Explanation and Realism

1. Explanation and Realism “If two theories can differ with respect to their explanatory power, even though they both predict all the same phenomena, and if explanatory power is evidence for the truth of a theory, then the empirical equivalence of theories will not imply evidential equivalence” (Ladyman, 196)

2. The battleground: realists and constructive empiricists on explanation • What is a scientific explanation? • Realists hold that the truth (or approximate truth) of a hypothesis is a necessary condition for it to be part of a genuine scientific explanation. That is why, for realists, explanations require more than empirical adequacy (what is ‘empirical adequacy?) • For the realists, inference to the best explanation (IBE) along with other super-empirical virtues show that empirical equivalence of theories does not imply evidential equivalence.

3. Scientific explanations • The constructive empiricist must offer a critique of IBE if they are to succeed in arguing their case. Before we get to van Fraassen’s criticisms of IBE, let us consider what we look for in an explanation. Consider the following: • 1) The cup broke because it fell on the ceramic tiles. • 2) The pressure of the gas rose because the volume was fixed and the temperature was increased. • 3) The stone fell to Earth because its natural place is at the centre of the universe. • 4) They are not answering the phone because they want to be left alone • What are the common features in these examples?

4. Explanations • 4) is an explanation by appealing to our background knowledge of how people’s behavior is related to their beliefs and desires. • 3) is an example of an Aristotelian (teleological) explanation. That kind of explanation was rejected during the scientific revolution because it did not appeal to causation • 1) & 2) are both examples of causal explanations; they attribute the structure of cause and effect to explain an event; 2) is an example of a nomic explanation (appeal to laws of nature)

5. Scientific explanations—the cover law model • Indeed 2) would appear to be the kind of explanation of choice in the sciences: explain the phenomena by discovering the laws of nature governing the phenomena. • Explain the tides by appeal to law of gravitation, the mass of water in the oceans, and the mass and position of the moon. • This is the cover law model. It was introduced by Carl Hempel. For him, to give a scientific explanation of an event is to show how it can be seen to follow from a law (or a set of laws) together with specific initial conditions.

6. Hempel’s deductive-nomological model • Schematic for the deductive-nomological model (DN model): • Laws: l1, l2, … • Conditions: C1, C2, … • _______________________________________ • Deductively entail explanadum O1, O2, … • Some terminology: • Explanadum: what is being explained • Explanans: what does the explaining

7. Deductive-nomological model • The DN model is also called the cover law model because the event explained is covered by some general law of nature. • Logical conditions of the DN model (202) • a) the explanans must deductively entail the explanadum • b) the deduction must make essential use of general laws • c) the explanans must have empirical content • d) the sentences in the explananas are true

8. Deductive-nomological model • a) requires that the explanadum must follow deductively from the explanans; i.e. it is a deductively valid argument • b) rules out the possibility of pseudo-scientific explanations, those that include laws to give the appearance of a scientific explanation • c) requires that the laws and the initial conditions must be empirically testable • d) ensures that the argument is deductively sound for a genuine scientific explanations appeal to true propositions

9. Deductive-nomological model • Example: why did the plant in the corner die? • 1) no sunlight was reaching the plant in the dingy corner • 2) sunlight is needed for photosynthesis • 3) photosynthesis produces the carbohydrates necessary for survival • Why is the above explanation a D-N explanation?

10. Problems with the DN model • Two general kinds of problems: • 1) criticize the DN model as not necessary • many explanations are perfectly acceptable without appeal to the DN model. So the DN model is too strict. • E.g. Why did the Titanic sink? • How would you defend the D-N model? • The explanation in the sinking of the Titanic is only a sketch of an explanation. The relevant laws can in principle be discovered.

11. Difficulties with Hempel’s model • 2) The D-N model is too liberal: it includes ‘explanations’ which are inadmissible. So the model isn’t sufficient • Defenders of the D-N model can account for objections under 1), but objections under 2) are much more difficult to answer • Consider the following objections:

12. Symmetry: the “thesis of structural identity” • Hempel: the D-N model shows that explanation and predictions are flip sides of the same coin; they have the same structure. The difference is that in explanation we already know that the conclusion of the argument (the explanadum) is true. • Symmetry: Information that allow us to predict a fact before we know it serves to explain it after its occurence. • E.g. Newtonian mechanics predicted the return of Haley’s comet in 1758, and once the comet has been observed, the same schema also explains why the comet returned when it did.

13. The problem with symmetry • Suppose the goal post at University stadium cast a shadow 20 m long yesterday. • How would you answer the question, why is the shadow 20 m long? • The goal post is 15 m high, and suppose the angle of elevation is 37 degrees. • The length is deduced from the height of the post, the angle of elevation, and laws of optics. • This explanation fits the D-N model from laws, initial conditions, you explain the length of the shadow.

14. The problem with symmetry • The problem with Hempel’s claim of symmetry is according to his account, the information about the length of the shadow and the angle of elevation is 37 degrees also explains the height of the post. • Schematic of argument: • Laws: optics—light travels in straight lines • Conditions: the shadow is 20 m, the angle of elevation is 37 degrees • Conclusion: the goal post is 15 m high • If Hempel is right about symmetry, then it is the length of the shadow that caused the goal post to be 15 m high. But that explanation is counter-intuitive.

15. The problem with symmetry • Hempel’s idea of structural identity in explanation and prediction is flawed because information that allow us to predict a fact before we know it may not serve to explain it after its occurrence. • Hempel’s DN model does not respect the fact that explanations are asymmetrical: If X explains Y, given the relevant laws and initial conditions, it is not true that Y will also explain X.

16. Other problems with the D-N model • Irrelevance: • In the example below, the explanans fit the model, but one (or more) of the explanans is not a relevant explanatory factor. • All salts dissolve in water • Father O’Brien put a sample of salt in holy water • Therefore, the sample of salt dissolved • What is wrong with this ‘explanation’?

17. Other problems • Pre-emption: an event that was going to happen for some reason happens earlier for another reason. For example: • Everyone who drinks a vat of Screetch gets really sick shortly thereafter • Wong drank a vat of Screetch • Therefore, Wong is really sick after the event. • What’s wrong with this ‘explanation’?

18. Other problems • Over-determination: when there are more than one set of causal conditions at play, but each one is sufficient to bring it about. For example: • All people who do not have sex do not get pregnant • John (who is a man) does not have sex • Therefore, John did not get pregnant • Note: the objections of over-determination and pre-emption overlap one another

19. Summary of problems with D-N model • The above counter-examples with regards to symmetry, irrelevance, pre-emption and over-determination show that the D-N model is not sufficient: The logical conditions set out do not exclude ‘explanations’ that are clearly not admissible. • The moral behind these counter-examples is that the D-N model fails to respect features that are essential to a ‘good’ explanation.

20. What about a causal account of explanation in science? • Given the problems with the D-N model, other accounts have been explored. • Some put more emphases in the notion of causality. They argue that a causal account is not identical with the D-N account. • The causal account would seem to respect, for instance, asymmetry of causes, and relevance of causal factors. • The trouble with the causal account is that it raises the thorny issue of the metaphysics of causality.

21. Constructive Empiricist alternative • How might a constructive empiricist respond to the question of exoplanation? • van Fraassen seeks to avoid unnecessary metaphysical commitments. His approach is to argue that the explanatory power of a theory must take into consideration a pragmatic account. • Consider the question, why did Bill buy that laptop? • The answer would depend on the context and interest of the person asking the question.

22. Constructive Empiricism vs. Realism • van Fraassen’s point: the explanatory power of a theory is pragmatic and not because the answer is compelled by the way the world is. • The realist, of course, claims the opposite, and the argument against anti-realism rests on the claim that the problem of under-determination (empirical equivalence is not identical with evidential equivalence) is solved by Inference to the best explanation type reasoning.

23. Problems with the DN model • The second kind of problem uses counterexamples to show that the DN model is too liberal; it includes explanations that are not genuine scientific explanations: • The problems can be grouped under irrelevance, preemption, overdetermination, and symmetry (203-204). • Let’s consider the problem of symmetry: Hempel observed that according to his model that explanation and predictions are flip sides of the same coin; they have the same structure • The difference is that in explanation we already know that the conclusion of the argument (the explanadum) is true.

24. The problem with symmetry • Suppose you notice that the goal post on Williston field today is casting a shadow 20 m long. • How would you answer the question, why is the shadow 20 m long? • The goal post is 15 m high, and the angle of elevation is 37 degrees. • The length is deduced from the height of the post, the angle of elevation, and laws of optics

25. The problem with symmetry • Suppose now we say that the goal post is 15 m high because the shadow is 20 m long and that the angle of elevation is 37 degrees. • Schematic of argument: • Laws: optics—light travels in straight lines • Conditions: the shadow is 20 m, the angle of elevation is 37 degrees • Conclusion: the goal post is 15 m high • This ‘explanation’ would also fit Hempel’s model but it is counter-intuitive. Why? • The shadow does not cause the goal post to be 15 m high.

26. The problem with symmetry • Hempel’s DN model does not respect the fact that explanations are asymmetrical • If X explains Y, given the relevant laws and initial conditions, it is not true that Y explains X. • Hempel’s idea of structural identity in explanation and prediction is also flawed; predictions and explanations part company as the flag pole example illustrates: Information that allow us to predict a fact before we know it does not serve to explain it after we know it.