The Korea Advanced Institute of Science and Technology (KAIST) Explainable Artificial Intelligence Research Center (Director: Professor Jaesik Choi) announced on the 27th that it has developed a plug-and-play explainable artificial intelligence framework that can provide explainability for artificial intelligence (AI) models without separate, complex settings or specialized knowledge, and released it as open source.
Explainable artificial intelligence (XAI) refers to technologies that explain key factors affecting the results of AI systems in a form that humans can understand.
Interest in and research in the field of XAI have increased as reliance on black-box AI models with opaque internal decision-making processes, such as recent deep learning models, has grown.
However, when utilizing XAI, the explanation algorithms differ depending on the type of deep learning model, and it was difficult to identify the parameters that could be applied to each algorithm. In addition, a separate tool was needed to evaluate the explanation algorithm.
The Plug and Play XAI Framework (PnPXAI Framework), which has recently been released as open source, was developed to solve these difficulties.
