ExplainabilityTrustworthinessAccountabilitycs.LG
Beyond single-model XAI: aggregating multi-model explanations for enhanced trustworthiness
Ilaria Vascotto, Alex Rodriguez, Alessandro Bonaita, Luca Bortolussi
October 13, 2025
TRUST-AI@ECAI
Key Takeaways
-
Single explanations may be unreliable: The paper demonstrates that relying on explanations from just one AI model can be problematic, as individual explanation methods lack robustness—meaning they may not consistently identify the true reasons behind a model's decisions.
-
Combining multiple models improves trustworthiness: By aggregating feature importance explanations from three different types of models (k-nearest neighbours, random forest, and neural networks), the researchers found that this multi-model approach increases overall system trustworthiness compared to single-model explanations.
-
Robustness of explanations is often neglected: The paper highlights that while explainability has become important for high-risk AI applications, developers frequently overlook whether those explanations are actually robust and reliable—a critical gap for building genuine trust in AI systems.
-
Practical relevance for regulated AI use: As AI regulation increases globally, this work addresses a key legislative concern: ensuring that explanations provided to users and regulators are dependable enough to justify AI decisions in real-world, high-stakes contexts like healthcare or finance.
Abstract
The use of Artificial Intelligence (AI) models in real-world and high-risk applications has intensified the discussion about their trustworthiness and ethical usage, from both a technical and a legislative perspective. The field of eXplainable Artificial Intelligence (XAI) addresses this challenge by proposing explanations that bring to light the decision-making processes of complex black-box models. Despite being an essential property, the robustness of explanations is often an overlooked aspect during development: only robust explanation methods can increase the trust in the system as a whole. This paper investigates the role of robustness through the usage of a feature importance aggregation derived from multiple models ($k$-nearest neighbours, random forest and neural networks). Preliminary results showcase the potential in increasing the trustworthiness of the application, while leveraging multiple model's predictive power.