Classification of Explainable Artificial Intelligence Methods through Their Output Formats

Machine and deep learning have proven their utility to generate data-driven models with high accuracy precision. However, non-linear, complex structures are often difficult interpret. Consequently, many scholars developed a plethora of methods explain functioning the logic inferences. This systematic review aimed organise these into hierarchical classification system that builds upon extends existing taxonomies by adding significant dimension—the output formats. The reviewed scientific papers were retrieved conducting an initial search on Google Scholar keywords “explainable artificial intelligence”; machine learning”; “interpretable learning”. A subsequent iterative was carried out checking bibliography articles. addition dimension explanation format makes proposed practical tool for scholars, supporting them select most suitable type problem at hand. Given wide variety challenges faced researchers, XAI provide several solutions meet requirements differ considerably between users, problems application fields intelligence (AI). task identifying appropriate can be daunting, thus need helps selection methods. work concludes critically limitations formats explanations providing recommendations possible future research directions how build more generally applicable method. Future should flexible enough posed widespread use AI in fields, new regulations.