Balancing Fairness and Accuracy in Sentiment Detection using Multiple Black Box Models

Almuzaini, Abdulaziz A.; Singh, Vivek K.

doi:10.1145/3422841.3423536

Cited by 6 publications

(1 citation statement)

References 21 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Inprocessing (also called in-training) approaches revise the training of the state-of-the-art models to achieve fairness. More specifically, they apply fairness constraints or design an objective function considering the fairness of predictions [18]. Such approaches assume that the sensitive attributes information are accessible in the training samples, and enforce fairness during the training process either by directly imposing fairness constraints and solving constrained optimization problems [19] or by adding penalization terms to the learning objective [20].…”

Section: Introductionmentioning

confidence: 99%

Reducing Bias in Sentiment Analysis Models Through Causal Mediation Analysis and Targeted Counterfactual Training

Da,

Bossa,

Berenguer

et al. 2024

IEEE Access

View full text Add to dashboard Cite

Large language models provide high-accuracy solutions in many natural language processing tasks. In particular, they are used as word embeddings in sentiment analysis models. However, these models pick up on and amplify biases and social stereotypes in the data. Causality theory has recently driven the development of effective algorithms to evaluate and mitigate these biases. Causal mediation was used to detect biases, while counterfactual training was proposed to mitigate bias. In both cases, counterfactual sentences are created by changing an attribute, such as the gender of a noun, for which no change in the model output is expected. Biases are detected and eventually corrected each time the model behaviour differs between the original and the counterfactual sentence. We propose a new method for de-biasing sentiment analysis models that leverages the causal mediation analysis to identify the parts of the model primarily responsible for the bias and apply targeted counterfactual training for model de-biasing. We validated the methodology by fine-tuning the pre-trained Bidirectional Encoder Representations from Transformers (BERT) model for sentiment prediction. We trained two sentiment analysis models using the Stanford Sentiment Treebank dataset and the Amazon Product Reviews, respectively, and we evaluated the fairness and prediction performances using the Equity Evaluation Corpus. We illustrated the causal patterns in the network and showed that our method achieves both high fairness and more accurate sentiment analysis than the stateof-the-art approach. Contrary to state-of-the-art models, we achieved a noticeable improvement in gender fairness without hindering sentiment prediction accuracy.

show abstract

Section: Introductionmentioning

confidence: 99%