IBE vs. Falsificationism
One hypothesis or Two (or more)?
• One big difference between falsificationism
and IBE is that IBE requires at least two competing hypotheses, whereas falsification can work with a single hypothesis.
• IBE says, in effect, “We shouldn’t reject H just because it has trouble predicting E, since the other hypotheses might be even worse!”
• Are there any cases (you can think of) where a single hypothesis is simply rejected, on account of some evidence, without any other hypothesis being put in its place?
• Such cases are problematic for IBE. They suggest that falsificationism is correct.
• What if, for example, you initially assume that a coin is fair (chance of heads = ½, and separate tosses are independent). But then you start tossing it, and get the outcome:
h t h t h t h t h t h t h t h t h t h t h t h t h …?
Surely this refutes the “null hypothesis”. But what alternative is there?
• The fact that the outcome has a “simple pattern” perhaps leads us to think that there must be some plausible explanation for it, even if we cannot think of it right now. (Some sort of bistable mechanism?)
• If the outcome were just some random‐looking string (which is equally improbable under the null hypothesis) then we wouldn’t see any need to look for an alternative cause.
• A detective is almost certain that Smith is the murderer. He’s about to make an arrest, but then Smith comes up with a cast iron alibi. CCTV footage places him definitively in a London Drugs store, many miles from the murder, at the time it must have occurred. Back to square one – there’s no other plausible suspect.
• Similar situations occur in science. A pattern might be observed, which cannot be explained by the present theory. Yet the pattern looks like something that should have an explanation.
• In biology, for example, examples of convergence are becoming increasingly numerous and striking. (Convergence is when the same structures, including proteins, evolve in separate lineages.)
Convergence
Marsupial and placental wolves
Vertebrate and cephalopod eyes
Prestin in bats and whales
• … two new studies in the January 26th issue of Current Biology, a Cell Press publication, show that bats' and whales' remarkable ability and the high‐frequency hearing it depends on are shared at a much deeper level than anyone would have anticipated ‐‐ all the way down to the molecular level.
• Stephen Rossiter of the University of London, an author on one of the studies. "Our study shows that a complex trait ‐‐
echolocation ‐‐ has in fact evolved by identical genetic changes in bats and dolphins.“
• "The results imply that there are very limited ways, if not only one way, for a mammal to hear high‐frequency sounds," said Jianzhi Zhang of the University of Michigan.
“…this suggests that the evolutionary routes may be much more restricted than usually thought. Of course the tree of life is vast and arborescent, ending in innumerable twigs. But if at any bifurcation the evolutionary possibilities are limited (as convergence surely indicates) then it might be that this tree is constructed on determinate principles.”
Simon Conway Morris: “Walcott, the Burgess Shale and rumours of a post‐ Darwinian world” Current Biology, Volume 19, Issue 20, R927‐R931, 3 November 2009.
• Conway Morris has been writing like this for a few years, hinting that evolution may be driven by endogenous (internal) factors as well as by the external forces of random mutation and natural selection. (In order to explain convergence.) But he hasn’t come up with an actual hypothesis. He just seems to feel that something more is needed.
Surprising Predictions
• What does IBE say about a hypothesis that makes surprising predictions, that turn out to be true? (e.g. Fresnel’s wave theory of light predicting the “Poisson spot”.)
• If an observation is predicted by almost every hypothesis being considered, then it isn’t surprising.
• A surprising prediction is one that is made by only implausible hypotheses.
• Hence surprising (and true) predictions are rather significant, as they favour hypotheses that were previously seen as implausible.
• In the Bayesian (probabilistic) version of IBE, a surprising prediction of H is when P(E) is very low, but P(E | H) is high. This means that P(H | E) is much higher than P(H). • P(H | E) = P(H) x P(E | H)/P(E)
Science and Prediction
• Remember Popper’s criterion to distinguish between science and metaphysics? A scientific hypothesis must be highly falsifiable.
• In order to be falsified by empirical data, the hypothesis must first make a (robust and precise) prediction about what will be observed.
• According to IBE, a good explanation must also make a precise and robust prediction.
A point of agreement?
• So perhaps Popper and IBE agree that one essential attribute of a scientific hypothesis is that it make robust and precise predictions.
Robust = the hypothesis really does make this prediction. It’s not a matter of subjective judgement. We all agree on it. There’s no backing out of the prediction.
Precise = the prediction is improbable, being satisfied by a very small set of possible data. E.g. predicting a value of 54.662, rather than 50‐60.
• On this “clear prediction” criterion of what counts as science, intelligent design seems to fail the test, since it’s not clear what kinds of organisms we would expect to see if ID was true.
• Unless we know a lot about the designer, it’s hard to predict what the designer would do. There is no scientific “divine psychology”.
Hand‐waving predictions
• How do we know what a theory predicts?
• E.g. Cartesian astronomy. Does the vortex theory predict that apples will fall from trees?
The tiny swirling particles of the vortex are forced outward by their rotation, and get pressed against the boundary with the sun’s vortex. This pressure gradient forces large bodies like apples inward, towards the earth.
• But notice that the particles of the vortex are forced outward. So why are apples forced inward? Is it because apples are bigger? Why does the size matter here? It’s vague, hand waving. I bet that, if apples did fall upward, then Descartes could predict that too.
Darwinian predictions?
• Berlinski (critic of Darwinism) claims that Darwinian biology doesn’t make clear predictions. There’s no mathematical demonstration, for example, that Darwinism predicts an increase in complexity.
• [video file 5:05]
• Compare with Miller_3, 6:13
“Firstly, evolutionary theorizing is rarely sullied by any specific predictions or retrodictions concerning organic events at any level of biological organization. Secondly, the theory seems to possess a disquieting amount of elasticity or flexibility with regard to explaining organic phenomena. Anything and everything in the empirical biological world seems to be compatible with evolutionary explanations. Refuting evidence or crucial experiments that could realistically jeopardize an evolutionary account seem extremely few and far between.”
Arthur Caplan, Erkenntnis 13 (1978) 261‐278.
• Caplan himself rejects this criticism of evolutionary theorizing:
“Nor does it seem fair to demand high philosophical standards of testability, systematic power, and predictability of scientific theories that seem to adequately serve the explanatory needs of empirical scientists for purely logical or philosophical reasons. The logical analyses of philosophers of science, which were undertaken as purely descriptively adequate enterprises, may have grown prescriptively unwieldy.” (p. 277)
(In other words, biologists are happy with their theory, so philosophers of science should lay off.)
(My own bias)
• I am working in this area of the origins of biological complexity. A paper of mine, “Self‐
Organisation in Dynamical Systems: A Limiting Result” was recently published in Synthese, a philosophical journal. (Sept. 2010)
• I argue that no physical theory, based on laws of the kind we know about, predicts novel complexity under any circumstances. It’s always “garbage in, garbage out”.
ABSTRACT
Self organization, or “order for free”, is an important (and expanding) area of inquiry. Self‐organized structures occur in many contexts, including biology. While these structures may be intricate and impressive, there are some limitations on the kinds of structure than can self‐organize, given the dynamical laws. (William Paley pointed out, for example, that a watch cannot be produced by “the laws of metallic nature”.) In this paper I will demonstrate that certain fundamental symmetries in the laws of physics constrain self organization in an interesting way. Roughly speaking, structures that are both large and non‐self‐similar cannot self organize in any dynamical system.