Beli: um guia para avaliação de usabilidade em sistemas machine learning de recomendação

Abstract

The evolution of digital technologies in a mutual influence with the human being, which aims to make our lives easier and more functional, favored the development of Machine Learning (ML) applications, a field of Artificial Intelligence. Since they imply economic and social benefits, these systems have been increasingly used to improve the usability of systems and offer a personalized experience. Recommendation machine learning systems are designed to help the user make better choices in large content catalogs, shaping the behavior of applications in various categories, however, it is necessary that these applications, increasingly present in everyday life, shock a interaction interaction quality beyond the accuracy of its algorithm. This work describes the development of a guide for usability evaluation in recommendation machine learning systems, called BELI. For this, surveys were carried out regarding the specific requirements of this type of system and how usability estimates are currently carried out in this field. Given the window of opportunity related to increasing the scope of usability estimates in machine learning recommendation systems, the opportunity to develop a usability evaluation guide to support evaluators to more effectively and efficiently find specific usability problems was observed. this type of system. A systematic review of the literature was developed to collect data related to the research question and generate resources to guide the case studies, carried out to, in a joint analysis with the systematic review, result in the activities listed in the guide. The BELI guide, divided into suggested activities related to the usability testing method, helps the designer when planning, executing and analyzing usability tests for recommender systems, considering their specific aspects.