Unreliable: Bias, Fraud, and the Reproducibility Crisis in Biomedical Research (Hardcover) (ISBN-13: 9780231216234)

Reproducibility is fundamental to the scientific method. After reading a paper describing research findings, a scientist should be able to repeat the experiment and obtain the same results. Yet an alarming number--perhaps as high as 90 percent--of published biomedical research papers face challenges in independent replication. Such issues range from honest mistakes to outright fraud. The scope of this crisis, however, underscores deeper systemic issues within the scientific community: its culture, incentives, and institutions.

In Unreliable, the distinguished scientist Csaba Szabo examines the causes and consequences of the reproducibility crisis in biomedical research, showing why the factors that encourage misconduct stem from flaws in real-world science. There are many culprits, including commonplace research methods and dubious statistical techniques. Academic career incentives, hypercompetition for grant funding, and a bias toward publishing positive results have exacerbated the problem. Deliberate data manipulation and fabricated findings churned out by "paper mills" are disturbingly common. Academic institutions and publishers, for their part, have perpetuated a culture of impunity.

Szabo explores how these failures have hindered scientific progress and impeded the development of new treatments, and he introduces readers to the "science sleuths" who tirelessly uncover misconduct. He proposes comprehensive reforms, from scientific training to the grant system through the publication process, to address the root causes of the crisis. Written in clear language and leavened with a keen sense of irony, Unreliable is an essential account of the reproducibility crisis that gives readers an inside look at how science is actually done.

Author: Csaba Szabo
ISBN-10: 0231216238
ISBN-13: 9780231216234
Publisher: Columbia University Press
Language: English
Published: 03/11/2025
Pages: 328
Format: Hardcover