Reproducibility failures are essential to scientific inquiry

Abstract

Current fears of a "reproducibility crisis" have led researchers, sources of scientific funding, and the public to question both the efficacy and trustworthiness of science (1, 2).Suggested policy changes have been focused on statistical problems, such as p-hacking, and issues of experimental design and execution (3, 4).However, "reproducibility" is a broad concept that includes a number of issues (5) (see also www.pnas. org/improving_reproducibility).Furthermore, reproducibility failures occur even in fields such as mathematics or computer science that do not have statistical problems or issues with experimental design.Most importantly, these proposed policy changes ignore a core feature of the process of scientific inquiry that occurs after reproducibility failures: the integration of conflicting observations and ideas into a coherent theory.Here we argue, using examples from mathematics and computer science, that current discussions of the reproducibility crisis overlook the essential role that failures of reproducibility play in scientific inquiry.This viewpoint that reproducibility is a key part of inquiry suggests several new perspectives and policies to promote good science.First, science needs to be given the time necessary to reconcile conflicting results.It typically takes decades to probe the parameter space of a discovery to identify and characterize the fundamental variables.Second, reproducibility failures are a critical part of this journey, and attention

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 66 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.