Hacker News Reader

To be fair, given recent reproducibility fiasco, social sciences were scorned for a good reason.

Is there anything to indicate that computer science and computer engineering will avoid their own reproducibility fiasco?

Well... there's computer science, and then there's computer science.

All (most of?) the stuff done on P vs NP is good science that will stand up.

Studies on which language features make it better for developers are social science, because they involve those pesky humans. That stuff is likely to suffer from a reproducibility crisis.

reproducibility in social sciences is usually a function of a)insurmountable costs of recruiting participants, b)complexity of the questions, and c) lack of standardization of humans. Sure social scientists would like to have N=1 billion, but they'd be lucky to get funding for 1,000.

One of the problems is that the humans used were pretty standard -- all psychology students. Also publication bias.

can't win. If your sample isn't diverse enough it isn't representative. It it is representative, its likely too small to get a reliable effect due to the number of confounds.

Reply on news.ycombinator.com