You’ve shared your code. You’ve shared your data.…so why can’t anyone reproduce your results?
In real-world research projects, reproducibility often breaks in subtle ways, missing dependencies, unclear workflows, undocumented steps, or environments that no longer work. These issues don’t appear all at once; they emerge gradually as research evolves. In this talk, I introduce the idea of Reproducibility Debt as a way to understand how these issues accumulate over time. Importantly, this debt is not simply the result of poor practice, it reflects real-world constraints such as time pressure, experimentation, and shifting project goals. While reproducibility is often framed as a technical problem, solved through tools like Docker, Conda, or better documentation, it is in reality a multi-faceted challenge shaped by the interaction between technical, organisational, and human factors. Rather than focusing on technical fixes alone, this talk takes a socio-technical perspective: how decisions, trade-offs, and team practices influence reproducibility outcomes. I will show how reproducibility can be approached as something to understand, assess, and manage throughout a project, rather than something to fix at the end.
The talk introduces a lightweight way of identifying and tracking reproducibility risks based on a structured view of common contributing factors. If you have ever struggled to rerun your own code, or someone else’s, this talk offers a different way to think about why, and what to do about it.
Reproducibility is widely recognised as important, yet in practice it is often treated as a final step, something to address once development is complete. In fast-moving research environments, this approach rarely works. This talk reframes reproducibility as a continuous, socio-technical challenge shaped by the interaction between code, environments, workflows, and human decision-making. Rather than presenting a checklist of tools, this talk emphasises that reproducibility is not only about using Docker, Conda, or improving code structure. These are useful, but they address only part of the problem. Reproducibility challenges arise from a broader set of interacting factors that need to be understood and managed over time.
We will cover:
The session includes a practical “before vs after” example and highlights how reproducibility can be treated as an ongoing, manageable concern rather than a one-time fix. This perspective is particularly relevant for research software engineers and developers working in dynamic, exploratory environments.
Audience
Key Takeaways
Zara Hassan is a researcher in software engineering at the Australian National University, focusing on the reliability and trustworthiness of research software. Her work explores reproducibility as a socio-technical challenge, examining how technical, organisational, and human factors interact in real-world projects. She develops practical frameworks and tools to support the understanding, assessment, and management of reproducibility debt in research software.
