From Andrew Glassner, Seattle, Washington, US
Your interview with Stuart Ritchie paints a bleak picture for reproducibility in science (22 August, p 36). But taking all of science to task like this may be too broad-brushed.
In computer science, particularly in fields such as computer graphics and artificial intelligence, publications are now expected to include a link to a public repository, such as GitHub, providing the complete source code. Reproducibility is a snap: install the code and run it.
For all but the most gargantuan systems, every claim made by the authors can be easily confirmed or demonstrated to be false. This explicit mechanism for easy, objective reproducibility may partly explain the explosive growth of these fields.
Advertisement
