An international team has developed a manual on artificial intelligence (AI) aimed at enhancing transparency and trust in this disruptive technology, particularly in sectors like healthcare, finance, and law.
Coordinated by the University of Granada (UGR), this resource offers a method to verify and certify the outcomes of complex models, contributing to the development of AI systems that are not only effective but also understandable and fair.
In recent years, the use of automated decision-support systems, such as Deep Neural Networks (DNNs), has significantly increased due to their predictive capabilities. However, their opaque nature complicates detailed interpretation of their behavior, raising ethical and legitimacy concerns.
To tackle this issue, the UGR team has introduced a comprehensive guide on explainable artificial intelligence (XAI) techniques.
This resource aims to serve as an essential guide for IT professionals seeking to understand and explain the results of machine learning models. Each chapter outlines applicable AI techniques for everyday situations, complete with examples and workbooks, along with a user manual detailing the requirements and benefits of each technique.
The guide is directed by Professor Natalia Díaz Rodríguez from UGR's Department of Computer Science and Artificial Intelligence, who is also a member of the Andalusian Interuniversity Institute for Data Science and Computational Intelligence (Instituto DaSCI).
“It is important to be aware of the capabilities and limitations of both advanced AI models and the explainability techniques that aim to argue and validate outcomes. Sometimes, explanations are neither satisfactory nor easily verifiable,” Díaz noted.
This work was conducted during Professor Díaz's tenure at the Polytechnic Institute of Paris and involves international collaboration with experts from the UK, France, and Austria, among other countries.