Version Control for Scientific Reasoning: A Paradigm Shift from Artifacts to Processes

Keywords: version control, meta-science

TL;DR: We propose a paradigm shift: treating scientific research as version-controlled computational processes rather than collections of artifacts.

Abstract: Scientific research suffers from fragmented reproducibility, scalability crises in peer review, and implicit assumptions that resist systematic tracking. We propose a paradigm shift: treating scientific research as version-controlled computational processes rather than collections of artifacts. Our Scientific Domain-Specific Language (DSL) formalizes research operations (observe, hypothesize, execute, analyze) within a directed acyclic graph structure that captures epistemic evolution, assumption dependencies, and collaborative decision-making. This approach enables granular attribution tracking, automated quality assurance, and systematic knowledge building across research communities. We demonstrate the framework through implementation studies showing significant improvements in research reproducibility and collaboration efficiency while addressing the AI validation paradox through multi-scale validation mechanisms. Our approach transforms scientific collaboration by making reasoning processes explicit, version-controlled, and systematically improvable. An example repository with attribution may be found at https://github.com/cosci-org/cosci-project-example.

Submission Number: 109