Vocabulary design considerations for agentic SE are five criteria proposed in Toward Agentic Software Engineering Beyond Code: Framing Vision, Values, and Vocabulary for designing and evaluating terminology in agentic software engineering (SE), offered as a precursor to formalizing the field’s taxonomy and knowledge areas. The paper motivates them by noting that although agentic SE is nascent, drift and discrepancies in terminology usage — some syntactic (e.g. “agentic AI software engineer” vs. “AI software engineer” vs. “agentic software engineer”) and others reflective of deeper semantic or philosophical issues — are already apparent.
Usage
The five considerations are: Relevance — fit for purpose and domain context, i.e. the extent to which a term is relevant to the domain, including whether traditional SE vocabulary should be revisited for relevance in the agentic SE context. Coverage — conceptual completeness and accuracy, i.e. the extent to which a term covers (nearly) all the goals and responsibilities of the process area, actor, activity, or artefact it aims to support (for example, since a software engineer is more than a coder, an autonomous coding agent needs to be able to do more than coding to be termed an AI software engineer). Acceptance — recognition and adoption by the community, i.e. the extent to which humans, especially those in the role being named, will agree to an AI agent being given their professional title. Consistency — uniformity of use, stability, and adaptability, i.e. the extent to which different technology and research groups use a term in the same way, in both form (syntax) and meaning (semantics), how stable that usage remains, and how it adapts over time. Philosophical alignment — reflexivity leading to awareness of underlying assumptions, power dynamics, and fairness, i.e. reflecting on who proposes terminology and why, who stands to gain or lose, and whether every stakeholder (including underfunded regions and the global south) has a voice.
Related Terms
Toward Agentic Software Engineering Beyond Code: Framing Vision, Values, and Vocabulary