What is it about?
Today, we are all subject to decisions made entirely or partially by algorithms. Many of these algorithms are known as black-box, meaning the user cannot see/understand their inner workings. In this context, the use of explainable/interpretable algorithms or the development of solutions to understand the results of black-box algorithms is fundamental.
Featured Image
Photo by Joshua Sortino on Unsplash
Why is it important?
The present review concluded that, in recent years, several works have emerged to try to help the process of explaining the results of black-box algorithms. However, in general, it is a separate set of approximations, which do not speak to other existing solutions and, often, do not have their performance compared. Few works design human-centered methods, and many of the explanations require users to be experts in the problem being addressed.
Perspectives
We believe explainable artificial intelligence (XAI) is fundamental to produce better and more reliable systems that can minimize any kind of bias or prejudice. The discussion about these methods is still quite recent, but extremely necessary among developers and users of artificial intelligence systems.
Dr. Luciano Digiampietri
Universidade de Sao Paulo Campus da Capital
Read the Original
This page is a summary of: Machine Learning post-hoc interpretability: a systematic mapping study, May 2022, ACM (Association for Computing Machinery),
DOI: 10.1145/3535511.3535512.
You can read the full text:
Contributors
The following have contributed to this page
