Science’s Biggest Fail • Hacker News Board

Re: Science’s Biggest Fail

#251

Post by msandford » Tue, Feb 03, 2015, 3:38 PM UTC

snowwrestler wrote: post #240

msandford wrote: post #229

Earlier quoted context omitted.

The orders of magnitude more nodes and inputs mean that there are orders of magnitude more opportunities for mistakes to be made in the models. It also increases the likelihood of failing to find a confounding variable. The upper bound on system complexity is factorial. I shouldn't need to explain how big that is when you start talking about many orders of magnitude.

That is why it is so difficult to predict the weather. But you don't need to predict the weather to predict changes to the global climate. Just like you don't need much of a model to predict that if a rogue planet passed through the solar system, existing orbits would be perturbed. Calculating the precise perturbations would be complex, but not predicting their existence. Likewise, it's difficult to predict exactly h…

Just because you can make a simplification doesn't mean you have achieved understanding. That understanding might be correct, or it might not. Even if you think you know the error bars are small, that doesn't mean that they are.

Given that the earth has had temperature excursions of +4C and -4C in the past 100k years it seems that there are some kind of forces that keep the temperature within that range. What are they? Climatologists are saying "tipping points" and whatnot, but if you believe the very long climate history as measured by proxies you would have to admit that there certainly SEEMS to be something that stops the temperature from rising further, and something that stops it from falling further. What are those mechanisms?

Re: Science’s Biggest Fail

#252

Post by api » Tue, Feb 03, 2015, 4:37 PM UTC

TeMPOraL wrote: post #249

rybosome wrote: post #231

Earlier quoted context omitted.

Honestly, I agree. In some ways when I think about the future I feel very hopeful; advancements in narrow AI and medical technology in particular are very easy to imagine as beneficial. Mostly though I feel very cynical and depressed by what appears to be the cultivation of ignorance as a virtue. I am convinced that we are fucked, barring a massive breakthrough in photovoltaic efficiency or nuclear fusion.

> Mostly though I feel very cynical and depressed by what appears to be the cultivation of ignorance as a virtue. I am convinced that we are fucked, barring a massive breakthrough in photovoltaic efficiency or nuclear fusion. I share that. I actually suffered a short-term deppressive period because of that (as weird as it sounds, for few months I felt guilty and afraid whenever I turned on the gas stove). We have may…

You were just sentient for a minute. Don't worry. It'll pass.

Re: Science’s Biggest Fail

#253

Post by TeMPOraL » Tue, Feb 03, 2015, 4:56 PM UTC

api wrote: post #252

TeMPOraL wrote: post #249

Earlier quoted context omitted.

> Mostly though I feel very cynical and depressed by what appears to be the cultivation of ignorance as a virtue. I am convinced that we are fucked, barring a massive breakthrough in photovoltaic efficiency or nuclear fusion. I share that. I actually suffered a short-term deppressive period because of that (as weird as it sounds, for few months I felt guilty and afraid whenever I turned on the gas stove). We have may…

You were just sentient for a minute. Don't worry. It'll pass.

Well if that is how sentience feels then I'm not surprised many prefer to live like brainless zombies. I was petrified.

Re: Science’s Biggest Fail

#254

Post by faske » Tue, Feb 03, 2015, 6:37 PM UTC

This is refreshing to hear that I am not the only one who see this happening to the general public. This is a good video that explains how to determine which "science" is trustworthy or not.

https://www.youtube.com/watch?v=y1RXvBveht0

Re: Science’s Biggest Fail

#255

Post by snowwrestler » Tue, Feb 03, 2015, 7:41 PM UTC

msandford wrote: post #251

snowwrestler wrote: post #240

Earlier quoted context omitted.

That is why it is so difficult to predict the weather. But you don't need to predict the weather to predict changes to the global climate. Just like you don't need much of a model to predict that if a rogue planet passed through the solar system, existing orbits would be perturbed. Calculating the precise perturbations would be complex, but not predicting their existence. Likewise, it's difficult to predict exactly h…

Just because you can make a simplification doesn't mean you have achieved understanding. That understanding might be correct, or it might not. Even if you think you know the error bars are small, that doesn't mean that they are. Given that the earth has had temperature excursions of +4C and -4C in the past 100k years it seems that there are some kind of forces that keep the temperature within that range. What are the…

That's what a planetary climate is: an energy equilibrium. But if a forcing or feedback changes, the equilibrium will change.

The history of Earth's climate shows the limits of these changes, but that doesn't mean there are guard rails keeping us safe. It's just a reflection of the physical limits of the forcings and feedbacks. Our orbit only varies so much. Our axial tilt only wobbles so much. The sun's output only varies so much. The atmosphere can only hold so much water vapor at a given temperature. Etc.

If the sun suddenly doubled in brightness, I doubt you would expect the climate to still stay within that +/- 4C range, right? Well, atmospheric gases are also a forcing, one that we know we're changing. Granted--not at that level, but change is change. Models help us think about the sensitivity to that change, but the equilibrium must shift, somehow.

Anyway, the Earth's climate could stay well within that +/- 4C range, and the warming would still cause mankind a ton of trouble if it happens too fast. Remember that these are global average energy levels...it doesn't take much to raise sea level a troublesome amount, for example, especially since societies have built right at the water line all over the world.

Re: Science’s Biggest Fail

#256

Post by shanusmagnus » Wed, Feb 04, 2015, 2:57 AM UTC

msandford wrote: post #248

shanusmagnus wrote: post #241

Earlier quoted context omitted.

Dude, I promise you that the reality I want to be true is not one where we have fucked up the earth's climate. I will only reiterate that if 1000 brain surgeons say X, you should probably believe X unless you have very particular reasons to believe you are way smarter than all of them and know things about brain surgery that they don't know. And of course once in a great while that happens. There was a guy who first…

Yeah I hear you loud and clear. I get it. You believe this to be true, and/or that the confidence interval is very tight. I think the confidence interval is large enough that "warming is caused by CO2 and we gotta stop" is plausible but not proven yet. How on earth do I have the arrogance to say this? 1. I'm trained as an engineer and a scientist. I've read a lot of papers and I've got a good grasp of significance. 2…

I get your objections to my objections. I appreciate your points and the general atmosphere around the discussion. So now it's pretty well-established that neither of us is gonna change the other one's mind, so it's fine if it ends here.

But I'm curious about something you said and would be interested in hearing you expand a bit on it if you're willing. You say (and I believe you) that you don't think there's a giant conspiracy; and you give an example of the Honda never revving past 3000, which, if one were to make the error you describe, is basically an illustration of a linear model with insufficient leverage to support the conclusions that are drawn from it. The Honda example is a good illustration of that problem, and I will steal it in the future.

For now, though, based on this example, and your point #2, I'm wondering how you think these kind of fundamental statistical and logical errors could be made on such a massive scale, by a huge group of scientists, distributed around the world?

The only way I can imagine such a thing being feasible is via some serious groupthink / path dependency; like if Einstein, Maxwell, and von Neumann were raised from the dead, transformed into climate scientists, and then, at the very start of the very first climate conference, one right after the other got up and loudly and confidently expressed certainty that climate change was caused by X. Such an event could theoretically lead subsequent scientists to parrot the new party line, and to have difficulty getting funded for research that diverted from these pre-conclusions about X.

That scenario obviously didn't happen. It seems exceedingly unlikely that anything comparable could happen on this scale. Of course, from time to time certain topics of inquiry go in or out of fashion, like the neural winter of the 90s and 2000s, which has ironically now reversed itself with the whole deep learning thing. But that happened because funding was coming principally from US funding agencies, the computational power and data was inadequate to demonstrate the value of the nascent techniques, and the body of research was overwhelmingly produced in US institutes, none of which apply here.

So I'm interested to know by what mechanism you think so many smart people who are also engineers and scientists, except also specialists in this very topic, are making fundamental logical and statistical errors?

Re: Science’s Biggest Fail

#257

Post by stephengillie » Wed, Feb 04, 2015, 3:43 AM UTC

loqi wrote: post #176

apl wrote: post #164

Earlier quoted context omitted.

Do you have any decent sources on that one?

Didn't hang onto any links, sorry. The stuff I saw will probably turn up in a search for "alcohol" and "total mortality", though.

http://www.psmag.com/health-and-behavior/truth-wont-admit-dr...

http://content.time.com/time/magazine/article/0,9171,2017200...

Re: Science’s Biggest Fail

#258

Post by lotharbot » Wed, Feb 04, 2015, 5:47 AM UTC

api wrote: post #155

cookiecaper wrote: post #142

Earlier quoted context omitted.

>I understand the nature of scientific theory, and that scientific theory is not dogma or absolute revealed truth, but does the public? And was that ever communicated? Yes, people are generally not hostile to science and understand it's a best-effort work-in-progress-type thing. The problem is that the Science Zealots haven't gotten this memo, and go around flashing "studies" in peoples' faces (often completely witho…

The Science-as-religion movement, which I personally call fundamentalist positivism, is doing more harm to science than a million creationists and anti-vaccinationinsts and faith healers could possibly do. It undermines science from within by performing a kind of deep epistemological bait and switch -- replacing scientific epistemology at the root with religious epistemology while leaving the layers above superficial…

I've seen this a handful of times, where someone with legit science credentials will say something like "the lesson of science is" and then follow it up with an entirely philosophical conclusion. Like "... there is no purpose to the universe" (that comes from Dr. Jerry Coyne.) I'm not aware of any scientific experiment or framework to test the hypothesis that there is or isn't a purpose to the universe, nor even the possibility of creating such an experiment. But a fairly well renowned scientist made that claim -- essentially a philosophical-religious claim -- under the label of science.

I wouldn't say that does more harm to science than creationists etc. Instead, I'd say it contributes to the same mentality. It treats science as a label for a certain belief system, rather than a label for a set of processes and the data/explanations tied to those processes. As you say, it replaces scientific epistemology with religious epistemology -- stripping away the thing that makes science universal, and replacing it with something that makes it tribal.

Re: Science’s Biggest Fail

#259

Post by JabavuAdams » Wed, Feb 04, 2015, 7:14 AM UTC

olentangy wrote: post #160

JabavuAdams wrote: post #130

Nutrition is a weird example to pick on, because it's dominated by marketing, financial interest, media mis-reporting etc. Nutrition works like this: your boss reads an article in some rag and thinks that there might be an opportunity to target a new market. They tell you to go find some studies or something that could back up your claims. Usually, this would not survive any kind of rigorous scientific investigation…

Global warming is a weird example to pick on, because it's dominated by marketing, financial interest, media mis-reporting etc. Global warming works like this: your boss reads an article in some rag and thinks that there might be an opportunity to target a new market. They tell you to go find some studies or something that could back up your claims. Usually, this would not survive any kind of rigorous scientific inve…

I don't understand what you're trying to say.

Re: Science’s Biggest Fail

#260

Post by totony » Wed, Feb 04, 2015, 8:42 PM UTC

erroneousfunk wrote: post #44

ohazi wrote: post #13

"How do you make people trust a system that is designed to get wrong answers more often than right answers?" You can start by teaching kids how real science is actually done in order to get them to understand and trust the process rather than the we-only-deal-in-absolutes pop-news headlines.

Exactly. "Science" != "believing scientists" There's a component of reading research papers, trying to understand other people's work, etc., but, fundamentally, science works and moves forward independent of "belief" -- that's the whole point. You don't like that "science" told you to drink 8 glasses of water a day, and now they're telling you not to? Great, go read the research and figure out how much water you need…

What you say is false. Science is all about trust. There is way too much fraud being done just for quick fame.

Who wrote those research paper you take for granted, scientists? Are you saying that these papers are always right in what they affirm, or that, worst case, their result cannot be forged?