I have not read the book Science Fictions, but will comment that the core problem in academic science is the intense structural reward of telling “stories” rather than focusing on high quality data generation and robust causal models. A snappy story gets you a paper in Cell, Science, or Nature, a grant award, and promotion. That is the pressure driving many of the distortions.<p>In contrast, a decade-long effort to generate foundational data sets gets you a footnote or acknowledgement. Think Tycho Brahe versus Copernicus.<p>The imperative of generating snappy, and preferably simple series of stories is that scientists focus intensely on reductionist systems and neat little “mechanisms”. To get to mechanism, the modus operandi is to trim away fundamental system complexity. Complex state-dependent feedback systems get converted into sad but comprehensible cartoons. And we are surprised that findings do not replicate? The one phrase that makes me cringe in the biomedical literature is “necessary and sufficient”. In your dreams!<p>Yuri Lazebnik’s wonderful “Can a Biologist Fix a Radio” is a classic commentary published in 2002, well worth reading in this context. “Could a Neuroscientist Understand a Microprocessor” by Eric Jonas and Konrad Kording (2017, PLoS Comp Biology) is the update for all of you on Hacker News.<p>The current structure of science usually does not reward high quality persistent data and metadata. This seems antithetical to what we are usual taught about the scientific method. The reality is that doing data justice has been getting lip service from day 1. I still see little evidence of serious efforts in this regard (genomics being a welcome exception) since the true cost of data preservation and sharing is staggering.<p>A salutary final note from Richard Hamming, that I do not mean in any way as an excuse: “In science, if you know what you are doing, you should not be doing it. In engineering, if you do not know what you are doing, you should not be doing it.”<p>In both cases though high quality data/metadata are just as important—perhaps more so, than the big ideas.