Towards Trustworthy Explanation: On Causal Rationalization
This addresses the issue of unreliable explanations in NLP for users needing trustworthy model interpretations, though it is incremental by applying causal inference to an existing framework.
The paper tackles the problem of spurious rationales in text rationalization by introducing causal desiderata to identify necessary and sufficient rationales, achieving superior performance on real-world datasets compared to state-of-the-art methods.
With recent advances in natural language processing, rationalization becomes an essential self-explaining diagram to disentangle the black box by selecting a subset of input texts to account for the major variation in prediction. Yet, existing association-based approaches on rationalization cannot identify true rationales when two or more snippets are highly inter-correlated and thus provide a similar contribution to prediction accuracy, so-called spuriousness. To address this limitation, we novelly leverage two causal desiderata, non-spuriousness and efficiency, into rationalization from the causal inference perspective. We formally define a series of probabilities of causation based on a newly proposed structural causal model of rationalization, with its theoretical identification established as the main component of learning necessary and sufficient rationales. The superior performance of the proposed causal rationalization is demonstrated on real-world review and medical datasets with extensive experiments compared to state-of-the-art methods.