Improving MBSE Tools UX with AI-Empowered Software Assistants

Abstract

Model Based Software Engineering (MBSE) consists in applying Model-Based Engineering principles - mostly abstraction and automation - to software engineering practices. As concluded by several empirical studies during the past 10 years, applying MBSE methodology is no more a question and we shall now wonder how to do it rather than if we should. One of the major criticisms of MBSE remains about the available tooling to apply its methodology. Multiple surveys show that the usability of tools and the lack of skills from the user are two key barriers that still slow down the spread of the MBSE approach. In the meantime, these surveys also highlight the difficulties software engineers often encounter during the modeling activity itself, regardless of the tool. Indeed, complex operations such as problem-to-model mapping, consistency checking or error identification are still often manually performed by humans. The recent advances in AIs introduced new software-based systems able to interact with their users to help them in their daily life. The purpose of this PhD is to investigate how such AIs (which we will call software assistants) could help software engineers to face the complexity of software modeling. Interactions between software assistants and users of MBSE tools will be the bulk of this work. We plan to implement software assistants and provide them an access to a high-quality knowledge repository on models that we will build to study these interactions. Based on these studies, we hope to contribute laying the foundations of interactions with software assistants. The end result will feature a knowledge repository and a Software Assistant acting together to help users modeling by catering for new ideas, recommendations and help. This work will also provide frameworks to create knowledge repositories and IDE-embedded software assistants.