Generative AI towards Inference to the Best Explanation (XBest)

The existing post-hoc explainable AI (XAI) approach to provide explanations involves developing a secondary model to explain the inner workings of the first black box AI model, which can often be unreliable and inconsistent.

Inference to Best Explanation (IBE) is a form of logical inference that aims to find the most convincing explanation for a given set of observations.

The XBest project is focused on utilizing the IBE concept to provide better explanations of the decisions made by machine learning models.

The project will infer the most warranted explanation that is the most understandable ("Loveliest") explanation for available observational data. XBest will be supported by a team comprising experienced researchers in AI.

During this 4- year project, in order to apply IBE in XAI, we will construct a rational model to provide the best explanations for the decisions made by the AI model using the prior knowledge of the dataset.

The project’s first two years will be dedicated to developing the generative modelling framework followed by the extension of the framework in years 3 and 4 to enhance the traditional explanation towards a more IBE approach through a novel combination of XAI methods. The project also guarantees to present the corresponding confidence level of the provided explanation.

Eventually, this research will significantly advance the state of the art in XAI methods and push forward the research frontier in the field.