Tag: evaluation

A critical look at Intelligent Design

A critical look at Intelligent Design

Intelligent design is defined by the Discovery Institute as a theory holding ‘that certain features of the universe and of living things are best explained by an intelligent cause, not an undirected process such as natural selection’ (Discovery Institute, 2016).

The Philosophical Apologist has spent considerable effort evaluating whether Intelligent Design (ID) should be regarded as science, and it failed to meet the suggested criteria. ID should not be regarded as science, at least at this stage of its development. We’ve noted that this doesn’t necessarily matter, and this verdict has little bearing on whether ID’s arguments are sound. It’s now time to examine them.

Initial objections

It’s common for opponents to dismiss ID as religion, often because it has creationist roots, and creationism is certainly religious in nature. This is a lazy option, as ID explicitly avoids identifying its designer, and this approach does not address ID’s arguments.

Similarly, we shouldn’t dismiss ID on the grounds of it being supernatural. If there really is reason to believe there is a designer, it must be extraordinarily capable, orders of magnitude more so than humans. That certainly sounds non-natural, and even God-like. If it is God as conceived by theists, we’d surely want to know.

The design inference

We seem quite capable at reliably inferring design produced by intelligent agents, almost without thinking. By experience we’ve learnt how to discriminate between artifacts produced by natural processes, and artifacts produced by human intelligence. It seems an almost intuitive process.

What about biological artifacts? We have a theory of how they might be produced by natural means. But they seem incredibly complex, and evolutionary theory hasn’t yet produced a convincing step-by-step explanation of how many (even most) biological features could have arisen.

ID is an attempt to apply our design intuitions to (primarily) biological artifacts to decide if they were intelligently designed, or if they merely display the appearance of design but are produced naturally.

So far this sounds reasonable, if rather vague. We need some kind of rigorous, clearly defined evaluation method if the conclusions are to be convincing. Obviously confirmation bias means theists are going to be easier to persuade than atheists, so this process needs to be as objective as possible.

ID’s general approach is to try determine what properties of artifacts are associated only with design, and then identify these properties in biological features. The argument is as follows:

  1. A property is identified as a marker of ID based on our own experience of that property and our understanding of its nature. For example, the purposeful arrangement of parts and functional information.
  2. Biological organisms display many of these properties.
  3. There are no credible natural explanations for these properties of organisms, and so design is the most credible explanation.

This seems a reasonable approach – providing such properties can be found. ID proponents focus on two properties that are closely related.

Complexity

Complexity seems a good candidate as a design marker. Many of our human artifacts (e.g. a computer) are extremely complex.  But there are three major problems.

Firstly, we don’t have an objective, language independent way of assigning complexity to any artifact that captures our intuitions. Until we can do so, complexity isn’t a viable candidate.

Secondly, even if we decided upon a complexity measure (e.g. Kolmogorov complexity), we don’t have a technique allowing us to calculate this for a given biological feature. I’m not aware of any complexity calculations performed in ID so far.

Finally, it is unclear how we could demonstrate only intelligent agents can produce such complexity.

Irreducible complexity

Enter irreducible complexity (IC), which tries to address the third issue above. ID’s claim is that there are certain biological structures unable to be produced by natural processes, only by intelligence. A system is IC if it is “a single system composed of several well-matched, interacting parts that contribute to the basic function, wherein the removal of any one of the parts causes the system to effectively cease functioning” (Behe, 1996).

This sounds promising, but in practice only restates the issue. Demonstrating that a feature is IC seems very difficult (e.g. co-opting of systems for purposes other than their original function is claimed to generate apparently IC structures), and is subject to accusations of a lack of imagination. There’s no broad consensus on any IC features – examples such as the bacterial flagellum and the blood clotting cascade are disputed by most biologists. Recasting the argument as a probabilistic one runs into the thorny issue of how to calculate the probabilities involved when we don’t know what the process is.

The analogy problem

ID works by analogy with human intelligent design.  It assumes human intelligence is representative of all advanced intelligence.  But it seems optimistic to assert that the intelligence required to create life (and perhaps the universe) is analogous to human intelligence. Do we really have any idea what the characteristic design trail of a superhuman intelligence would be like? Would they create IC structures, for example?

We can reliably infer human design as we know a lot about humans. We know far less about super-intelligent designers – and ID proponents are reluctant to provide any details. Crucially, super-intelligent designers can’t be alive in the sense that we are familiar with, because they can’t have designed themselves. So why assume their minds have anything in common with our own? 

Inference to best explanation

It has become common for ID proponents (initiated by Stephen Meyers) to frame their method of enquiry as an Inference to Best Explanation (IBE), largely (I think) to build a case for ID being a scientific explanation. IBE involves taking competing explanations (e.g. evolution vs ID) and deciding which one best fits the empirical data. It is particularly suitable for comparing hypotheses for non-repeatable events from the past.

God of the gaps?

ID is often accused of being a “God-of-the-gaps” argument, which means invoking God for natural phenomena that science hasn’t explained.

Although this is a pejorative, it’s worth noting that there might well be genuine gaps in nature that could be empirically detectable. If so, these would suggest a supernatural designer.  The existence of such gaps is a legitimate question – the problem being demonstrating that such gaps exist as illustrated by the concept of irreducible complexity. Historically, many gaps have been closed, which has discredited this approach, and of course the supernatural element helps discredit gaps as science.

ID’s use of IBE neatly avoids the charge of being God-of-the-gaps. Rather, it is a comparison of competing hypotheses based on available empirical evidence.

Some issues with IBE

There are some issues with IBE both as a method, and more specifically as applied to ID.

Best of a poor lot?

The first is that the outcome of using IBE is only as good as the explanations being compared. What if we settle on the best of a poor lot? Typically, ID is compared against evolutionary theory.  Are there any other options? And what do we mean by ID? Are there different theories of how this designer operated?

What if God programmed the laws of evolution so they can produce IC structures? What if God has programmed the laws of the universe to make human life inevitable? In this case, evolution does create IC structures – but the process is intelligently designed.

What makes a good explanation?

Is an “intelligent designer” really an explanation of anything? The theory seems to lack any empirical content. When and how did this designer act? What are its characteristics?

Conspiracy theories

The frustrating thing about conspiracy theories are that they are infinitely malleable, and are always able to explain all the facts. They can be made to explain circumstances perfectly. But (much of the time!), they are wrong. If the intelligent designer is not specified to some degree, it too may become a malleable explanation that can account for any evidence.

With IBE we must cull the most unlikely explanations. One way of doing this is to examine if there are independent reasons to believe the hypothesis. Are there independent reasons to believe that there is an intelligent designer? It is difficult for IDers to answer yes when they trying to be “scientific” and are avoiding saying they think the designer is God.

Setting the bar higher

Finally, a super-intelligent designer is akin to a supernatural designer. We don’t want to rule out the supernatural by fiat, but it seems reasonable to set the bar higher for what is the “best explanation” if it involves something that might be supernatural.

Conclusion

There’s clearly some merit to the design inference, as shown by our intuitions about human designs. The problem is distilling these intuitions into a rigorous method that applies to biological features and only selects intelligent agency. 

A major issue is the reluctance of ID proponents to identify the designer. I suspect this is to partly to avoid being tarred with the YEC brush, as well as an attempt to appear more scientific. It’s a catch-22, as without more details on the designer, there is little content to an ID theory and it fails most tests for science.

It may be best for ID proponents to propose a rigorously defined supernatural designer, and argue that this does not preclude ID being science (if being science is that important to them).

The IBE approach is promising, but it needs these details for ID to have any chance of being the best explanation. We also need theories about what the designer has done and when. We might get more detailed predictions that are falsifiable.

Donald Prothero’s observation on the current state of ID seems rather apt for now:  “they don’t offer any new scientific ideas or a true alternative theory competing with evolution. All they argue is that some parts of nature seem too complex for them to imagine an evolutionary explanation” (Prothero, 2010, 418). 

References

Behe, Michael J. 1996: Darwin’s Black Box: The Biochemical Challenge to Evolution. New York: Free.

Discovery Institute, 2016: Frequently Asked Questions. [online] Available at:
http://www.discovery.org/id/faqs/ [Accessed 19 Jan. 2016].

Prothero, Donald, 2010: Science and Creationism. In Rosenberg, A. and Arp, R. (eds). Philosophy of biology. Chichester, U.K.: Wiley-Blackwell.

Evaluating Intelligent Design 1

Evaluating Intelligent Design 1

Having allowed that ID is not necessarily religious in nature and is capable of transcending its creationist origins, the task is to evaluate ID according to the proposed strategy outlined earlier. The three necessary criteria will be examined in detail, followed by the four indicators.

Testability

ID’s historical claims will need to be evaluated for testability by examining their causal explanations. Its hypotheses about design detection must also be examined for testability.

Historical testability

The basic hypothesis of ID states that ‘certain features of the universe and of living things are best explained by an intelligent cause’. However this is a very general hypothesis too vague to generate meaningful causal explanations. According to Elliott Sober, ‘testing the design hypothesis requires that we have information about the goals and abilities the designer would have’ (Sober, 1999).

Yet ID proponents explicitly deny any knowledge of the designer. Michael Behe states that ‘the reasons that a designer would or would not do anything are virtually impossible to know unless the designer tells you specifically what those reasons are’ (Behe, 1996). Similarly, Stephen Meyer says ID ‘does not claim to be able to determine the identity or any other attributes of that intelligence’ (Meyer, 2009, 428). A corollary is that the designer’s mechanism for creation is also unknown.

This lack of knowledge is used as a defence by ID proponents against the existence of apparently sub-optimal designs in nature – how can we know that a designer would not create them when we know nothing about their intentions? This, however, highlights the difficulty of testing such a minimalist version of intelligent design – what empirical observations would we expect from a designer whose attributes are unknown?

According to Behe and Meyer, ‘the conclusion that something was designed can be made quite independently of knowledge of the designer’ (Behe, 1996, 197), as ‘we know from experience only conscious, intelligent agents produce large amounts of specified information’ (Meyer, 2009, 429).

In terms of historical testing, their basic claim is that intelligent design can be reliably detected, and this is empirical evidence of a design event some time in the past. No auxiliary assumptions about the designer are necessary, and this evidence of design is Cleland’s ‘smoking gun’ that selects the design hypothesis over evolution.

This is a dubious claim, as it is dependent on an unstated assumption that human intelligence acts in a way analogous to designer intelligence, despite the very significant differences in capabilities and the denial of any knowledge about the designer. It is also vague enough to deny that sub-optimal designs could not be the product of this intelligence, as well accommodating any other empirical evidence that might be available. There is no indication as to how or why the intelligent designer does what it is claimed to do, and so very little is explained – in fact it provides no more explanation than the claim ‘God did it’ does. It has to be concluded that without a more developed theory, ID is untestable as a historical science.

Testability of design detection using CSI

William Dembski’s complex specified information (CSI) is at the heart of ID’s design detection claims. In terms of testability, the claim is that a biological feature that exhibits CSI has been designed by an intelligent agent.

This use of CSI is undermined by a theoretical flaw that makes this claim difficult to substantiate. Both the explanatory filter and Dembski’s more recent simplified approach of ruling out chance hypotheses (which now include regularities) require that currently unknown natural laws are rejected to conclude design. According to Fitelson, rejecting all possible chance and regularity hypotheses requires ‘ kind of omniscience’ (Fitelson et al, 1999) – both in the estimation of probabilities that are poorly defined, and because of these unknown laws. Given PMN’s presumption of naturalistic explanations, it is hard to see how a conclusion of design could be arrived at unless extraordinary empirical evidence confirms it. ID is not necessarily supernatural as has been discussed, but it is potentially supernatural and so still requires such evidence.

Is the design claim testable, despite this flaw? In principle, a series of tests could be devised that calculate CSI for human-designed and natural artefacts to validate the Dembski formula for human intelligence. However the auxiliary assumption that a designer would act in a way analogous to human intelligence would be required, as with historical testing. It also does not solve Fitelson’s issue, and so it is difficult to see how any tests can validate CSI’s applicability to design in biological organisms.

Testability of irreducible complexity

The first issue regarding testability is that IC is primarily a test of evolution rather than ID. If a truly IC feature was found, evolution would be falsified – but this would not necessarily rule out an alternative natural mechanism.

A further problem is that without any details of the designer and the mechanism, we can’t know an intelligent designer would build IC structures. Similarly, IC cannot falsify ID. If an evolutionary explanation is later found for a feature claimed to be IC, ID proponents can claim that they were mistaken about this feature being IC.

In terms of historical testing, an IC feature would certainly favour the ID hypothesis instead of evolution, but again the PMN requirement means that an extraordinary level of evidence is required for a feature to be considered IC.

How is IC determined? Luskin states that IC ‘can be tested for by reverse-engineering biological structures through genetic knockout experiments to determine if they require all of their parts to function’ (Luskin, 2011). But does this mean a structure is IC? To be definitive would surely involve examining all conceivable evolutionary scenarios for constructing the feature rather than just determining if all parts are required. This would need to include the possibility of exaptation, where the function of a biological trait can change during its evolutionary history. H. Allen Orr complicates things further by suggesting ‘an irreducibly complex system can be built gradually by adding parts that, while initially just advantageous, become— because of later changes—essential’ (Orr, 1996).

Behe himself notes ‘an exhaustive consideration of all possible roles for a particular component can’t be done’ (Behe, 1996, 111). Behe rather is claiming that it is extremely unlikely that certain features he has identified as irreducibly complex could have evolved gradually, i.e. his is a probabilistic argument: ‘one can not definitively rule out the possibility of an indirect, circuitous route. As the complexity of an interacting system increases, though, the likelihood of such an indirect route drops precipitously’ (Behe, 1996, 40).

However Behe’s discussions of supposedly IC features such as blood clotting, the bacterial flagellum and the immune system do not perform any probabilistic calculations (Behe, 1996). They describe the systems in some detail and highlight their complexity, as well as the current lack of knowledge concerning their evolution. How the relevant probabilities could be calculated for an unknown process is not discussed, and it is difficult to see how a conclusive decision could be made. They primarily seem to be arguments from incredulity, and so IC’s testability is problematic.

Testability of evolution criticisms

ID’s various criticisms of evolutionary theory listed above also offer no direct positive evidence for ID, and are again tests of evolution rather than ID.

But when comparing historical hypotheses, criticisms of one hypothesis are indirect support for the other. According to Stephen Meyer, ‘since design hypotheses are often formulated as strong claims about intelligence as the best causal explanation of some particular phenomenon, these hypotheses entail counter-claims about the insufficiency of competing materialistic mechanisms’ (Meyer, 2009, 482). This seems reasonable – provided that ID is providing a causal explanation. But we have already concluded that ID provides little explanation at all, and so this claim seems very weak.

ID’s predictions

Stephen Meyer lists ‘a dozen ID-inspired predictions’ (Meyer, 2009, 482-483). Representative examples are briefly examined below.

One prediction is that instances of ‘bad design’ will eventually show evidence of degenerated original designs, or functional reasons for the supposedly bad design – for example, the backward wiring of the retina.

Another is that the fossil record should show evidence of sudden appearances of major forms of life, and limits to the amount of change organisms undergo discovered.

A final example is that ‘no undirected process will demonstrate the capacity to generate 500 bits of new information starting from a non-biological source’, which draws on Dembski’s CSI.

These predictions are not testable. Some predictions (e.g. ‘bad design’) are not falsifiable in that they predict an event vindicating ID at an unspecified future point. Other predictions of future events that would falsify ID (e.g. generating bits of information) cannot easily be tested. And finally some predictions (e.g. concerning the fossil record) are vague enough that in practice they are not falsifiable.

Conclusion

So is ID testable? Without any knowledge of the designer, there are no specific empirical predictions that can genuinely test the theory. Both CSI and IC have significant theoretical flaws that cast doubt on the veracity of design detection, which is the key claim in ID’s quiver. There is also little evidence to suggest that these concepts can be used in practice. Criticisms of evolution are tests of evolutionary theory, not ID. So ID’s key tenets are currently not testable, and we have also concluded that ID is not testable as a historical science. ID fails the stated criteria for science at the first hurdle.

 Part 5: Evaluating Intelligent Design – part 2