Scoping Software Engineering for AI: The TSE Perspective

In recent years, important advances in Artificial Intelligence (AI), and, in particular, in Machine Learning (ML), including Deep Learning (DL) and Large Language Models (LLMs), have caused a substantial increase of submissions to all Software Engineering (SE) venues (conferences and journals) related to SE with and for AI. They are commonly referred to as AI for SE and SE for AI. On the one hand, AI techniques have been used to provide better solutions to problems with which software engineering researchers have struggled for a long time (e.g., code completion, fault localization, program repair, and test case generation), as well as solve problems for which automated solutions did not exist in the past, or were very limited, e.g., automated bug reproduction, code review, or the generation of complete, non-trivial program elements. Contributions along these lines are commonly described as “AI for Software Engineering” and are welcome at IEEE Transactions on Software Engineering (TSE). The questions of what constitutes novelty and significance of such papers are interesting and complex, and we will address them in a future editorial. On the other hand, certain AI artifacts, e.g., ML models, can be seen as software components forming part of a more complex software system. Thus, the engineering of ML components might be considered to be of a core interest to SE. In fact, many top SE venues, including IEEE TSE, have been publishing a broad range of contributions on testing, verifying, repairing, understanding, and optimizing ML components, under the broad umbrella of “software engineering (SE) for AI”.