How do we detect design? V4!

I continue to update this item so I have started to put version numbers on it.

The intelligent design (ID) movement claims to have a formal method for detecting design by observing the results of the design process (this form of words get tedious so I will use the word “outcome” to indicate one or more events or objects that are the consequence of a designer at work). ID proponents have biology in mind, but they claim their approach works for detecting design in any sphere. There are many (including me) who disagree with the ID movement, but it is true that we can, and often do, detect design by observing the outcome. We can deduce with great certainty that Stonehenge is designed and not just a pile of rocks, and that an elaborate sandcastle is designed and not just the result of the tide. We can do this by inspecting the results – not by observing the designer at work. So anyone who wants to criticise ID ought to explain how we make this deduction, what makes it reasonable, and how it differs from the ID approach.

The ID inference is very simple in essence. ID proposes that outcomes may be the result of either necessity (the sun must rise because of universal natural laws); chance (it was bad luck that Katrina struck New Orleans); or design (the planes struck the World Trade Centre because that was the terrorists’ intention). Therefore, if we can rule out necessity or chance then the cause must be design. In fact most of the complexity in ID is deciding the rules for when chance can be ruled out. ID rules out chance on the basis that if the outcome is too improbable on the basis of chance then it cannot have been caused by chance. When the first plane struck the World Trade Center I think most observers put it down to an awful accident. When the second plane struck most spectators realised it was a deliberate plot. The chances of two planes accidentally hitting the World Trade Centre on the same day were so small as to be effectively impossible. The events of September 11th are a graphic illustration of the ID argument – some things are just too improbable to be accidents.

Now consider a different scenario. Suppose the twin towers had been hit not by planes but by two meteors, large enough to seriously damage the towers, and for the purposes of this fantasy suppose the meteors had no physical connection that we could discern - they were from different sources and were following unrelated trajectories. Such meteor strikes are less frequent than commercial airplanes flying into buildings – remember Schipol. So the probability of two such meteors striking the same location in the same day is even smaller than two planes striking the same location. Yet we would hesitate to ascribe the two meteors to design. We would certainly be less inclined to ascribe them to design than the two planes. We would ask questions such as “how could anyone control where the meteors land?” Instead of ascribing design, we would almost certainly want to look more deeply into the causes of these meteors to see if there was some factor that could explain why these two freakish events happened at the same place on the same day.

Can we describe this reasoning more formally and does it justify our intuitive reaction? I think we can.

First is important to recognise that when we decide between the causes of outcomes we compare the plausibility of possible causes (this sounds trivial but I believe it is vital – so hold on to it). Let’s first look at the process of comparing causes when design is not under consideration. For example, I might be trying to decide whether a damp patch is due to condensation or a leaky roof. I would do this by considering two things: how likely is it that each possible cause (condensation or a leaky roof) would lead to that particular damp patch in that location and how likely is it that there is condensation or a leaky roof in the first place. For example, I might start off thinking: “that is where I would expect to get a damp patch if the roof leaked and it seems unlikely that condensation would cause damp that far from the window ". But then suppose I remember that the roof has just been thoroughly checked and I had noticed that the room is prone to condensation in the past. This might well cause me to change my mind about the most likely cause. I cannot make a fair comparison of the alternative explanations unless I take into account both factors – the likelihood of the causes and the probability of the outcomes given the causes. Notice also that in order to assess the probability of the outcomes given the causes I want to ask questions about “how” – how could condensation lead to damp in that part of the room?

Can we extend this principle to the case where one of the causes involves design? Absolutely, and it immediately explains our intuitive reaction to the two twin towers scenarios above. So in the case of the two planes hitting the towers we compare the plausibility of it being a double accident to the plausibility that someone planned it. The second explanation wins by an enormous margin. The two collisions are far more likely if someone had planned them and there is a definite possibility that such people might exist and one can imagine an account of how they might have done it. We pretty much dismiss accidental explanations – even such explanations that might change the likelihood estimate (e.g. a fault in navigation systems that was used on both aircraft) because the awful explanation involving terrorists is all too plausible.

In the second case of the meteors we are much less likely to accept the design explanation – even though the likelihood of it being accidental are even smaller – because the idea of a designer who can control meteor strikes is so wildly implausible. We cannot conceive how they could do it. The probability of the outcome (twin meteor strikes) given the cause (someone who can control where meteors land) is actually very high – much higher than the freakish accident. But the likelihood of there being such a cause in the first place are so small that we dismiss that cause and in practice investigate the accidental cause more closely. Note, however, that if someone came up with a vaguely convincing account of how someone might control meteors then we would reconsider.
So it appears we can account for the way we detect design in just the same way we detect any other cause. We assess the probability of the outcome given the proposed cause and we assess the likelihood of the cause existing in the first place. Finally we compare these assessments to alternative explanations.

How does this compare to ID? ID concentrates on just one thing: the improbability of the outcome given that the cause is chance. It does not address the likelihood of the rival causes existing (chance or design). It goes even further than that. It explicitly forbids any exploration of design as a cause because it refuses to be drawn into who or what the designer is or how they implement their design. This means it provides no valid basis for deducing design. Ironically it also means that virtually all the discussion among ID proponents is about the improbability of chance and not the probability of design. They have nothing else to discuss! It is perhaps the greatest irony of ID that its proponents spend most of their time talking about Darwinian evolution. It is one of the biggest testimonies to the power of Darwin’s theory that even its greatest opponents cannot conceive of an alternative explanation.

Responding to a challenge from Paul

Paul (http://exilefromgroggs.blogspot.com/) has issued a challenge to ID opponents to come up with a way formalising the detection of design. I can resist this. But the response is rather long so I have put it here.

Paul

You are assuming that there is a universal way of detecting if something is designed. I would claim that you can only decide if something is designed given some context. For example, I am pretty sure Stonehenge was designed. Why? Because my experience tells me something about how stones generally get arranged in nature. (But you might be interested in http://w3g.gkss.de/staff/storch/pdf/misuses.pdf ) and also something about how people arrange stones intentionally. I can look at the result and compare the chances of the result happening given the two explanations. I can also investigate details of the process. If Stonehenge was designed it would have required extraordinary dedication and resources but it is feasible.

In another context it is far more difficult to make the call. Suppose I we discover a configuration of stalactites that happen to form a similar pattern to Stonehenge (roughly circular). Did someone make that happen by design or was it a freak of nature? We can’t decide this simply by studying the result. We have to look at the details of how the pattern might have been generated naturally or by design. Was there anyone around at the time the stalactites formed? How could they have implemented the pattern? Alternatively what natural process might have lead to the pattern? An interesting possibility might be that we discover that stalactites often form patterns like this even though we are not sure why. Do we now deduce there is a stalactite designer masterminding these patterns or do we simply say – let’s investigate how this might have happened?

There is another important aspect to this deduction. To estimate the probability of explanation X for outcome Y I need to do two fundamental things. One is estimate the probability of Y given X – e.g. what are the chances of getting this result if was designed? The other is to estimate the a priori probability of X. Then I can invoke Bayes’ theorem P(XY) = P(X)*P(YX). In the case of Stonehenge I know independently of Stonehenge that there were people around at the time capable and very likely motivated to build something of this type. So the probability of there being a designer P(X) is high. So both P(X) and P(XY) are high and I can reasonable deduce P(XY) is high. In the case of the Stalactite pattern I have no knowledge of any such designer so P(X) is low. The extreme case is when anyone claims that the world is the way it is because an omnipotent God made it that way. Now P(YX) = 1 – if there is such a God and it is truly omnipotent and it wants to make things this way then it is certain it would succeed. But P(XY) is not high because we have no a priori evidence that there is such a God who wanted things to be this way. P(X) is very low.

That’s a lot of writing – I hope it helps.

Cheers

Some Questions about Intelligent Design

I have to admit to being a bit of an Intelligent Design (ID) addict. I believe it is utterly wrong and should not be taught as science. I believe that there is overwhelming evidence for neo-Darwinism (ND) and that while there are some phenomena that ND finds it difficult to explain, this no more destructive to ND than an inexplicable orbit for a planet would be destructive to Newton’s laws of motion. I am also concerned that the rise of ID may lead to the rise of more sinister movements which may lead to the degradation of the scientific approach which has brought human kind so many benefits (and this is not just a problem for the USA - see http://education.guardian.co.uk/higher/news/story/0,,1714171,00.html). Nevertheless, the questions that ID raises, and forces us to answer, are interesting ones which deserve to be taken seriously. I would certainly approve of discussing the questions in a philosophy class as long as the teacher understood the science and the issues.

However, I have some questions about ID. I expect someone has already asked these questions and that someone else has answered them. But I have never seen the questions or the answers. Obviously I think they pose problems for the ID movement, but I am hoping that an ID proponent will take this in the spirit of enquiry and not as personal attack and respond appropriately. So here goes:

1 What exactly is being designed?

Is it, for example, an individual organism, a species, a mechanism such as blood clotting, or just the process which generates all of this? How much did the designer have in mind and how much did the designer just design the process and then see how it turned out?

2 In the case of biological phenomena what is the specification?

This question requires some explanation. William Dembski uses the concept of complex specified information as proof of design. This can be explained using the Dembski’s own example. Suppose an archer shoots at an arrow at a large wall. The arrow hits the wall and then the archer draws a ring round the arrow and declares he has hit the bullseye. The chances of the arrow hitting that exact point in the wall are tiny (or at least they were tiny before he shot the arrow). However, this tells us nothing about the archer’s skill. Compare this to an archer who draws a target on the wall and then shoots a hundred arrows that all hit the target. Both outcomes are highly improbable but the second is the result of the design of the archer, the first is not. Dembksi says the outcome in the second case is specified because the outcome fits a pattern which is independent of the event itself. The circle was drawn on the wall before a single arrow had been shot.

This is all fair enough. But how does this extend to biological phenomena? Suppose we observe a biological system such as the much discussed flagellum of bacteria, or the system of blood clotting, and we decide that it is indeed highly improbable that such a system could have evolved according to standard Darwinian processes (I don’t accept this – but assume it for the sake of the argument). So far we have the equivalent of an arrow in the wall. For Dembski to use this as evidence of design he needs the equivalent of the circle which has been independently drawn. Most intelligent design proponents would say the independent pattern is that the mechanism fulfils its function in the organism. E.g. how extraordinary that this combination of proteins should combine to allow blood clotting. But this needs defining in more detail. Does this mean a mechanism using this particular combination of proteins? In which case it is equivalent to drawing the circle after the arrow has been shot. Or is it something broader? Perhaps any mechanism that allows for blood to clot? Or any mechanism that prevents excessive bleeding in case of injury? Or any mechanism that fulfils the functions of blood without excessive risk from injury? Each one is a broader target, equivalent to bigger target on the wall, and allows for more possible solutions. Furthermore as soon as the specification moves to something broader than this specific combination of proteins it becomes pretty much impossible to calculate the probability of hitting it.

3 How can you conclude that because a phenomenon appears improbable according to a ND explanation that it is improbable for all chance explanations?

I believe that an ND explanation is by far the best explanation for the evolution of most biological phenomena and those phenomena that appear difficult will either be resolved according to ND or by some slight modification of ND. However, suppose one day an incontrovertible counterexample is found e.g. a large number of new fossils destroying the picture of divergence from a common ancestor (mammalian fossils in the pre-Cambrian etc). Or let us suppose for the sake of argument that systems such as blood clotting are shown to be quite impossible to evolve through ND methods. How, do we jump from this to ID? Remember the design argument requires that an outcome be improbable and specified. It is not sufficient to show that an outcome be improbable according to one chance explanation. It has to be shown to be improbable according to any chance explanation. If evolutionary biologists should come across serious counter-evidence to ND they would seek alternative explanations or modifications of the ND e.g. Larmarckism. These would be proposed processes which could then be tested by inspection of molecular biology, the fossil record, taxonomy etc. Why should they conclude that there is no chance explanation because ND failed?

It is perhaps a testimony to the power of the ND explanation that even ID proponents cannot think of alternative except to invoke an unspecified designer using an unspecified process.