Agentic AI helps maintain code for reproducible research
Reproducibility in the Age of Agentic AI: Context Engineering at the Timescale of a Codebase
Software EngineeringComputers and Society
Summary
Keeping research reproducible is about carefully organizing code, tests, and records so that others can confirm results. The authors argue that AI coding agents lower the effort needed to maintain these parts of a project while making their advantages clear right away. Even so, humans still need to check these outputs and the decisions behind them to ensure correctness. This suggests a collaborative future where AI supports but does not replace human judgment in research.
What this means in practice
- •For software development teams: Improve codebase maintainability by integrating AI agents to assist with testing, version history, and documentation upkeep.
- •For quality assurance engineers: Use AI coding agents to reduce overhead in verifying software build artifacts and their supporting records during reproducibility checks.
A position paper. It proposes an approach and reports no results.
Authors
Lorena A. Barba
Abstract
Reproducible research practices are context engineering for AI coding agents. I argue that agents lower the cost of maintaining tests, commit histories, repository structure, instructions, and decision records while making their benefits immediate. Researchers remain responsible for verifying these artifacts and the scientific judgments they encode.