Reasons to believe
I’ve been thinking recently that I believe in the Theory of Evolution on about the same level as in the Theory of Plate Tectonics. I have grown up being taught that both are true, and I am capable of doing research in either field, or at least reading the literature to examine them for myself. I have not done so in either case, to any reasonable extent.
I am not swayed by the fact that some people consider the former (and not so much the latter) to be controversial, primarily because those people aren’t scientists. I tend to be self-congratulatory about this fact, but then I think that I am essentially not interested in examining the evidence, but I am essentially taking it on faith (which the creationists are quick to point out). I think I have good Bayesian reasons to take science on faith (rather than, say, mythology that is being offered in its stead), but do I therefore have good reasons to accept a particular well-established scientific theory on faith, or is it incumbent upon me to examine it, if I think its conclusions are important to my life?
In other words, is it epistemologically wrong to rely on an authority that has produced a number of correct statements (that I could and did verify) to be more or less correct in the future? If I think of this problem as a sort of belief network, with a parent node that has causal connections to hundreds of children, I think such a reliance is reasonable, once you establish that the authority is indeed accurate. On the other hand, appeal to authority is probably the most famous fallacy there is.
Any thoughts? If Eliezer or other people have written on this exact topic, a reference would be appreciated.
Curiously chosen examples; I find the two quite disanalogous. With plate tectonics, I could listen to an expert give a detailed but made-up account which wouldn’t ring false in an immediate sense, with evolution (at least at the basic level of iterated mutation-selection cycles) any alternate theory lacking its basic dynamic would contradict not only authority figures and the actual evidence, but also fundamental notions about change and time, not to speak of the many daily-life encounters with evolution in different incarnations (from flu vaccines to social dynamics).
If evolution turned out to be a hoax, I’d feel lost in my understanding of reality in deeply permeating ways, much more so than if the more insular theory of plate tectonics turned out to be false. You can’t introspect your way towards a theory of plate tectonics (at least I can’t) in the same way as you can infer that when there’s change (take that Parmenides) and a benchmark, there’s evolution.
What you are talking about is a lay sense of evolution. Sure, things change, and the more adapted thing should survive with higher frequency, this much is obvious even to creationists. It is also obvious to me (as it was to Aristotle), that things which are in motion tend to come to rest. Turns out, it’s not really true. Just because a theory is intuitive, doesn’t mean that’s how the world really works. You only need to think about Heliocentrism, let alone something like quantum physics.
One problem that Darwin had was the lack of mechanism for evolution (i.e. genetics). If I were alive at the time he wrote his books, I would have liked his theory, but would have been forced to acknowledge that it does not truly explain how the world works. I am told that now this is all solved, but have been taking that largely on faith.
It also may be that the theory is wrong, but there is no better theory to replace it, a la physics in 1900. If that were the case, I’d like to know that too.
We should expect some amount of evolution by natural selection ‘a priori’, from various obvious premises such as
(1) There is a reproduction process in which characteristics are inherited (2) Things with X characteristics in Y environment die/live etc.
There seems to be an absence of similarly parsimonious explanations, and the account given by natural selection is compelling. I suspect that even a small amount of knowledge of the empirical evidence for natural selection would establish a lower bound on the share of evolution it causes, such that searching for equally significant factors for evolution of life in general should be expected to fail.
If one set up a mathematical representation of a population that took into account characteristics, life, and death, etc. then natural selection would be the name for a provable behaviour of the system, even if the system were just axiomatised by more basic facts such as (1) and (2). I’m not convinced that the same is true of Aristotelian physics.
I struggle far more to fabricate accounts of our observations without natural selection than I did to get to grips with Newtonian mechanics. As in, accounts that don’t leave me more confused (e.g. ‘God did it’, which is a mysterious non-answer).
Quantum mechanics I do not know well enough (and I’m not sure anyone does) at the level where mathematical reductionism meets theoretical physics, but I would not be surprised if it turned out to be extremely parsimonious given even a small number of our empirical observations.
Heliocentrism also seems much more contingent than natural selection, although possibly less than one thinks, given how prevalent star-planet systems are.
This merely says that the surviving characteristics will move towards optimal in an unspecified range, assuming there are no other significant influences.
It is not explained how exactly the things with less likely characteristics have appeared in the first place. Is there perhaps another process that causes non-optimality? How could you know the process is not actually stronger than the process of natural selection?
It is also not obvious that this process of improvement can cross boundaries between species. You proved that a healthy dog is more likely to survive than a sick dog. You didn’t prove that the super-healthy dog will evolve to a lion or an eagle.
So this naive observation is actually not in contradiction with creationism. (Of course only after the creationism is updated to include it. But hey, scientists change their opinions, too.)
To move further, we need to know that the information which encodes the organism can change randomly (mutation), and that all species use the very same mechanism, so it is possible in theory to modify the information step by step starting with dog and ending with an eagle, in each step getting organisms which would be optimized for some environment. We need some kind of “universal DNA” hypothesis, even if we don’t know the exact formula for the DNA. And it’s not really obvious that dogs and eagles and fish and trees are all encoded in the same way.
On the other hand, once we have a microscope, the “universal DNA” hypothesis becomes rather simple to prove, because it is enough to explore what we have here and now.
I agree that understanding the extent to which natural selection generated things is easier nowadays than it was previously.
I’m not concerned here with logical disproofs of creationism, moreso with something like arguing that time spent {worrying that evolution by natural selection is significantly overrated} is probably misallocated, or arguing that people are privileging the question of ‘Is evolution by natural selection a good explanation?’ (I’m not entirely sure what my motivation is, but it feels like it’s at least partly something along those lines.)
I also think that there’s a lot of hard-to-enumerate background information available to a person nowadays that should be enough for a naturalistic reductionist to intuit that evolution will not arise from bolted-on complex mechanisms like creation, but rather from mechanisms like natural selection that are inherent to populations with a few basic properties that we have actually observed (e.g. finite lifespans that vary according to characteristic-environment interaction, inheritance of characteristics, etc.). It’s possible you understand this and I’m misinterpreting the point at which you’re challenging my argument, but I would very strongly disexpect somebody thinking like me along the lines of ‘provable properties of populations under observed axioms’ (or the probabilistic/continuous generalisation thereof) to be talking about creationism; rather, I would expect challenges to come in the form of other equally basic, low-complexity processes that arise from what we already know.
I’m not sure DNA/the exact method of inheritance is relevant; I would still consider it to be a win for natural selection if we had, say, Lamarckism/inheritance of acquired characteristics/epigenetics as a significant force.
Well, we have tons of observation that nature can be understood, and that the useful explanations do not involve gods or magic. So yes, it would be a reasonable prior expectation for the origins of species, too.
I agree that certainly some evolution would follow from your premises (1) and (2). But imagine that we also have independent evidence that Earth is 1 million years old. In that case, I’d be forced to say that the Theory of Evolution can’t account for the evidence of life we observe, given mutation rates, etc. This is the sort of thing I am worried about when I say I haven’t looked at the evidence. As far as I know there isn’t any contradictory evidence of this sort, but there may be specific challenges that aren’t well-explained. Creationists like to cite irreducible organs and claim that those exist (i.e. where it can’t evolve from anything that has any evolutionary advantage) and are contrary to the theory. I know about this objection, but it would be a lot of work to truly evaluate it in depth.
As far as having an alternative: this isn’t necessary. I’d be reluctant to go with “God did it”, so I’d be fine with “the theory explains 95% of the evidence, and about the other 5% we don’t know yet, and we have no better theory”.
Just realised you’re the post author, so: Thanks for posting this, it’s something I’ve wondered about in relation to myself, as well. :)
1: No tentacles
This reminds me of something Eliezer once said—”How would I explain the event of my left arm being replaced by a blue tentacle? The answer is that I wouldn’t. It isn’t going to happen.” We do not observe a young (even 10^6) Earth, and by suggesting the possibility of one as counterevidence against the strength of the ‘a priori’ reasoning I advocated, you must be smuggling in a circular assumption that young Earth models have significant probability.
Your argument as I understand it is roughly that since my a priori reasoning would fail in young Earth scenarios, then that reasoning is unreliable. But if our prior for young Earth scenarios is extremely low, then it will only very rarely happen that my reasoning will fail in that particular way. Therefore for your argument to go through, you would have to place a high prior probability on young Earth scenarios.
To put it another another way: If observing a young Earth would be evidence against my a priori reasoning, then by conservation of expected evidence, our actual observation of a non-young Earth must be evidence in favour of that reasoning.
People in a modern day situation, and LW’ers in particular, are better placed to understand that ‘naturalistic’ explanations are preferable, and that magic ones should incur huge complexity penalties. Therefore we should have low priors on young Earths, because most of our probability will be concentrated in models where intelligent life arises from nonintelligent (hence slow) processes as opposed to intelligent (e.g. God) processes.
Moreover, the more intelligent the process that generated us, the more we push the explanatory buck back onto that process. God is an extreme case where the mystery of the apparent improbability of human intelligence is replaced with the mystery of the apparent improbability of divine intelligence. But even less extreme cases like superhumanly (but still decidedly ‘finitely intelligent’) entities simulating us incurs a penalty for passing the explanatory buck back. Natural selection is so elegant and formidable because it is an existing behaviour of what we already observe.
2: Insanity screens off charity
If—even before accepting evolution by natural selection—you can put an extremely high probability on creationists spewing objections like ‘irreducible organs’ regardless of the veracity of evolution by natural selection, then you can pretty much write off the counterarguments of creationists, because your observation of creationists making these arguments is extremely weak evidence that there is anything to these counterarguments. Now, this is circular if your only reason for not taking creationists seriously is that they are wrong natural selection/irreducible organs, but there’s ‘any number’ of other reasons to suspect creationists are engaging in motivated cognition.
The same way that if natural selection provides a substantial portion of the explanation of evolution, then, you need look no further, if cognitive biases/sociology provide a substantial portion of or even all of the explanation for creationists talking about irreducible organs, then their actual counterarguments are screened off by your prior knowledge of what causes them to deploy those counterarguments; you should be less inclined to consider their arguments than a random string generator that happened to output a sentence that reads as a counterargument against natural selection.
I think you are interpreting my comments with too much emphasis on specific examples I give. Sure, Earth being 1 million years old is unlikely, but there could be some equally embarrassing artifact or contradictory evidence. I can’t give a realistic example because I haven’t studied the problem—that’s my whole point. You seem to be saying that the Theory of Evolution is unfalsifiable, at least in practice. That would be a bad thing, not a good thing. Besides, surely, if someone runs cryptological analysis software on the DNA of E. Coli, and get back “(C) Microsoft Corp.”, that would rather undermine the theory?:)
In actuality, for me it comes down to trust: I expect if there was important contradictory evidence, someone would report it. Creationists think that biologists are all in on a conspiracy to hide the truth and would not change their mind if they see such evidence—that is rather unlikely from my point of view. That is to say, like you, I am not spending a lot of time evaluating the underlying facts because I think one side is reliable and the other is not. But it feels wrong to me to ignore evidence because of who says it. I understand your argument that you expect some evidence to be presented by them and that makes it unnecessary to examine it, but I think you are wrong. You do have to examine it in case it turns out that their evidence is in fact overwhelming your prior. They could be right in a specific case even if it’s unlikely. Even a stopped clock is right twice a day.
Let’s be a bit more precise. Evolution is a mechanism. It works given certain well-known preconditions. The fact that it works is not contested by anyone sane.
What actually is contested by creationists is that the mechanism of evolution is sufficient to generate all the variety of life we see on Earth and that it actually did, in fact, generate all that variety. *That* claim is falsifiable -- e.g. by showing that some cause/mechanism/agency other than evolution played an important part in the development of life on Earth.
You’re steelmanning the creationist position. That’s fine, but by saying “what actually is contested...” you’re also asserting that creationists only believe your steelman and not the position it’s a steelman of.
There may be some who do.
However, there are plenty of creationists who think that evolution cannot work. Some of their arguments include: “Evolution would mean order comes out of disorder, which is impossible” and “Evolution is a violation of the Second Law of Thermodynamics.”
Sure, but we are not really talking about the creationists here, we’re talking about whether evolution is falsifiable and in which sense.
As the person who asked the question, I’d like to say that I don’t particularly care about what creationists believe either.
I think it better be true that both of these are falsifiable (and they both are). I agree that the former is overwhelmingly likely and no one I’d care to talk to disputes it. In any event I am only talking about the latter. The fact that it completely explains the variety of life on Earth is the very thing I am accepting on faith, and that’s what I don’t like.
I’ve just discovered Argument Screens Off Authority by EY, so it seems I’ve got an authority on my side too:) You can’t eliminate an argument even if it’s presented by untrustworthy people.
Argument screens off authority only if you have already considered the argument. This doesn’t mean you should consider all arguments by anyone.
In what sense does it matter for your life? If it matters because you want to do molecular biology, than it makes sense to understand evolution on a deeper level.
That’s a good point. I suppose it has no practical implications for me, except that I’d like to have an accurate model of how the Universe works. Although if I were a young-earth creationist, it would have mattered a lot.
But let’s take global warming. That one does matter in a practical sense.
When it comes to global warming there are two separete issues.
The first is talking about global warming. It’s good for a society is a broad public debates important issues.
On the personal level however the significance of being wrong about global warming is relatively low for most people. Given that’s low you can just go with what your favorite authority says. If you however work in a field where it’s not low, I would again recommend that you get a better understanding of the subject.
I agree that such a reliance is reasonable.
As Jaynes points out, many “logical fallacies” are actually examples of correct Bayesian inference.
Relevant: http://lesswrong.com/lw/aq2/
Thanks, that was interesting, although didn’t specifically address my question.
I wish people would stop conflating “probabilistic/statistical inference” and “Bayesian inference.” Bayesians do not have a monopoly on the use of probability theory.
It seems to me that Bayesians do have a ‘monopoly’ on the use of probabilities to express personal uncertainty, which is the topic under discussion. Are you worried that the phrase “Bayesian inference” seems exclusive, or something else?
Are you saying that if I am not sure what the average causal effect is, I better get my Bayesian union card? I am perfectly happy to deal with uncertainty using probability theory without such a card!
If you express “not being sure” in terms of numerical probabilities assigned to various levels of surety, or a numerical point estimate, then it seems reasonable to describe you as acting in a Bayesian manner.
It does not seem reasonable to insist on identification as a Bayesian to use any piece of the Bayesian toolkit. As I’ve said in the past, Frequentists don’t disagree with Bayes’s law. I hoped to implicitly disavow this view by putting monopoly in quotes.
It wasn’t clear to me what you thought was being conflated. There are ‘fallacies’ which are examples of correct inference using Bayes’s rule, and to object to calling that “Bayesian inference” seems odd to me.
Back when Pearl was arguing for the adoption of probability over logic in AI (for example, in the intro to his 88 book), he was talking about things like “rumor propagation” that logic calculus does not handle well, but probability calculus does. There is not much mention of B vs F in his defense, just properties of probability theory itself.
What Jaynes’ quote is really about is that logic calculus and probability calculus are not the same. Logic obeys locality, probability does not, etc. Bringing issues of epistemology into this is either a confusion of issues or a deliberate bait and switch.
I publish in a conference called “UAI” (Uncertainty in AI) a lot. There is a word “uncertainty” in the title, but I assure you, the community is far from uniformly Bayesian.
Agreed, and I would go further and say Jaynes is making the claim that probability calculus is superior to logic calculus.
I was confused- I flipped “interpret probabilities as beliefs” and “interpret beliefs as probabilities.”
The issue I still haven’t resolved is whether you’re objecting to calling the use of Bayes’ law Bayesian inference instead of probabilistic/statistical inference, and why. I think I see why you would want to, but my speculation is unclear.
Yes I do (because Bayes’ law is a triviality—it’s like saying any use of fractions is frequentist).
I see where you’re coming from, and I agree that probabilistic inference is a more precise term that should be used instead. But I think my sense of shouldness there is much smaller than your sense of shouldness, and I can’t tell if that’s more tolerance for verbal sloppiness on my part or some sort of regret at narrowing the term Bayesian inference. I think I endorse the first but not the second explanation.
I just find LW’s often uninformed “Bayesianism” annoying, that’s all. It’s classic belief as attire—I am willing to bet having an opinion on these sorts of things make zero (0) practical difference in most people’s lives.
In fact, lots of things in the “LW cluster” are like this—MWI for instance.
edit: lest you think I am joking, here’s someone who calls the g-formula “an ad hoc frequentist device”:
http://lesswrong.com/lw/hwq/evidential_decision_theory_selection_bias_and/9crr
Here’s a post by Nyan Sandwich
http://lesswrong.com/lw/irj/crush_your_uncertainty/
which amounts to saying “get more data!” but which starts as follows:
“Bayesian epistemology and decision theory provide a rigorous foundation for dealing with mixed or ambiguous evidence, uncertainty, and risky decisions.”
which is a complete non-sequitur to the main point. I could keep going, but I hope the point is clear. There is an alarming lack of clue coupled with an alarming amount of “rah rah Bayes”. (Note: in case it is not obvious, I have no horse in the B vs F race at all, I am happy to write papers in either formalism. My point is not about B vs F at all.)
Do you also get annoyed when people say “prior” when they mean “probability estimate”? As in, “my priors have been updated”.
One could affirm probabilism while denying (especially stronger forms of) Bayesianism (without contradiction), and, in fact, many philosophers do.
At a first cut, you want confidence that it’s the same type of question you’re asking now that they’ve asked in the past.
I am not sure I got that. Is “the question I am asking now” referring to a theory whose truthfulness I am evaluating? And “the asked in past” the ones whose truthfulness I have verified? It’s confusing because chronologically it’s the other way around: most of these theories are old and were accepted by me on faith since school days, and I could only verify a few of them as I grew older.
Yes and yes.
Asking now = not yet verified. Asking in the past = already verified.
It would have been better for me to say:
Which statements pertinent to both evolution and plate tectonics did you verify?
Essentially none. I have a lot of evidence of science being right (at least as far as I can reasonably tell) in some other subject areas such as parts of physics, chemistry, cognitive science, etc.
I’ve read some FAQs on both, but it doesn’t count as verification. I suppose I can look at the map of S. America and Africa and see coastlines roughly match, that is some evidence for plate tectonics. Also, as I mentioned in reply to other comments, it seems correct that with genetics being right (that I strongly believe), natural selection would certainly work to cause some species to change. I think even creationists nowadays are forced to agree with this.
Look at something like psychology. If you’d deferred to the leading authorities over the past 100 years, you would have been an introspectionist, then a behaviourist, then a cognitive scientist and now you’d probably be a cognitive neuroscientist. Note that these paradigms primarily differ on what they think counts as evidence, rather than quality or quantity of evidence. They all performed experiments. They share many of the same experimental methods. They all had numerous results they could point to and a neat story about how the same method could be carried on to explain everything else.
Unfortunately, the authorities get divided up into schools of thought before even they have examined all the alternatives. Typically the mainstream school has a way of dismissing alternatives without examining them. A school can become mainstream for all sorts of reasons (it provides ideological support, it’s sexier, there’s a lack of alternatives, mere persistence, it has charismatic advocates, etc). So I think you have to be very careful who you take to be an authority on a given subject. Assessing authorities probably isn’t much easier than assessing the subject directly.
I think you are right, but is it so bad? If I were living at the time of the introspectionists, was there a better alternative for me? I suspect that unless I personally worked out some other theory (unlikely), I’d have to either take that one or something equally bad. Maybe it’s slightly different around boundaries of these paradigm shifts where I could possibly adopt the new ideas before the mainstream did, but most of the time it wouldn’t happen. I am far from being confident that I’d do a better job personally then the general consensus, even if that tends to be very conservative.
Once, for a Wittgenstein course, I read through the entirety of William James’s 1890 Principles of Psychology. It was of course absurdly outdated, but I learned a lot from it. One of the things was surprise at how much time James felt he had to spend in the book attacking theories involving souls.
So yes, you could do much worse than being an introspectionist.
I’m not sure about introspectionism, but I’m sure you could find theories that have produced bad outcomes and had mainstream acceptance, particularly in medicine. I suppose the alternative is to remain noncommittal.
That’s very useful, actually. I think I have a tendency to just accept the latest medical theory/practice as being the best guess that the most qualified people made with the current state of evidence. Which may be really suboptimal if they don’t have a lot of evidence for it, and perhaps it should be independently examined if it concerns you personally. I am not sure what degree of belief to assign such things, though, because I have no experience with them.
Do you, or anyone, have an idea of how trustworthy such things generally are, in the modern age? Are there statistics about how often mainstream approaches are later proven to be harmful (and how often merely suboptimal)?
Sir Karl Popper argued against induction and against authority in Conjectures and Refutations (and in most of his books). Provisional trust in scientific claims can be useful, but all conjectures are subject to critical examination and perhaps refutation. Popper said the scientific method is defined by use of falsifiability, not by authoritty.
That’s the standard scientific point of view, certainly. But would an Orthodox Bayesian agree?:) Isn’t there a very strong prior?
Speaking of falsifying evolution—a new paper (via Yvain’s blog) hints that maybe Lamarck wasn’t completely wrong after all? If so, the standard Theory of Evolution will need to be… adjusted :-)
What probability do you place on your level of “faith”? 10%? 50%? 90%? And at what threshold of probability should you move from “faith” to “knowledge”? (I don’t think there is a threshold).
(I put “faith” in scare quotes because it’s one of the most often abused words in these sorts of discussions concerning evolution vs. creationism. It could really benefit from being taboo’d )
I don’t think I can have “knowledge” in Science. It’s done by humans, therefore it makes errors. For any given proposition, if I examine the evidence and find it compelling, sure. But my whole point is whether I can rely on it without specifically examining it.
By default humans don’t necessarily place probabilities on their beliefs. When it comes to people who don’t faith seems to be a reasonable word.
It actually doesn’t take too long to get a decent grasp of evolution. Just reading the wiki page in detail will probably take less than an hour, and give you a decent grasp on it. Possibly also read some of Eliezer’s stuff on how people get evolution wrong.
If you want to go a bit further, I’m sure you can find a middle school or high school level textbook that could explain more.
It doesn’t take to long to get a decent grasp of Le Sage’s theory of gravitation either. Most problems have solutions that are clear, simple, and wrong.
I’ve been thinking about this sort of thing as well. There are lots of books published by creationists and I am sure they are quite compelling (I haven’t actually read those either), otherwise they wouldn’t write those. Essentially, reading someone’s summary is again putting yourself into the hands of whoever wrote it. If they have an agenda, you’ll likely end up believing it. So, really, you need to read both sides, compare their arguments, etc. Lots of work.
The problem with textbooks is that they are the “pink slime” of publishing. College textbooks are slightly better than grade school textbooks, in that they have slightly greater than zero market forces acting on them (not many, but some) and can assume an adult level of comprehension.
See also this paper about common misrepresentations of evolution in textbooks and science literature.