Estado da Arte sobre Engenharia de Requisitos e Explicabilidade em Sistemas Baseados em Aprendizado de Máquina

With the recent growth in the use of Machine Learning (ML)-based software, concerns arise regarding explaining the results generated. Explanations help with transparency and increase stakeholder trust. Explainability, a term used to refer to these explanations, is considered a non-functional requirement (NFR) that substantially impacts the quality of ML systems. Explainability has become a mandatory requirement outlined in various laws in several countries. Additionally, Explainable Artificial Intelligence (XAI) is a field that studies methods supporting explainability in ML-based systems, focusing mainly on technical explanations. This study is not limited to technical explanations but provides a comprehensive overview of Requirements Engineering (RE) and the explainability requirement in AM-based systems. To achieve this, we planned and executed a Systematic Mapping Study protocol, adopting automatic searches in six databases. From the 200 returned articles, after applying selection criteria, we analyzed and reported the results of 27 articles. Our findings reveal that explainability is an emerging quality NFR in ML-based systems, challenging classical RE paradigms.