Paper ID: 2306.05731

Multimodal Explainable Artificial Intelligence: A Comprehensive Review of Methodological Advances and Future Research Directions

Nikolaos Rodis, Christos Sardianos, Georgios Th. Papadopoulos, Panagiotis Radoglou-Grammatikis, Panagiotis Sarigiannidis, Iraklis Varlamis

The current study focuses on systematically analyzing the recent advances in the field of Multimodal eXplainable Artificial Intelligence (MXAI). In particular, the relevant primary prediction tasks and publicly available datasets are initially described. Subsequently, a structured presentation of the MXAI methods of the literature is provided, taking into account the following criteria: a) The number of the involved modalities, b) The stage at which explanations are produced, and c) The type of the adopted methodology (i.e. mathematical formalism). Then, the metrics used for MXAI evaluation are discussed. Finally, a comprehensive analysis of current challenges and future research directions is provided.

Submitted: Jun 9, 2023