Domain Faithfulness through Counterfactually Robust Learning

Ananth Balashankar, Ankit Bhardwaj, Neelabh Madan, Thomas Wies, Lakshmi Subramanian

Keywords: domain faithful, counterfactual data augmentation, robustness

Abstract: Domain-specific knowledge can often be expressed as suggestive rules defined over subgroups of data. Such rules, when encoded as hard constraints, are often not directly compatible with deep learning frameworks that train neural networks over batches of data. Also, domain-experts often use heuristics that cannot be encoded as logical rules. In this work, we propose a method where domain-experts' knowledge expressed as domain-specific rules over subgroups of data is leveraged in training domain faithful deep learning models using the modular components of counterfactual data augmentation, concept-based robust regularization, and parameter optimization. This translation of domain knowledge into custom primitives that can be augmented to existing state-of-the-art deep learning models improves the ability of domain experts to faithfully interpret and express model behavior, intervene through changes in the modeling specifications, and improve the overall performance of the model as compared to existing frameworks that incorporate deterministic declarative predicates. On one synthetic and three real-world tasks, we show that our method allows iterative refinement and is demonstrably more accurate.

Submission Number: 78