Hacker News Reader

Comment by havnagiggle | Hacker News Reader

Glad it worked out for OP, but NGL that is a long program. Probably the most disappointing part of the essay is still the idea that good experiments proving hypothesis wrong is not publishable. Is that a problem with the research community, or a journal problem, or self-imposed problem? Probably all of the above.

I left my PhD program after 3ish years. It was not working out with my advisor. I felt he was abusing research time, more interested in university research as a business than hard science. I'm sure the line is thin for many programs, but he did get in trouble for it years later after I left. His students were shuffled to other advisers and he was removed. So many years lost for so many students. I'm very glad I left when I did, but it's also one of my biggest disappointments as a lover of hard science and engineering.

View on HN

A well designed experiment will produce publishable results regardless of the specific outcome. As long as your error bars are small enough to learn something new it's a success. Not publishing unfavored but valid results is not normal, and myself and my colleagues would tend to place that in the territory of academic dishonesty. It's also counterproductive career-wise.

The most important grad school decision by far is choosing an advisor. The overall culture of the university and academia in general have a very small effect on things relative to the culture of the group you join.

My EngD supervisor explained to me that it's harder to prove something doesn't work than that it does.

A negative result thus has to pass a much higher bar and prove that it wasn't methodologically flawed in any way, whereas a positive result has a much lower bar to pass.

That's an interesting theoretical argument but in practice the much larger reason is that a negative result is not sexy.

Regardless, I don't agree with this philosophy when it comes to the functioning of the scientific system. A field like biology has many real results that are highly context-dependent. If we are going to allow people to regularly publish highly controlled studies and then make grand conclusions based on them (with real $$$ impact btw), we sure as shit need people publishing their observations about when something didn't generalize for them.

Not every paper needs to be a rock solid scientific conclusion, and it's pretty obvious that in practice many papers now aren't anyway. Why we play this silly game of pretend with scientific formalisms is beyond me. Science is getting larger and more complex, and there is 0 reason to assume that the way to scale is to have every lab try to tackle questions from start to finish in a nearly identical way.

I'm inclined to agree. Another weird effect I've heard of (2nd hand) act against publishing negative results is the social cost of casting doubt over a colleagues work. If science isn't reproducible its basically useless

Reply on news.ycombinator.com