DAnTE: a taxonomy for the automation degree of software engineering tasks

TL;DR

This paper introduces DAnTE, a taxonomy categorizing automation levels in software engineering, evaluates existing tools within this framework, especially AI-based ones, and discusses future tool development.

Contribution

The paper proposes a new taxonomy for automation in software engineering and applies it to analyze current and future AI-based tools.

Findings

Existing tools are positioned within the taxonomy.

AI tools have specific limitations identified.

Future tools may evolve within the proposed framework.

Abstract

Software engineering researchers and practitioners have pursued manners to reduce the amount of time and effort required to develop code and increase productivity since the emergence of the discipline. Generative language models are just another step in this journey, but it will probably not be the last one. In this chapter, we propose DAnTE, a Degree of Automation Taxonomy for software Engineering, describing several levels of automation based on the idiosyncrasies of the field. Based on the taxonomy, we evaluated several tools used in the past and in the present for software engineering practices. Then, we give particular attention to AI-based tools, including generative language models, discussing how they are located within the proposed taxonomy, and reasoning about possible limitations they currently have. Based on this analysis, we discuss what novel tools could emerge in the middle and long term.