Análisis inmersivo para Inteligencia Artificial Explicable: un enfoque conceptual

Explainable artificial intelligence (XAI) addresses the "black box" problem in machine learning models by providing interpretable information about model decisions. Although visual explanations are a widely used strategy in XAI, they are often based on two-dimensional (2D) representations, which limits the understanding of complex, high-dimensional data. This paper proposes a conceptual approach that integrates immersive analytics with XAI through a visualization pipeline. This preliminary test demonstrates the feasibility of the proposed pipeline by using SHAP (SHapley Additive exPlanations) to generate local explanations, which are then transformed into interactive 3D visualizations within a virtual reality environment using a head-mounted display (HMD). The results show the effectiveness of the pipeline to generate immersive visualizations from previously calculated explanations. With the possibility of incorporating dynamic and interactive elements, transforming the representation of traditionally static visual explanations into new and more informative forms.