Explaining Text Classifiers with Counterfactual Representations
This work addresses the problem of interpretability and bias mitigation in text classification for AI practitioners, offering a novel approach but with incremental improvements over existing counterfactual methods.
The paper tackles the challenge of generating plausible counterfactual explanations for text classifiers by proposing a method that intervenes in text representation space, bypassing limitations of direct text manipulation, and validates it through experiments on synthetic and realistic datasets, showing alignment with ground truth counterfactuals.
One well motivated explanation method for classifiers leverages counterfactuals which are hypothetical events identical to real observations in all aspects except for one feature. Constructing such counterfactual poses specific challenges for texts, however, as some attribute values may not necessarily align with plausible real-world events. In this paper we propose a simple method for generating counterfactuals by intervening in the space of text representations which bypasses this limitation. We argue that our interventions are minimally disruptive and that they are theoretically sound as they align with counterfactuals as defined in Pearl's causal inference framework. To validate our method, we conducted experiments first on a synthetic dataset and then on a realistic dataset of counterfactuals. This allows for a direct comparison between classifier predictions based on ground truth counterfactuals - obtained through explicit text interventions - and our counterfactuals, derived through interventions in the representation space. Eventually, we study a real world scenario where our counterfactuals can be leveraged both for explaining a classifier and for bias mitigation.