Computer Science > Software Engineering
Title:Trusting Trust in AI Software Engineering Agents: The Why, What, and How?
Abstract:AI software engineering agents are increasingly integrated into development workflows, from autocomplete assistants to autonomous repair agents and multi-agent software development pipelines. Trust matters because these agents can produce, revise, and pass on artifacts that developers, other agents, or organisations may later rely on under uncertainty and risk. Yet trust in such systems is often treated as a single measurable property of an AI system. This is insufficient for software engineering, where trust depends on who relies on the agent, how that reliance is structured, and which properties must hold in that deployment. We propose a structural framework for reasoning about trust in AI software engineering agents. Inspired by the recursive nature of trust in software systems, we decompose trust into three analytical questions: Why trust?, identifying the trust scenario; What trust?, identifying the trust configuration; and How much trust?, identifying the trust properties. From these questions, we derive three scenarios, four configurations, and five trust-relevant properties, state which of these are closed by construction and which remain open to revision, and instantiate the resulting profile on a ChatDev-style multi-agent workflow. The framework is conceptual, and is offered as a foundation for the empirical work that would validate it, rather than as a numerical trust metric.
| Subjects: | Software Engineering (cs.SE) |
|---|---|
| Cite as: | arXiv:2611.21634 [cs.SE] |
| (or arXiv:2611.21634v1 [cs.SE] for this version) | |
| https://doi.org/10.48550/arXiv.2611.21634 arXiv-issued DOI via DataCite |
Submission history
Access Paper:
Additional Features
- Audio Summary
Current browse context:
References & Citations
- Google Scholar
- Semantic Scholar
BibTeX formatted citation
Bookmark
Bibliographic and Citation Tools
Code, Data and Media Associated with this Article
Demos
Recommenders and Search Tools
arXivLabs: experimental projects with community collaborators
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.