ReScience C: A Journal for Reproducible Replications in Computational Science
Citation: Nicolas P. Rougier, Konrad Hinsen (2019/06/29) ReScience C: A Journal for Reproducible Replications in Computational Science. International Workshop on Reproducible Research in Pattern Recognition (RSS)
DOI (original publisher): 10.1007/978-3-030-23987-9_14
Semantic Scholar (metadata): 10.1007/978-3-030-23987-9_14
Sci-Hub (fulltext): 10.1007/978-3-030-23987-9_14
Internet Archive Scholar (search for fulltext): ReScience C: A Journal for Reproducible Replications in Computational Science
Download: https://doi.org/10.1007/978-3-030-23987-9 14
Tagged: Computer Science
(RSS)
Summary
The author identifies some cases where the act of trying to replicate a study brought up new research questions (author's citations 3, 5, 6, 9, 10, 12). Most existing journals do not accept replication results because they prioritize originality. Therefore, the author co-founded ReScience C ReScience C, a journal which publishes replications of computational work. See also Sustainable computational science: the ReScience initiative, which was written at the launch of ReScience C. In ReScience C, the reviews are conducted openly on GitHub; accepted experiments are archived in Zenodo. The GitHub PR process is somewhat limiting, since it is not automated yet, but overall a better platform for communicatoin than traditional journals. In practice, Rescience C reviews and authors have been polite., however a failed replication can be "equivalent to publicly accusing the authors of the target work of having made a mistake, which is a potential source of conflict." The authors consider a new model, where replicability is attempted pre-publication, and the work is published with its replicability study. A new journal, ReScience X, will focus on replicability for physical (not computational) experiments.
Terminoligy note: Reproducible means that the work results can be exactly reproduced, using the exact same code. Replicable means that main results can be approximately recreated by conceptually equivalent software. Reproducible replication is a replication experiment (new code written from the high-level description) that is reproducible (code and data are available).
