LIB-analysis

This repository contains the bias analysis methods based on the Linguistic Implicit Bias theory proposed in Tell-tale Signs of Implicit Bias: Language Abstraction for Automated Bias Analysis, and Implicit Bias in Peer Review: Through the Lens of Language Abstraction.

Usage

To use the abstraction scoring regression model proposed in Implicit Bias in Peer Review: Through the Lens of Language Abstraction, run the abstract_scoring_peer_review.ipynb jupyter notebook file. The data used for training and the experiments can be found from sources listed in the paper.

To use the abstraction span extraction model proposed in Tell-tale Signs of Implicit Bias: Language Abstraction for Automated Bias Analysis, run the abstract_instruct_span_exaction.ipynbfile on Kaggle notebook after acquiring access of the LLAMA-instruct model via Kaggle. The finetuning dataset is listed in the repository as wsj_data_300.json. The processed datasets used for experiments can be found in intergroup and POLUSA.

Citation

If you use the abstraction scoring regression model, please cite the paper - Implicit Bias in Peer Review: Through the Lens of Language Abstraction with the following:

@inproceedings{DBLP:conf/lrec/ZhangMC26,
  author       = {Xulang Zhang and
                  Rui Mao and
                  Erik Cambria},
  editor       = {Stelios Piperidis and
                  N{\'{u}}ria Bel and
                  Henk van den Heuvel and
                  Nancy Ide and
                  Simon Krek and
                  Antonio Toral},
  title        = {Implicit Bias in Peer Review: Through the Lens of Language Abstraction},
  booktitle    = {Proceedings of the Fifteenth Language Resources and Evaluation Conference,
                  {LREC} 2026, Palma de Mallorca, Spain, May 13-15, 2026},
  pages        = {9569--9580},
  publisher    = {{ELRA} Language Resource Association / {ACL}},
  year         = {2026},
  url          = {https://doi.org/10.63317/4uxwwzgswmqc},
  doi          = {10.63317/4UXWWZGSWMQC},
  timestamp    = {Tue, 11 Aug 2026 16:59:37 +0200},
  biburl       = {https://dblp.org/rec/conf/lrec/ZhangMC26.bib},
  bibsource    = {dblp computer science bibliography, https://dblp.org}
}

If you use the abstraction span extraction model, please cite the paper - Tell-tale Signs of Implicit Bias: Language Abstraction for Automated Bias Analysis with the following:

@inproceedings{zhang2026tell,
  title={Tell-tale Signs of Implicit Bias: Language Abstraction for Automated Bias Analysis},
  author={Zhang, Xulang and Mao, Rui and Ge, Mengshi and Cambria, Erik},
  booktitle={Proceedings of the Annual Meeting of the Cognitive Science Society},
  volume={48},
  year={2026}
}