People are asserting that these reversals are a normal part of science and therefore good. I think they are a normal part of science but reveal a huge flaw in the method: scientists frequently over extrapolate and end up drawing erroneous conclusions based on limited data. The widespread presence of basic methodological flaws in many scientific papers such as misunderstanding the meaning of p values means a lot of so-called science is built on faulty reasoning. This is worse than merely presenting a model that imperfectly represents the data, it is simply wrong: reaching conclusions that the data do not justify and never could.
You see this most often with efforts to reduce complex systems down to the effect of single variables; such efforts are almost guaranteed to miss the point. The serotonin story is a good example. The notion that the complexity of human psychology that led to depression could be reduced to the activity level of a single neurotransmitter should strike us as an absurd exercise on its face. It's not even wrong, it's a basic misreading of the problem space.
You see this most often with efforts to reduce complex systems down to the effect of single variables; such efforts are almost guaranteed to miss the point. The serotonin story is a good example. The notion that the complexity of human psychology that led to depression could be reduced to the activity level of a single neurotransmitter should strike us as an absurd exercise on its face. It's not even wrong, it's a basic misreading of the problem space.