I realized a couple of days ago that I'd never blogged about what Software Carpentry needs to accomplish in order to change the practice of science fundamentally and permanently. In a nutshell, we need to convert a fifth of scientists to our way of thinking. Once we do that, the odds are better than 50-50 that every time someone sends a paper out for review, at least one reviewer will ask hard questions about how the computational work was done. I get that number by assuming:

Number of Reviewers | Fraction of Papers |

2 | 10% |

3 | 40% |

4 | 40% |

5 | 10% |

which means the probability that *none* of a paper's reviewers will ask the right questions is 0.1×0.8^{2} + 0.4×0.8^{3} + 0.4×0.8^{4} + 0.1×0.8^{5} = 46.5%. It's a grossly simplistic model, but at least it gives us something to shoot for.

*This post originally appeared in the Software Carpentry blog.*