Comment by losvedir | Hacker News Reader

Comment by losvedir | Hacker News Reader

Parent Full thread

I'm not sure if this article brings up anything new but it fans my crisis of epistemology all the same.

I've always been a good "the academics know what they're talking about" believer, but lately I've started having doubts.

I went to MIT and many of my good friends and peers went onto PhDs in the hard sciences (Princeton/CERN, Harvard CfA, CalTech chemistry) so I've always been impressed, almost reverential, towards them and the discoveries of their fields.

But as we branch away from physics, astronomy, chemistry, what I used to almost take on faith I'm now growing doubtful about. This excellent blog post[0] by Columbia statistics Professor paints the state of modern Psychology research unflatteringly. And then there's this post[1] looking at a parallel universe of scientists studying, peer reviewing, and publishing (in their own journals) parapsychology (being "psychic" and all that). I'd call it "cargo cult" science, but at much more than a superficial level it resembles many "real" scientific disciplines in terms of meta analyses, p-values, etc. I personally majored in Math and Econ, and for my Econ classes I did my fair share of modeling and teasing out effects from their confounding variables, etc, and was always a little leery about it all. It doesn't really surprise me that the Fed has written that Economics, too, is suffering a replication crisis[2].

But then how far does it go, and where do I go from here? I don't have the time or inclination to study to an expert level all these soft science fields and form my own opinions. What fields are as sturdy as particle physics? Medicine? Biology? Climate change? It's a slippery slope to anti-intellectualism but I feel myself slipping.

[0] http://andrewgelman.com/2016/09/21/what-has-happened-down-he...

[1] http://slatestarcodex.com/2014/04/28/the-control-group-is-ou...

[2] https://www.federalreserve.gov/econresdata/feds/2015/files/2...

It's a slippery slope to anti-intellectualism but I feel myself slipping.

Au contraire. You may be being truly intellectual for the first time. Do you know the motto of the Royal Society, the organization most responsible for the birth of modern science? "Take nobody's word for it." That is the essence of science -- replication and verifying things for yourself. You are not anti-intellectual just because you question the authorities in charge of a particular institution (academia).

What fields are as sturdy as particle physics? Medicine? Biology? Climate change?

IMO, fields where there isn't a quick feedback loop between controlled, real-world experiments and obvious result are almost all unsound. That includes macroeconomics, psychology, long-term nutrition, etc. In these situations, you can always play with the variables to get pretty much any result. Thus academics who either are good at playing politics and PR, or who have a politically convenient message, will rise to the top, not the academics who are rigorous and honest. You can find good academics, but they are rare and usually aren't the famous and acclaimed ones.

I think Economics is a fairly big outlier though. There's quite a good book by Economics Prof. Steve Keen called Debunking Economics, which is very accessible and goes over how most of the core tenants of neoclassical (mainstream) economics have been known to be flawed for decades, yet it's all still taught as fact.

It's not a scholarly work, but references a great deal of scholarly work.

It's fascinating to read and imagine how the profession has not been reformed in light of these known flaws (and the fact of the huge failure of mainstream Economics to predict the Great Recession).

There is some hope though in some student movements that are starting to challenge how economics is taught (like Rethinking Economics and the Post-crash Economics Society). It may be that like some times in science, the profession will advance "one funeral at a time" as the new generation comes in.

Neoclassical synthesis is still taught because, even though it's flawed, we have nothing better. Economists know that the assumptions behind neoclassical models aren't always fulfilled.

All science is always done with models and we just use the best models we have, while taking into account their limitations. People don't usually object to this in other fields, but in economics they do because they have stronger Opinions about economics than the natural sciences and the models sometimes disagree with their Opinions. If they had a better model, economists would welcome it, I'm sure, but they don't and they still want to keep their Opinions, so instead they blabber about how the current models are flawed. They're not wrong.

The Marxian school is a dead end and the Austrian school lacks scientific rigor. There's no point in teaching along with neoclassical synthesis things that are less rigorous than neoclassical synthesis while also having less breadth. If they were taught, the situation would be similar to what it is in psychology, where dubious approaches like psychoanalysis are taught alongside cognitive psychology.

This is a commonly held belief, but it's wrong, the dominant economic models in use today are neither rigorous (internally inconsistent, extremely poor empirical verification, major aggregation errors, lack of essential inputs) nor representative of an economy (a dynamic system, not an extremely stylized static equillibrium model). You should read Keen's book, it's point is not that those assumptions don't always hold, it's that they under _any_ real world circumstance will invalidate themselves.

https://www.youtube.com/watch?v=60__fqFjLMw

I blame language. As with all models, the further we move from blissfully perfect abstraction the less certain we can be about our model's predictions. But our language isn't very lenient with expressions of uncertainty.

We speak about a single psychology study's findings with the many of the same words of concrete certainty that we do with the deepest mathematical theorems because that is what is most natural. We say something is true when we really mean that we're pretty sure something is true under certain conditions. We assume that everyone understands the touch of chaos underneath the words based on the context. But they don't. And we don't either. We know better but we fool ourselves more or less by accident and repetition and it shakes us when contradictions inevitably arise from messy reality.

And on the other end of the spectrum, if any uncertainty at all is spoken of it seems to cast the whole statement into doubt.

The language of statistics and uncertainty is cumbersome and unnatural. If I could wish for anything and have it come true for the sake of advancement of humanity it would be for a standard modifier or tense for levels of uncertainty. I would trade that increase in complication for the reduction of confusion.

I tend to use:

"possibly" for 0<x<40% where x is the probability of some event

"maybe" for 0%<x<50%

"probably" for 75%<=x<=100%

And of course the sarcasm gods dictate that "proba-maybe" means x<10%

All models are wrong, some models are useful. And what is useful is subjective.

The Newtonian model works brilliantly at human-scale resolution. But if you are a quark, it is totally unconvincing. So the quarken-creatures' understanding of physics would have developed totally differently.

Similarly, SM/GR seems basically fine at solar-system resolution. But at an intergalactic scale it's completely unworkable – "regular" matter is only 5% of the universe! You are not going to get very far through the galaxy on that.

In reality, even "hard" science is very squishy. It's closely tied to the sort of creatures doing it, and different creatures would have done it differently. "Our" physics can answer a lot of questions – unless you want to leave the galaxy or learn about 95% of the stuff in the universe in which case we have no idea. But that doesn't matter (yet) because what we've got is good enough for a bunch more planetary expeditions.

Similarly, economics has solved some problems and not solved others. The difference is, lots of the "unsolved" problems seem quite urgent. Dark energy might be useful someday, but minimum wage theory would be useful right now.

But that has nothing whatsoever to do with science, or with correctness (all models are incorrect). It has more to do with our goals as a society, and how closely our discoveries map onto problems we think are urgent. A quarken society would have completely different set of priorities than we do, and the way they think about "sturdiness" would be entirely different as a result.

Some of the problem is it's much harder to judge softer fields from the outside. There is plenty of great research in Education for example, but it's much harder to evaluate without a lot more domain specific knowledge.

Reply on news.ycombinator.com