CLIF: Concept-Level Influence Functions for Transparent Bottleneck Models
For practitioners needing interpretable NLP models in high-stakes domains, this work offers a method for transparent debugging and concept-level understanding, though it is incremental as it applies existing influence functions to bottleneck models.
The paper proposes using influence functions to enhance interpretability in NLP models at sample and concept levels. Experiments on CEBaB and Yelp datasets show that adjusting influential samples restores model performance to baseline without retraining, and concept-level analysis in Concept Bottleneck Models reveals key concepts affecting predictions.
In recent years, the black-box nature of deep learning models has limited their application in high-stakes domains such as medical diagnosis and finance, where interpretability is essential. To address this, we propose a novel approach using influence functions to enhance interpretability in NLP models at both the sample and concept levels. Experiments on CEBaB and Yelp datasets show that influence functions effectively identify the most impactful training samples, both helpful and harmful, on model predictions. By adjusting the labels and weights of these samples, we demonstrate that model performance can be restored to baseline levels without retraining, confirming the value of influence functions for efficient data debugging. Furthermore, our concept-level analysis identifies key concepts within Concept Bottleneck Models (CBM) that significantly affect predictions. Modifying these concepts alters model behavior observably, providing clear insights into the decision process.