Reproducibility and Integrity of Science: Problems, Solutions, and Problems of Solutions
Richard Shiffrin is a Distinguished Professor and Luther Dana Waterman Professor of Psychological and Brain Sciences at Indiana University.
Abstract: Recent years have seen an enormous amount of discussion and writings, by scientists, science writers and lay persons, about scientific publications whose reported results are unlikely to be reproducible. The problem is real and is presently being addressed by the scientific community. Publicity about this issue is itself a large problem, calling the scientific process into question in the eyes of the public, funding agencies and lawmakers. Much of the bad publicity is aimed at social psychology reports because lay people believe they are capable of judging the plausibility and importance of such reports, but reproducibility is a problem for all of science, and particularly for publications about health, given the societal consequences of actions based on unreproducible results. Although I have been publishing technical papers using Bayesian inference to assess reproducibility, I will in this talk focus on the general issues: What are the problems, what are the causes, what solutions are being proposed, what solutions might work, why some remedies will make things worse rather than better, why science works despite the present problem, and how the scientific community is moving forward to improve matters.