Computational notebooks are a staple of interactive data analysis, letting users mix code and outputs and execute cells in any order. That flexibility is good for exploration but creates a reproducibility problem: outputs can become inconsistent with the current state of the code.

The paper, FlowBook: Enforcing Reproducibility in Computational Notebooks, addresses this issue directly. The abstract frames the problem and introduces FlowBook as a system for enforcing reproducibility. The provided text ends mid-sentence, so the specific mechanism is not described in the available excerpt.

The significance is that it tackles a well-known pain point: notebooks that look fine but can't be rerun reliably. By enforcing reproducibility, FlowBook could make shared analyses more trustworthy.