To make an impartial judgement on the status of ID as a science or pseudoscience, a demarcation strategy must be adopted. It must allow fields widely accepted as science to be classified as such, and must exclude the most obvious cases of pseudoscience.
The proposal presented here acknowledges the difficulties Laudan pointed out with a necessary and sufficient set of demarcation criteria. However it rejects his conclusion that demarcation is a pseudo-problem, and instead takes the view that while there are no clear-cut demarcation boundaries, demarcation is still possible in many instances.
The proposal suggests three minimal necessary criteria for demarcation, coupled with a variable group of indicators. The necessary criteria are not regarded as sufficient, but serve to draw a relatively uncontroversial boundary around science. The more additional indicators that are satisfied by a field, the more firmly it is placed within the realm of science. It may be that some fields satisfy the necessary criteria, but few if any of the indicators, and it may not be possible to come to a firm conclusion in these cases. But this is not a fatal objection: rather it might be that a field is in the early stages of development as a science and its status may become clearer in the future.
The first and most important proposed criterion is testability. Science is an empirical investigation about the real world, and so scientific theories must be checked against the real world – they must be testable. Testability is closely related to falsifiability – a field can only be falsified if it can be tested. Additionally, the Duhem–Quine thesis tells us that a theory cannot be tested in isolation – theories require a number of auxiliary assumptions, and these are also tested when a theory is tested. Crucially, there should be independent evidence for auxiliary assumptions, to prevent their ad hoc invention to ensure a theory produces the expected results.
Historical or operational science?
Many creationists and ID supporters have distinguished between ‘historical science’ and ‘operational science’. A detailed case was presented by Geisler and Anderson (Geisler et al, 1987). They claimed historical sciences investigate phenomena that have occurred in the past (such as a creation event) and are not subject to repeatable experiment, unlike operational (or experimental) science.
This distinction has been used to argue that creationism and ID should not be subject to a testability criterion, and that they have equivalent scientific status to evolutionary theory, another historical science.
There is a valid basis to claims of methodological differences between historical and operational sciences. Stephen Jay Gould made the distinction, as have philosophers of science such as Carol Cleland. Of course, it should be noted that evolutionary theory uses both historical and operational methodologies – experimental evolution is an increasingly important field.
Do these differences grant an exemption from testability? The answer must be no. A science’s theories must be confirmed against the real world, and that can only be performed via some form of testing. As Gould stated: ‘We cannot see a past event directly, but science is usually based on inference, not unvarnished observation. The firm requirement for all science – whether stereotypical or historical – lies in secure testability, not direct observation. History’s richness drives us to different methods of testing, but testability is our criterion as well’ (Gould, 2000, 282).
Similarly, Michael Shermer wrote that ‘we cannot rerun the past, alter a variable here or there, and then observe the effects. This does not mean, however, that we cannot make causal inferences from what has already transpired’ (Shermer, 2002, 317).
As Gould and Shermer note, testing in the historical sciences must be performed differently to testing in operational science. Both Forber et al and Cleland argue that multiple competing hypotheses must be considered, and the hypothesis providing the best causal explanation selected. Forber et al argue for multiple independent lines of evidence that converge on a hypothesis’ estimates (Forber et al, 2011).
According to Cleland, predictions made by historical hypotheses are usually too vague to fail, and function more as educated guesses. Instead, ‘hypotheses are accepted and rejected by virtue of their power to explain as opposed to predict the evidence that supports them’. The assumption is that ‘seemingly improbable associations among present-day traces of the past are best explained in terms of a common cause’ (Cleland, 2013). So the process is to compare between common cause hypotheses for their ability to explain all the available evidence, hoping for ‘smoking gun’ evidence that discriminates between them.
In the historical sciences, then, testing proceeds by comparing multiple hypotheses to see which provides the best common cause explanation for the available empirical evidence. Is ID a historical science? Certainly, many of ID’s claims concern historical events, particularly those pertaining to criticisms of evolutionary theory. Examples include claims about the lack of intermediate fossils, and Stephen Meyers’ argument about the Cambrian explosion (Meyer, 2013), which attempts to provide a competing explanation to evolutionary theory. However ID also makes claims about being able to detect design in living organisms. While the initial design event may be regarded as a historical claim based far in the past, it is very much a matter of operational science to demonstrate the veracity of design detection techniques.
This means investigating testability will need to consider testing from both a historical and operational point of view, depending on the claims being made.
The second necessary criterion is empirical adequacy. A theory is empirically adequate if its predictions approximate relevant observable aspects of the world. Empirical adequacy is strongly related to testability. If theories are not testable, they will automatically fail to be empirically adequate. Testability is not sufficient, though – theories must be confirmed by successful testing to be empirically adequate.
Of course, testing will not always be successful. Test failures cannot be ignored – they must be acknowledged and eventually resolved. By contrast, pseudosciences are ‘selective in considering confirmations and disconfirmations’ (Thagard, 1978).
The final necessary criterion is the presumption of methodological naturalism (MN) – the principle that ‘all hypotheses and events are to be explained and tested by reference to natural causes and events’ (Kurz, 1998). MN is widely considered to be a working assumption of science, and science’s successes based on MN are a strong inductive justification for requiring it as a necessary feature of doing science.
An important question is whether MN should permit the investigation of supernatural phenomena. In the context of creation science and ID, this is a pertinent question, as creation science explicitly appeals to the supernatural (Creation Ministries International, 2015). MN was a key demarcation criterion used in numerous court cases to reject creation science and subsequently ID because of their apparent reliance upon the supernatural.
Restricting science to natural phenomena is problematic in several ways. Firstly, the concept of the supernatural is ill-defined. Secondly, there seems to be no a priori reason why supernatural events could not exert a causal influence on the physical, and if we suspected that they did, we would surely want to investigate. Finally, there have been numerous attempts to investigate a variety of supernatural phenomena by scientific means, e.g. Benson et al’s study on the effects of intercessory prayer (Benson et al., 2006), and this implies that the supernatural is amenable to scientific investigation.
Boudry et al argue that rather than considering MN as an intrinsic limitation of science, it should be a ‘provisory and empirically grounded commitment to naturalistic causes and explanations, which in principle is revocable by extraordinary empirical evidence’ (Boudry et al, 2010). They justify this inductively, based on ‘the pattern of consistent success of naturalistic explanations’ (Boudry et al, 2010), and label this Provisory (or Pragmatic) Methodological Naturalism (PMN). PMN does not exclude the supernatural by definition, accepting that supernatural forces ‘would have empirically detectable consequences, and these are in principle open to scientific investigation’ (Boudry et al, 2012).
This proposal suggests adopting PMN as a necessary feature of science. It means that supernatural explanations cannot be excluded from science by philosophical decree, but naturalistic explanations are to be preferred unless ‘extraordinary empirical evidence’ is brought to light.
This has implications for testing in the historical sciences, implying that extraordinary ‘smoking gun’ evidence would be required to prefer a supernatural hypothesis over a naturalistic hypothesis.
Four optional indicators have been chosen that are characteristic of fields that have undisputed scientific status, such as physics, chemistry and biology. Note that some of the indicators are not uniquely associated with scientific fields, but are also characteristic of other academic fields.
The first indicator is progressiveness. Thagard states that for a theory or discipline to be pseudoscience it ‘has been less progressive than alternative theories over a long period of time, and faces many unsolved problems … the community of practitioners makes little attempt to develop the theory towards solutions of the problems’ (Thagard, 1978). He defines progressiveness as ‘a matter of the success of the theory in adding to its set of facts explained and problems solved’ (Thagard, 1978).
A progressive field continually identifies anomalies and shortcomings of its theories and actively works to resolve them. It develops, modifies and discards theories as necessary – this is sometimes known as revisability, and also implies that theories are tentative.
If an alternative theory exists, comparative progressiveness becomes important. A pseudoscience’s practitioners may be doggedly persisting over many years when an alternative theory is solving problems that they continue to struggle with.
Progressiveness is an indicator rather than a necessary criterion because many theories we regard as scientific have been superseded and are no longer progressive. For example, Newtonian gravity was superseded by general relativity, while steady-state cosmology was superseded by big bang cosmology.
Theories should explain why things are as they are. Typically, this should be by reference to natural laws, although adopting PMM means this is not absolute.
Scientists generally share the results of their research – informally, at conferences, and by publication in peer-reviewed journals. This allows their claims to be scrutinised by the broader scientific community.
Scientists borrow relevant knowledge from adjacent disciplines relevant to their research. They do not repudiate or ignore knowledge that tends to disconfirm their theories.
Summary of proposal
The proposed demarcation strategy thus consists of evaluating a field against the three necessary criteria of testability, empirical adequacy and PMN, and the optional indicators of progressiveness, explanatory power, peer review, and knowledge borrowing.
Finally, to be considered pseudoscience, a field must both fail the above demarcation test and meet Hansson’s criterion of deliberately masquerading as science.
Evaluating the proposal
The strategy will be briefly evaluated against a field widely thought to be a pseudoscience – astrology. As a comparison, quantum field theory, an undisputed scientific field, is also evaluated.
Astrology is somewhat testable, as evidenced by the work of statistician Michel Gauquelin, and more recently Carlson (Carlson, 1985). However it is not empirically adequate, having failed any rigorous tests it has been subjected to. Given that no mechanism is suggested, the PMN criterion is difficult to assess.
Astrology fails to meet most of the proposed indicators. It is not progressive or explanatory – according to Thagard astrology ‘has changed little and has added nothing to its explanatory power since the time of Ptolemy’ (Thagard, 1978). There is no active research community or peer-reviewed publication record. Astrology does borrow some knowledge from astronomy, and this criterion is satisfied in a weak sense.
Based on the necessary criteria and the indicators, astrology fails the demarcation test, which firmly places it outside the realm of science.
Quantum field theory
Quantum field theory (QFT) is a framework for elementary particle physics. The Standard Model is an extremely successful QFT which has been tested extensively and confirmed to an extraordinary degree of accuracy, thus satisfying the necessary criteria. The Standard Model has progressed significantly over the last 60 years or more, it has broad explanatory power, there is a vast research community generating thousands of peer-reviewed articles, and it draws on many different areas of physics. It is science.
In each of the examples, the proposed demarcation strategy performs as expected. Astrology is decisively classified as non-science, and QFT is clearly shown to be science.