You will conduct research on the role of software testing in providing guardrails for AI coding agents.
These agents generate and modify code with increasing autonomy.
A key challenge is to ground the guardrails in requirements and stakeholder intent, which are clarified and communicated through developer conversations with agents.
The research also addresses how to maintain human control when expectations are incomplete, ambiguous, or conflicting.
The primary goal is to guide and verify AI coding agents through two connected artifacts: human-readable behavior specifications and executable tests, both grounded in requirements and stakeholder intent.
The specifications let stakeholders understand, discuss, and validate the intended behavior.
The tests give deterministic, automated checks that AI-generated code and later modifications conform to that behavior and signal misalignment or drift.
Depending on the research questions developed during the project, the work may include: • Investigating how intended behavior can be captured and refined from requirements, stakeholder input, developer-AI interactions, and development artifacts, and how gaps, ambiguities, and conflicts can be detected and resolved with humans in the loop.
• Developing and evaluating methods that connect human-readable behavior specifications with executable tests while keeping traceability to stakeholder intent.
• Exploring how specifications and tests can guide and verify agent-generated changes, and how they can be maintained as requirements and software evolve.
• Evaluating the effectiveness, reliability, and practicality of testing-based guardrails, with a focus on defect detection, stakeholder understanding, and support for human oversight.
The work combines method and tool development with empirical evaluation, including experiments and industrial case studies.
Specific research questions, methods, and technical solutions will be developed together with the supervision team.
We are looking for a candidate with strong software engineering and programming skills and an interest in software testing and agentic software engineering.
The Ph.D. position is a part of the strategic research area in IT and mobile communication ELLIIT, which is a partnership between Linköping University, Lund University, Halmstad University, and Blekinge Institute of Technology.
Rules and benefits for Ph.D. students in Sweden apply to this position.
The Ph.D. student will develop an individual study plan in software engineering together with his/her supervisors.
The position may include up to 20% of other services, e.g., teaching at the undergraduate and graduate levels.