Machine Learning Contributes to a "Reproducibility Crisis" Within Science

Scientific discoveries made using machine learning techniques can not be automatically approved, warned a statistician of Rice University.

A growing trend: Scientists in many disciplines are increasingly using machine learning systems to refine and accelerate data analysis. This speeds up their ability to make new discoveries, for example by discovering new pharmaceutical compounds.

The problem? Dr. Genevera Allen, Associate Professor at Rice University, warned that the adoption of machine learning techniques was contributing to a growing "reproducibility crisis" in science, where a disturbing number of research findings only can be repeated by other researchers, casting doubt on the validity of the first results. "I'd dare to say that a lot of that comes from the use of machine learning techniques in science," Dr. Allen told the BBC. In many situations, discoveries made in this way should not be trusted until they are verified, she explained.

On the positive side: Work is underway on the next generation of machine learning systems to ensure they are able to assess the uncertainty and reproducibility of their predictions, said Dr. Allen.

Sign up for our daily newsletter The Download to get your dose of the latest news from the world of emerging technologies.