Technology

From the point of view of society, technology is the reason for doing science.

Calculating Probabilities

When it comes to religion I tend to use the  'three kinds of people" model. Roughly speaking, you're either a theist, an atheist, or an agnostic. Which of these three is the most defensible position? Which makes the most sense? Which is most consistent with science? What criteria should be used to evaluate your argument for your position on the question of whether there is a god?

I'm not sure that I know the best arguments for each of these three positions, but here's my best shot at evaluating them.

Theists have a problem when it comes to science. They're probably just never going to have anything like scientific evidence for their position. Depending on the particular brand of theism, it may be logically impossible for them to have evidence. For example, one form of theism says that god, being omnipotent and omniscient, can manipulate the universe such that his presence is simply not detectable by scientific means. If you're that kind of theist then it's no surprise to you that science isn't finding god. You're problem, then, is an epistemological one: how do you know that there is a god? Do you have some method for creating knowledge that is different from science? Or do you just choose to believe, even though you don't actually know?

Atheists also have a problem when it comes to science, and for exactly the same reason. The proposition that there is a god who can hide the evidence of his own existence is completely unscientific. It cannot be addressed by science. Atheists will tell you that the claim that there is a god is similar to the claim (without evidence) that there is a teapot in orbit. However, these claims are not similar. The god-claim is an attempted explanation of observed phenomena: there is a universe, there is life, there are sentient beings. The god-claim says these things are here because god made them. The teapot claim has no connection to observed phenomena. Atheists say that we can dismiss the teapot-claim because the probability that it is there is vanishingly small. Without evidence of the teapot, or a reason why it should be in orbit, this is absolutely correct. But we can't make the same argument against the existence of god because, 1) if the god-claim is true then we should expect to find no evidence of god (a weak point, but not able to be defeated), and 2) the god-claim explains things that are currently unexplained, or inadequately explained. Ultimately, the problem with the atheist position is also an epistemological one: how do you know there isn't a god?

Agnostics take the easy way out. They admit that there is no scientific evidence for the existence of a god, but also admit that this doesn't mean that there actually isn't a god. As a result, agnosticism embraces a wide spectrum of people from near-believers to near-disbelievers.

So, a question for those of you who are theists: Do you admit that there is no scientific evidence for the existence of a god? And if so, what method for creating knowledge have you used to discover that god exists?

A question for atheists: If you reject the proposition that there is a god, not because you can prove there is no god, but because you believe that it is highly improbable that there is a god, then how do you compute the probability that there is a god?

Agnostics get a pass.

This Blew My Mind

The following is an excerpt from a fascinating article in Wired.

Dunbar tells the story of two labs that both ran into the same experimental problem: The proteins they were trying to measure were sticking to a filter, making it impossible to analyze the data. “One of the labs was full of people from different backgrounds,” Dunbar says. “They had biochemists and molecular biologists and geneticists and students in medical school.” The other lab, in contrast, was made up of E. coli experts. “They knew more about E. coli than anyone else, but that was what they knew,” he says. Dunbar watched how each of these labs dealt with their protein problem. The E. coli group took a brute-force approach, spending several weeks methodically testing various fixes. “It was extremely inefficient,” Dunbar says. “They eventually solved it, but they wasted a lot of valuable time.”

The diverse lab, in contrast, mulled the problem at a group meeting. None of the scientists were protein experts, so they began a wide-ranging discussion of possible solutions. At first, the conversation seemed rather useless. But then, as the chemists traded ideas with the biologists and the biologists bounced ideas off the med students, potential answers began to emerge. “After another 10 minutes of talking, the protein problem was solved,” Dunbar says. “They made it look easy.”

When Dunbar reviewed the transcripts of the meeting, he found that the intellectual mix generated a distinct type of interaction in which the scientists were forced to rely on metaphors and analogies to express themselves. (That’s because, unlike the E. coli group, the second lab lacked a specialized language that everyone could understand.) These abstractions proved essential for problem-solving, as they encouraged the scientists to reconsider their assumptions. Having to explain the problem to someone else forced them to think, if only for a moment, like an intellectual on the margins, filled with self-skepticism.

I think there is really something in the idea that reasoning by analogy, though potentially a logical mistake, can spur creative and new understanding. And I definitely agree that being forced to explain without reference to standard models helps to advance one's thinking.

I'm Starting To Get It

After considerably more reading, this is the way I'm understanding it. Let me know what I'm missing or getting wrong.

The key to Bayes’ theorem is that it makes explicit the relationship between two events, and the probability of one event given that the other event has occurred.

Here’s a simple example: Say you want to know if it rained last night, so you go outside and touch the grass to see if it’s wet. Before you touch the grass there is some probability that it rained during the night, maybe based on a forecast of 40% chance of rain. But after you touch the grass and feel that it is wet, there is a new (higher) probability that it rained. Bayes’ theorem gives you a way to calculate the new probability that it rained last night, given the evidence of wet grass.

You can’t conclude that it rained last night based on the fact that the grass is wet, because the sprinklers may have come on or it may just be dew. But if you know something about the relationship between the event of the grass being wet and the event of it having rained, then you can calculate how likely it is that rain is the cause of the wet grass (CORRECTION - that reference to 'cause' is objectionable, we're only discussing correlation). The real insight of Bayes’ theorem is that the likelihood that rain is the cause of the wet grass is related to the probability of the grass being wet when it hasn’t rained.

If you happen to know that there is a very low probability of the grass being wet in the morning after a night with no rain (maybe you don’t have sprinklers, and you live in a dry climate with very little dew), then wet grass is a strong indicator of rain. But if there is a high probability of the grass being wet on a morning after no rain, then wet grass is a very weak indicator of rain.

This is fairly intuitive, and so maybe it doesn’t seem very revolutionary. However, what Bayes’ theorem does is it makes this kind of reasoning explicit and calculable. Bayes’ theorem justifies this kind of reasoning by formally spelling out how and why it works. An understanding of Bayes’ theorem also helps one avoid mistakes of probabilistic reasoning, e.g. thinking that wet grass is a stronger indicator of rain than it really is. 

Science Ruckus

Quick post: there's a commotion going on in the comments over at ThinkMarkets. Do you have an opinion about what counts as Science?

Scientific Claims and Falsifiability

In my previous post I argued against an economist's criticisms of scientist's claims on truth. Arnold Kling does a much better job of making a similar criticism stick.

Arnold's argument boils down to this: Scientists sometimes claim that they will soon make particular discoveries or uncover particular knowledge, when in fact there is great doubt about whether they will.

These claims of impending success are common, and commonly wrong. More to the point (as Arnold suggests) these claims are unscientific, and therefore unbecoming to professed scientists, precisely because they are non-falsifiable. There is no theoretically possible method to disprove these claims (at the time they are made).

Science deals exclusively with claims that are (in principle) falsifiable. This is an extremely important part of why science works at all. By limiting scientific inquiry to falsifiable claims science is made testable and, ultimately, meaningful. If I believe that a scientific claim is false, I need only devise an experiment that will demonstrate a conflict or contradiction in order to prove that it is false.

Imagine what would happen if non-falsifiable claims were included under the domain of science. Under such a system scientists would concern themselves with claims that are mutually exclusive, but with no ability to distinguish between the truth or falseness of either. The discovery of knowledge would slow as time and effort were consumed in pointless and unending argument...

This is a concept that I'm sure all scientists are taught at some point early in their education. However, as important and fundamental as falsifiability is, many scientists seem to forget about it, or even become confused about what it means. Take, for example, Richard Dawkins' invocation of Russell's Teapot in support of militant Atheism. The bizarre thing here is that Dawkins uses Russell's Teapot as an example of a non-falsifiable claim, and therefore outside of scientific notice, while he simultaneously argues in favor of militant Atheism - which is a non-falsifiable claim as well. How can a scientist of such standing be so confused about one of science's founding principles?

Note: It  is important to recognize that Russell's Teapot may actually be a falsifiable claim, and therefore within  the purview of science. If you claim that there is a teapot in orbit, AND describe the orbit and the teapot in sufficient detail, then the claim is clearly falsifiable because the absence of the teapot in the specified orbit could, in principle, be observed. Militant Atheism, however is not a falsifiable claim because it is not possible, even in principle, to observe the absence of a god who can tweak the universe and human observation to achieve whatever end he/she/it desires. Militant Atheism must perforce fall outside of the notice of science.

Soft vs Hard Science?

Eric Falkenstein has made some sweeping generalizations about scientists. It's always a bit of a puzzle to me when I hear these kinds of arguments. When I look at the world around me I see the application of science everywhere. If scientists don't have a special claim to truth, then either my DVD drive shouldn't work, or communism should.

OK, that needs some explanation. First, my DVD drive:

Einstein discovered the principle of Light Amplification by Simulated Emission of Radiation (LASER, of course) while working on another problem. A few decades later his work was demonstrated to be correct when the first functional laser was built. Einstein's claim to truth is irrefutable, as are the claims of subsequent scientists and engineers who gradually refined the understanding of the concept until my DVD drive could be mass produced and sold to me.

What about communism? Well, economics is something of a science, but it's not a terribly successful one. Communism is a failure precisely because economic principles are not well enough understood to construct powerful and useful technologies for planning and coordinating the efforts of millions of people. Economics doesn't have a LASER equivalent. But, even though economics has not produced a lot of useful or even agreed upon knowledge, there are still many economists who have a need to publish in order to move their careers along. I think this is at the heart of what Eric Falkenstein is criticizing.

But why attack scientists altogether? To my mind, there is a genuine distinction between the hard and soft sciences, and that distinction is most visible in the technologies that emerge, or fail to emerge, from the knowledge that various kinds of scientists produce. After all, the purpose behind science isn't only to gain understanding, but to gain useful understanding that can be applied to better our lot.There's no need to conflate physicists with economists when their relative accomplishments are so distinct.

Alas, Eric doesn't note the distinction, and instead slanders 'physical' scientists while levying accurate criticisms against economists and other 'social' scientists. Maybe he does this because he's not prepared to countenance the disparity.

Does Natural Selection Account for All Biophenomena?

No. But too many people who like to dabble in evolutionary explanations assume that it does.

The book, How Women Got Their Curves and Other Just So Stories is a good example of this. Many hypotheses are put forward for this or that characteristic or behavior, all explained in terms of selective pressure acting upon our forbears. We are as we are precisely because our clever genes have tried out various reproductive strategies, and the cleverest genes have won out.

What the authors fail to consider is the missing evidence.

Hair color can illustrate what I mean. Has hair color been strongly selected for? Are brunettes more common than redheads because brunettes possess an evolutionary advantage? Probably not. After all, there are many different common shades and hues, and no obvious advantage of one color over another presents itself. One might point out that light-colored hair is correlated with light-colored skin, which does possess an evolutionary advantage for people in northern lands. But one needs only to travel to the Caucasus to find that light-skinned people can have dark-colored hair.

So, maybe hair color has not been strongly selected for. Perhaps it is merely the result of chance, isolated populations, and group identity that Swedes tend to be blonde while Han Chinese are nearly uniformly dark-haired. Indeed, a fundamental mechanism at work within evolution is chance mutation. Only after a feature has appeared, as the result of a genetic accident, can the feature become the subject of selective pressure.

Imagine, for a moment, a world in which a Great Calamity early in human history has by chance killed off the ancestors of all modern humans except for a small group who were to become the forbears the Chinese. In such a world, nearly everyone would have dark hair since we all would have descended from the dark-haired survivors. Scientists studying evolution in that world would perhaps, upon considering themselves and their fellows, conclude that it must be the case that there had been some strong advantage to their ancestors in having dark hair, since human chemistry could permit other hair colors but no living humans were in fact so colored. The missing evidence that could have revealed the truth, died with the proto-Europeans in the Great Calamity.

Such scenarios have in fact happened repeatedly throughout the history of life. Why are the creatures that have built universities, governments, and shopping malls descended from the ancestors of marmots instead of from the ancestors of falcons? Because of a chance extinction and climate change that gave mammals an opening against dinosaurs.

Indeed, assuming that a trait has been selected for merely because it exists, or even because it is common or exclusive, is bad science. Science demands evidence to connect premises with conclusions (reasoning alone is not sufficient). Evolution is powerful science, and may be invoked in the work of others who are seeking to describe the world, but there is more to Evolution than selective pressure.