Author:
Hani Hagras
Affiliation:
The Computational Intelligence Centre, School of Computer Science and Electronics Engineering, University of Essex, Wivenhoe Park, Colchester, CO43SQ, U.K.
Keyword(s):
Explainable Artificial Intelligence, Fuzzy Logic Systems.
Abstract:
We are entering a new era which is characterized by huge amounts of data which are generated from almost every application in our everyday lives. It is getting easier to organise such huge amounts of data via efficient data bases and ever growing and cheaper data storage systems (which can nicely scaleup in cloud based solutions). Due to the huge sizes, high dimensionality and complex relationships of such data, Artificial Intelligence (AI) technologies are well placed to handle such data and generate new services, business opportunities and even provide breakthroughs to completely change our lives and realise new industrial revolution as anticipated. The vast majority of AI technologies employ what is called opaque box models (such as Deep learning, Random forests, support vector machines, etc) which produce very good accuracies but it is quite difficult to analyse, understand and augment such models with human experience/knowledge. Furthermore, it is equally difficult to understand
, analyse and justify the outputs of such opaque AI models. Hence, there is a need for Explainable AI (XAI) models which could be easily understood, analysed and augmented by the users/stake holders. There is a need also for such XAI models outputs to be easily understood and analysed by the lay user. In this paper, we will review the current trends in XAI and argue the real-world need for true XAI which provides full transparency and clarity at the model and output level.
(More)