Explainability and Fairness in Machine Learning
Year: 2021
Type: dissertation
Abstract: Over the years, when speaking about prediction models, the focus has been set on improving their accuracy, at the cost of loosing any comprehension of how the model predicts. Consequently, it has also been lost the ability of knowing if the behavior of the model is correct. Moreover, due to the fact that the addresses of the predictions do not have information about how ethic or fair the model is when predicting, persons become reticent to use such type of models. Therefore, in the last years there have been developed investigations aiming to explain such predictions in order to make them intelligible for humans, using techniques like LIME or SHAP, responsible for explaining in an interpretable way what happens behind the prediction. This work addresses this issue and reviews recent literature on the topic.
Author: Villalba Pizarro, Fátima del Pilar
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7028543786
