Search papers, labs, and topics across Lattice.
This study investigates the Rashomon effect in decision systems by introducing a consistency criterion that combines ensemble margin and local prediction variability to audit incorrect predictions. The authors demonstrate that larger ensembles can effectively approximate the consistency scores of the expected model from the Rashomon set, leading to a significant reduction in unchecked incorrect predictions while only moderately increasing the number of instances requiring human review. Their experiments with transformer models in natural language understanding and fine-tuning of large language models for tabular data classification reveal that ensembling models from the Rashomon set enhances auditing accuracy compared to single models.
Ensembling models from the Rashomon set can drastically reduce the risk of unchecked incorrect predictions in decision systems while only slightly increasing human review demands.
The Rashomon effect is a machine learning phenomenon where equally accurate models produce different predictions for the same inputs (predictive multiplicity). Existing work primarily focuses on multiplicity within individual models, but in more complex decision systems, the impact of the Rashomon effect is less well understood. In this work, we study multiplicity from the perspective of auditing incorrect ensemble predictions, where the decision to divert an instance for human review is based on a consistency criterion that combines the ensemble margin with a measure of local prediction variability for each constituent model. With mild assumptions about stability and smoothness, we show that the consistency scores of finite ensembles converge to the corresponding consistency score of the expected model from the Rashomon set as the ensemble size and the number of samples used to measure local prediction variability increase. To demonstrate the efficacy of the proposed criterion, we evaluate the framework with respect to transformer models applied to natural language understanding tasks and parameter-efficient fine-tuning of large language models used for tabular data classification tasks. Our experiments show that ensembling models from the Rashomon set substantially reduces the risk of incorrect predictions going unchecked compared with auditing a single model, while incurring only a moderate increase in the number of diversions. Moreover, the auditing behavior of the full Rashomon set can be closely approximated by finite ensembles of relatively modest size, with the risk approaching zero for some datasets. We further demonstrate that the proposed measure exhibits stronger agreement with established predictive multiplicity metrics than existing consistency measures, providing a more reliable way to capture multiplicity in the Rashomon set.
