2020

Fair Hate Speech Detection through Evaluation of Social Group Counterfactuals

Davani, Aida Mostafazadeh, Omrani, Ali, Kennedy, Brendan et al.

Understand

Approaches for mitigating bias in supervised models are designed to reduce models' dependence on specific sensitive features of the input data, e.g., mentioned social groups.

However, in the case of hate speech detection, it is not always desirable to equalize the effects of social groups because of their essential role in distinguishing outgroup-derogatory hate, such that particular types of hateful rhetoric carry the intended meaning only when contextualized around certain social group tokens.
Counterfactual token fairness for a mentioned social group evaluates the model's predictions as to whether they are the same for (a) the actual sentence and (b) a counterfactual instance, which is generated by changing the mentioned social group in the sentence.
Our approach assures robust model predictions for counterfactuals that imply similar meaning as the actual sentence.

Built on

Incorporating Priors with Feature Attribution on Text Classification
Original
Liu, F.; and Avci, B. 2019 · 1906
Earlier work this paper cites.
A survey on bias and fairness in machine learning
Original
Mehrabi, N.; Morstatter, F.; Saxena, N.; Lerman, K.; and Galstyan, A. 2019 · 1908
Earlier work this paper cites.
Bidirectional recurrent neural networks
Schuster, M.; and Paliwal, K. K. 1997 · 1997
Earlier work this paper cites.
A model of (often mixed) stereotype content: competence and warmth respectively follow from perceived status and competition
Fiske, S. T.; Cuddy, A. J.; Glick, P.; and Xu, J. 2002 · 2002
Earlier work this paper cites.
NLTK: the natural language toolkit
Loper, E.; and Bird, S. 2002 · 2002
Earlier work this paper cites.
StereoSet: Measuring stereotypical bias in pretrained language models
Original
Nadeem, M.; Bethke, A.; and Reddy, S. 2020 · 2004
Earlier work this paper cites.
Hate speech and stereotypic talk
Haas, J. 2012 · 2012
Earlier work this paper cites.
Detecting hate speech on the world wide web
Warner, W.; and Hirschberg, J. 2012 · 2012
Earlier work this paper cites.

Similar

Equality of opportunity in supervised learning
Hardt, M.; Price, E.; Srebro, N.; et al. 2016 · 2016
Cited alongside, same era.
Counterfactual fairness
Kusner, M. J.; Loftus, J.; Russell, C.; and Silva, R. 2017 · 2017
Cited alongside, same era.
Hate speech dataset from a white supremacy forum
Original
de Gibert, O.; Perez, N.; García-Pablos, A.; and Cuadros, M. 2018 · 2018
Cited alongside, same era.
Measuring and mitigating unintended bias in text classification
Dixon, L.; Li, J.; Sorensen, J.; Thain, N.; and Vasserman, L. 2018 · 2018
Cited alongside, same era.
Adversarial removal of demographic attributes from text data
Original
Elazar, Y.; and Goldberg, Y. 2018 · 2018
Cited alongside, same era.
Adversarial logit pairing
Original
Kannan, H.; Kurakin, A.; and Goodfellow, I. 2018 · 2018
Cited alongside, same era.
Learning adversarially fair and transferable representations
Original
Madras, D.; Creager, E.; Pitassi, T.; and Zemel, R. 2018 · 2018
Cited alongside, same era.
The Gab Hate Corpus: A collection of 27k posts annotated for hate speech
Kennedy, B.; Atari, M.; Davani, A. M.; Yeh, L.; Omrani, A.; Kim, Y.; Coombs Jr., K.; Havaldar, S.; Portillo-Wightman, G.; Gonzalez, E.; Hoover, J.; Azatian, A.; Cardenas, G.; Hussain, A.; Lara, A.; Omary, A.; Park, C.; Wang, X.; Wijaya, C.; Zhang, Y.; Meyerowitz, B.; and Dehghani, M. 2020a
Cited in the paper.
Contextualizing Hate Speech Classifiers with Post-hoc Explanation
Kennedy, B.; Jin, X.; Mostafazadeh Davani, A.; Dehghani, M.; and Ren, X. 2020b
Cited in the paper.

Then

Reducing gender bias in abusive language detection
Original
Park, J. H.; Shin, J.; and Fung, P. 2018 · 2018
Later among the works it cites.
Mitigating unwanted biases with adversarial learning
Zhang, B. H.; Lemoine, B.; and Mitchell, M. 2018 · 2018
Later among the works it cites.
Gender bias in coreference resolution: Evaluation and debiasing methods
Original
Zhao, J.; Wang, T.; Yatskar, M.; Ordonez, V.; and Chang, K.-W. 2018 · 2018
Later among the works it cites.
Pushshift Gab Corpus
Gaffney, G. 2018 · 2019
Later among the works it cites.
Counterfactual fairness in text classification through robustness
Garg, S.; Perot, V.; Limtiaco, N.; Taly, A.; Chi, E. H.; and Beutel, A. 2019 · 2019
Later among the works it cites.
Language models are unsupervised multitask learners
Radford, A.; Wu, J.; Child, R.; Luan, D.; Amodei, D.; and Sutskever, I. 2019 · 2019
Later among the works it cites.
Detection of Abusive Language: the Problem of Biased Datasets
Wiegand, M.; Ruppenhofer, J.; and Kleinbauer, T. 2019 · 2019
Later among the works it cites.

Beyond the bibliography

alphaXiv searches the wider corpus for related work and actual follow-ups.

Open on alphaXiv

alphaXiv is searching for related work…

Reading the bibliography…