Live data from Hacker News

People have no idea which sciences are robust

arstechnica.com

61–63 of 63 posts

Re: People have no idea which sciences are robust

#61
post #32

Earlier quoted context omitted.

I agree that the problem is complex systems, not biology per se. But physics is able to be quite predictive because it is able to isolate one basic phenomenon at a time and model it with great precision (gravity, electromagnetism, particle physics, etc). Then, if we want to build devices based on these phenomena from the ground up, we can also do that and predict their behavior with great accuracy (e.g., behavior of…

> Physics does have predictive problems when it is applied to weather, climate, etc, because those are complex systems. But that kind of the thing is a minority of the subject matter in physics. There is a far greater number of humans working in applied physics than in characterizing isolated aspects of theoretical systems so I'd question how you judged "minority" there :) Perhaps our disagreement is just in choice o…

Well, "physics" really encompasses just about everything, including biology, so I am implicitly limiting it to things which would not fall into another, more specific field, such as engineering or meteorology. Right, word choice.

Using any definition for "physics" close to this, while the percentage of biological questions that involve complex systems is close to 100%, it is much lower in physics. The subsets of physics problems that do involve complex systems will suffer the same predictive problems.

In fact, this conversation has got me wondering whether "complex system" really means anything more than "a system whose behavior is hard to predict". I know that complex systems have other common attributes, but really the unpredictability seems to be the defining feature.

This is a long way of saying "I agree that we don't really disagree" :)

Re: People have no idea which sciences are robust

#62
post #42

Earlier quoted context omitted.

There's nothing squishy about string theory. The line between currently untested hypotheses and those supported by reliable experimental results is quite clear in physics. You frequently see journalists representing the results of unreplicated garbage psychology experiments as scientific fact yet you never hear anyone claiming string theory is any such thing. We know how to test string theory but we lack the ability…

The comment I was responding to was specifically calling out biology as "squishy". As a computational biologist myself, I have experience in getting my purely computational papers rejected before I learned that you have to collaborate with experimentalists to get a really good paper.

(Fellow CB here) I have had the same experience. Although I think this phenomenon is because of a heavily empirical bias in biology, not because the experimental data that is usually paired with comp bio models is actually all that useful in supporting the model.

"Generate hypothesis with CB and present experimental data that is mostly irrelevant to it" seems to be a good recipe for getting Nature papers, actually.

Re: People have no idea which sciences are robust

#63

The sample studied by Broomell and Kane contained 217 people! While the results may be "interesting" and "evocative", the size of the sample really does not allow any conclusions to be drawn. The arstechnica post ignores the details; read the original paper.

An online survey at that, meaning that they can't even properly gauge if that sample was representative at all.
Post reply on HN