Generating Intuitive Fairness Specifications for Natural Language Processing

Florian E. Dorner, Momchil Peychev, Nikola Konstantinov, Naman Goel, Elliott Ash, Martin Vechev

Keywords: Individual Fairness, Style Transfer, NLP, Crowdsourcing, Human Evaluation

TL;DR: We provide new methods for generating individual fairness specifications for NLP based on LLMs and validate them in a human study.

Abstract: Text classifiers have promising applications in high-stake tasks such as resume screening and content moderation. These classifiers must be fair and avoid discriminatory decisions by being invariant to perturbations of sensitive attributes such as gender or ethnicity. However, there is a gap between human intuition about these perturbations and the formal similarity specifications capturing them. While existing research has started to address this gap, current methods are based on hardcoded word replacements, resulting in specifications with limited expressivity or ones that fail to fully align with human intuition (e.g., in cases of asymmetric counterfactuals). This work proposes novel methods for bridging this gap by discovering expressive and intuitive individual fairness specifications. We show how to leverage unsupervised style transfer and GPT-3's zero-shot capabilities to automatically generate expressive candidate pairs of semantically similar sentences that differ along sensitive attributes. We then validate the generated pairs via an extensive crowdsourcing study, which confirms that a lot of these pairs align with human intuition about fairness in toxicity classification. We also show how limited amounts of human feedback can be leveraged to learn a similarity specification.