Abstract
Replicability and reproducibility of computational models has been somewhat understudied by “the replication movement.” In this paper, we draw on methodological studies into the replicability of psychological experiments and on the mechanistic account of explanation to analyze the functions of model replications and model reproductions in computational neuroscience. We contend that model replicability, or independent researchers' ability to obtain the same output using original code and data, and model reproducibility, or independent researchers' ability to recreate a model without original code, serve different functions and fail for different reasons. This means that measures designed to improve model replicability may not enhance (and, in some cases, may actually damage) model reproducibility. We claim that although both are undesirable, low model reproducibility poses more of a threat to long-term scientific progress than low model replicability. In our opinion, low model reproducibility stems mostly from authors' omitting to provide crucial information in scientific papers and we stress that sharing all computer code and data is not a solution. Reports of computational studies should remain selective and include all and only relevant bits of code.
Highlights
Public controllability of research and reliability of results are the cornerstones of science, so it is no surprise that recent doubts about researchers’ ability to consistently duplicate findings in a number of scientific fields have caused quite a stir in the scientific community (Button et al 2013; Loken and Gelman 2017; Maxwell et al 2015)
We introduce the notions of repeatability and reproducibility as they are used in physics, chemistry and medicine, and discuss the various types and functions of replication studies in psychology, where the recent crisis has inspired serious methodological reflection on the subject
We describe a basic methodological distinction between direct and conceptual replications
Summary
Public controllability of research and reliability of results are the cornerstones of science, so it is no surprise that recent doubts about researchers’ ability to consistently duplicate findings in a number of scientific fields have caused quite a stir in the scientific community (Button et al 2013; Loken and Gelman 2017; Maxwell et al 2015). We introduce the notions of repeatability and reproducibility as they are used in physics, chemistry and medicine, and discuss the various types and functions of replication studies in psychology, where the recent crisis has inspired serious methodological reflection on the subject. According to Schmidt (2009), replication studies in psychology serve a wide range of functions, such as aiming to discover false positives, controlling for artifacts, addressing researcher fraud, attempting to generalize a result to a different population and trying to confirm a previously supported hypothesis using a different experimental procedure. I.e. direct replications performed by an independent team of researchers, reduce the likelihood of false positives, especially those stemming from experimenter effects (Rosenthal 1966) and tacit knowledge They provide information needed to establish the size of an effect, which the original investigators are prone to overestimate. This is all the more significant because a vast majority of theoretical studies in neuroscience and cognitive science are based on computational modeling (Busemeyer and Diederich 2010)
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have