I see the same issues also in computer science especially when looking into recent trends such as AI or blockchain/NFTs before that. There are definitely areas that are more rigorous than others but the replication crisis is a problem in many many scientific fields. If your results are not completely outlandish and don’t go against the vibe, no one will ever bother to check your results.
There are so many different areas of computer science though… Everything from pure mathematics (e.g ‘we found a new algorithm that does X in O(logx)’) to the absurdly specific (‘when I run the load tests with this configuration it’s faster’). The former would get published. The latter wouldn’t. And the stuff in the middle ranges the gamut from ‘here’s my new GC algorithm that performs better in benchmarks on these sample sets’ to ‘looks like programmers have fewer bugs when you constrain them with these invariants’. All the way over on the other side, NFT/Blockchain/AI announcement crap usually doesn’t even have a scientific statement to be expressed, so there’s nothing to confirm or deny. There are issues with some areas, but I’m not sure that replication is really the big one for most of these. Only one it commonly applies to IMO are productivity or bug-frequency claims which are generally hella suss
I see the same issues also in computer science especially when looking into recent trends such as AI or blockchain/NFTs before that. There are definitely areas that are more rigorous than others but the replication crisis is a problem in many many scientific fields. If your results are not completely outlandish and don’t go against the vibe, no one will ever bother to check your results.
There are so many different areas of computer science though… Everything from pure mathematics (e.g ‘we found a new algorithm that does X in O(logx)’) to the absurdly specific (‘when I run the load tests with this configuration it’s faster’). The former would get published. The latter wouldn’t. And the stuff in the middle ranges the gamut from ‘here’s my new GC algorithm that performs better in benchmarks on these sample sets’ to ‘looks like programmers have fewer bugs when you constrain them with these invariants’. All the way over on the other side, NFT/Blockchain/AI announcement crap usually doesn’t even have a scientific statement to be expressed, so there’s nothing to confirm or deny. There are issues with some areas, but I’m not sure that replication is really the big one for most of these. Only one it commonly applies to IMO are productivity or bug-frequency claims which are generally hella suss