How Gender Interacts with Political Values: A Case Study on Czech BERT Models
This addresses concerns about bias in language models for Czech language applications, but it is incremental as it focuses on a specific language and model type.
The study tackled the problem of political biases in pre-trained Czech BERT models by comparing them with a representative value survey, finding that the models do not assign statement probability based on value-driven reasoning and show no systematic gender differences.
Neural language models, which reach state-of-the-art results on most natural language processing tasks, are trained on large text corpora that inevitably contain value-burdened content and often capture undesirable biases, which the models reflect. This case study focuses on the political biases of pre-trained encoders in Czech and compares them with a representative value survey. Because Czech is a gendered language, we also measure how the grammatical gender coincides with responses to men and women in the survey. We introduce a novel method for measuring the model's perceived political values. We find that the models do not assign statement probability following value-driven reasoning, and there is no systematic difference between feminine and masculine sentences. We conclude that BERT-sized models do not manifest systematic alignment with political values and that the biases observed in the models are rather due to superficial imitation of training data patterns than systematic value beliefs encoded in the models.