Interpretability-by-Design with Accurate Locally Additive Models and Conditional Feature Effects
Abstract
Generalized additive models (GAMs) offer interpretability through independent univariate feature effects but underfit when interactions are present in data. GAMs add selected pairwise interactions which improves accuracy, but sacrifices interpretability and limits model auditing. We propose \emph{Conditionally Additive Local Models} (CALMs), a new model class, that balances the interpretability of GAMs with the accuracy of GAMs. CALMs allow multiple univariate shape functions per feature, each active in different regions of the input space. These regions are defined independently for each feature as simple logical conditions (thresholds) on the features it interacts with. As a result, effects remain locally additive while varying across subregions to capture interactions. We further propose a principled distillation-based training pipeline that identifies homogeneous regions with limited interactions and fits interpretable shape functions via region-aware backfitting. Experiments on diverse classification and regression tasks show that CALMs consistently outperform GAMs and achieve accuracy comparable with GAMs. Overall, CALMs offer a compelling trade-off between predictive accuracy and interpretability.
Cite
@article{arxiv.2602.16503,
title = {Interpretability-by-Design with Accurate Locally Additive Models and Conditional Feature Effects},
author = {Vasilis Gkolemis and Loukas Kavouras and Dimitrios Kyriakopoulos and Konstantinos Tsopelas and Dimitrios Rontogiannis and Giuseppe Casalicchio and Theodore Dalamagas and Christos Diou},
journal= {arXiv preprint arXiv:2602.16503},
year = {2026}
}