadi, Farzan and Omrani, Ali and Li, Zhihe and Kennedy, Brendan and Chochlakis, Georgios and Reimer, Nils Karl and Reyes, Melissa and Cheng, Kesley and Wei, Mellow and Merrifield, Christina and Khosravi, Arta and Alvarez, Evans and Dehghani, Morteza},
  booktitle = {Proceedings of the Fifteenth Language Resources and Evaluation Conference (LREC 2026)},
  month = {May},
  year = {2026},
  pages = {6383--6407},
  address = {Palma, Mallorca, Spain},
  publisher = {European Language Resources Association (ELRA)},
  editor = {Piperidis, Stelios and Bel, Núria and van den Heuvel, Henk and Ide, Nancy and Krek, Simon and Toral, Antonio},
  doi = {10.63317/2b6xmbq3kphf},
  abstract = {Moral framing and sentiment can affect a variety of online and offline behaviors, including donation, environmental action, political engagement, and protest. Various computational methods in Natural Language Processing (NLP) have been used to detect moral sentiment from textual data, but achieving strong performance in such subjective tasks requires large, hand-annotated datasets. Previous corpora annotated for moral sentiment have proven valuable and have generated new insights both within NLP and across the social sciences, but have been limited to Twitter. To facilitate improving our understanding of the role of moral rhetoric, we present the Moral Foundations Reddit Corpus, a collection of 16,123 English Reddit comments that have been curated from 12 distinct subreddits, hand-annotated by at least three trained annotators for 8 categories of moral sentiment (i.e., Care, Proportionality, Equality, Purity, Authority, Loyalty, Thin Morality, Implicit/Explicit Morality) based on the updated Moral Foundations Theory (MFT) framework. We evaluate baselines using large language models (Llama3-8B, Ministral-8B) in zero-shot, few-shot, and PEFT (Parameter-Efficient Fine-Tuning) settings, comparing their performance to fine-tuned encoder-only models like BERT (Bidirectional Encoder Representations from Transformers). The results show that LLMs continue to lag behind fine-tuned encoders on this subjective task, underscoring the ongoing need for human-annotated moral corpora for AI alignment evaluation}
}