Framework improves what if explanations by handling model uncertainty
AVCG: A Generalized Variational Framework for Counterfactual Generation under Hypothesis Distributions
Machine LearningArtificial Intelligence
Summary
When people ask 'what if' questions about data, like how changing something might affect an outcome, current methods usually assume one fixed prediction model. This can lead to explanations that stop working well if the model changes or has uncertainty. The authors created a new framework called AVCG that considers many possible models together instead of just one, making the explanations more reliable even when the model updates or has uncertainty. Tests show this method keeps explanations valid without slowing down the process.
Counterfactual explanationsPredictive uncertaintyBayesian posteriorRashomon setVariational optimizationAmortized inferenceModel retrainingHypothesis distributionMachine learning robustnessWhat-if analysis
Authors
Jamie Duell, Alejandro Jimenez Rodriguez, Mahault Albarracin
Abstract
Counterfactual explanations formalize "what-if" scenarios by identifying modifications to an input instance that obtain a desired alternative prediction. Traditionally, whether generated via instance-specific optimization or amortized single pass models, these approaches rely on a single, deterministic point-estimate predictor. However, this ignores predictive uncertainty and hypothesis variability, leading to brittle explanations that frequently become invalid if the underlying model is retrained or updated. To address this fragility, we propose the Amortized Variational Counterfactual Generator (AVCG), a generalized optimization framework that formulates counterfactual generation as optimization over an arbitrary distribution of plausible predictive hypotheses rather than a single deterministic predictor. This formulation naturally accommodates Bayesian posteriors, Rashomon-restricted hypothesis spaces, and other uncertainty representations within a unified optimization framework. Evaluation across multiple benchmark datasets demonstrates that the AVCG framework produces counterfactual explanations that remain highly valid under predictive uncertainty and model changes, while maintaining competitive plausibility and single-pass runtime performance.