Ethical Perspectives in AI: A Two-folded Exploratory Study From Literature and Active Development Projects

Abstract

Background: Interest in Artificial Intelligence (AI) based systems has been gaining traction at a fast pace, both for software development teams and for society as a whole. This increased interest has lead to the employment of AI techniques such as Machine Learning and Deep Learning for diverse purposes, like medicine and surveillance systems, and such uses have raised the awareness about the ethical implications of the usage of AI systems. Aims: With this work we aim to obtain an overview of the current state of the literature and software projects on tools, methods and techniques used in practical AI ethics. Method: We have conducted an exploratory study in both a scientific database and a software projects repository in order to understand their current state on techniques, methods and tools used for implementing AI ethics. Results: A total of 182 abstracts were retrieved and five classes were devised from the analysis in Scopus, 1) AI in Agile and Business for Requirement Engineering (RE) (22.8%), 2) RE in Theoretical Context (14.8%), 3) Quality Requirements (22.6%), 4) Proceedings and Conferences (22%), 5) AI in Requirements Engineering (17.8%). Furthermore, out of 589 projects from GitHub, we found 21 tools for implementing AI ethics. Highlighted publicly available tools found to assist the implementation of AI ethics are InterpretML, Deon and TransparentAI. Conclusions: The combined energy of both explored sources fosters an enhanced debate and stimulates progress towards AI ethics in practice.