A Practical Method for Generating String Counterfactuals
This provides a practical tool for researchers and practitioners to better understand and control bias in language models, though it is incremental as it builds on existing intervention methods.
The paper tackles the challenge of interpreting representation-space interventions in language models by converting them into string counterfactuals, enabling analysis of linguistic changes and bias mitigation through data augmentation.
Interventions targeting the representation space of language models (LMs) have emerged as an effective means to influence model behavior. Such methods are employed, for example, to eliminate or alter the encoding of demographic information such as gender within the model's representations and, in so doing, create a counterfactual representation. However, because the intervention operates within the representation space, understanding precisely what aspects of the text it modifies poses a challenge. In this paper, we give a method to convert representation counterfactuals into string counterfactuals. We demonstrate that this approach enables us to analyze the linguistic alterations corresponding to a given representation space intervention and to interpret the features utilized to encode a specific concept. Moreover, the resulting counterfactuals can be used to mitigate bias in classification through data augmentation.