Neural additive model (NAM) is a recently proposed explainable artificial intelligence (XAI) method that utilizes neural network-based architectures. Given the advantages of neural networks, NAMs provide intuitive explanations for their predictions with high model performance. In this paper, we analyze a critical yet overlooked phenomenon: NAMs often produce inconsistent explanations, even when using the same architecture and dataset. Traditionally, such inconsistencies have been viewed as issues to be resolved. However, we argue instead that these inconsistencies can provide valuable explanations within the given data model. Through a simple theoretical framework, we demonstrate that these inconsistencies are not mere artifacts but emerge naturally in datasets with multiple important features. To effectively leverage this information, we introduce a novel framework, Bayesian Neural Additive Model (BayesNAM), which integrates Bayesian neural networks and feature dropout, with theoretical proof demonstrating that feature dropout effectively captures model inconsistencies. Our experiments demonstrate that BayesNAM effectively reveals potential problems such as insufficient data or structural limitations of the model, providing more reliable explanations and potential remedies.

Explanations of machine learning models are important, especially in scientific areas such as chemistry, biology, and physics, where they guide future laboratory experiments and resource requirements. These explanations can be derived from well-trained machine learning models (data-driven perspective) or specific domain knowledge (domain-driven perspective). However, there exist inconsistencies between these perspectives due to accurate yet misleading machine learning models and various stakeholders with specific needs, wants, or aims. This paper calls attention to these inconsistencies and suggests a way to find an accurate model with expected explanations that reinforce physical laws and meet stakeholders' requirements from a set of equally-good models, also known as Rashomon sets. Our goal is to foster a comprehensive understanding of these inconsistencies and ultimately contribute to the integration of eXplainable Artificial Intelligence (XAI) into scientific domains.