Differential Privacy, Linguistic Fairness, and Training Data Influence: Impossibility and Possibility Theorems for Multilingual Language Models
This addresses the challenge of balancing privacy, fairness, and transparency in multilingual AI models, which is crucial for ethical deployment but incremental in scope.
The paper investigates whether multilingual language models can simultaneously satisfy differential privacy, linguistic fairness, and training data influence sparsity, finding that privacy and fairness are compatible but conflict with transparency. Experiments on NLP tasks show trade-offs, indicating a need for joint optimization approaches.
Language models such as mBERT, XLM-R, and BLOOM aim to achieve multilingual generalization or compression to facilitate transfer to a large number of (potentially unseen) languages. However, these models should ideally also be private, linguistically fair, and transparent, by relating their predictions to training data. Can these requirements be simultaneously satisfied? We show that multilingual compression and linguistic fairness are compatible with differential privacy, but that differential privacy is at odds with training data influence sparsity, an objective for transparency. We further present a series of experiments on two common NLP tasks and evaluate multilingual compression and training data influence sparsity under different privacy guarantees, exploring these trade-offs in more detail. Our results suggest that we need to develop ways to jointly optimize for these objectives in order to find practical trade-offs.