In The Grand Design, physicist Stephen Hawking expressed his opinion that the word “real” simply means “accurately predicts the evidence” and that two theories which predict the evidence equally well cannot be said to be more real than the other. He even goes as far as to say that the creationists who believe God planted the fossils in the rocks and created the world with age have a theory which is no more or less real than the theory that the earth is old and that these fossils are the remaining impressions of long-dead plants and animals.
I say, “poppycock.” If Hawking were on a jury, and he had to make a decision on whether to convict a man accused of murder, would he think that the theory that the man was guilty was no less real than the theory that aliens killed the victim and planted all of the evidence to get this man convicted? Something is deeply wrong here.
While I agree with the general thesis that a statement only means what it predicts, I would add that there are two different types of predictions a sentence can make: primary and secondary. Here’s an example: the primary prediction that the statement “A killed B last night at ten thirty” makes is that if you were there last night at ten thirty watching A, you would have seen him kill B. All of this can translated into a set of visual and other sensory experiences, which constitute the primary predictions the statement makes. The statement “A killed B last night at ten thirty” also makes a huge number of what I call secondary predictions, which are basically all the little clues that detectives and policemen look for in order to find out the probable truth or falsity of the statement, since it cannot be directly observed after it has happened.
Hawking makes a huge mistake when he reduces the meaning of a statement down to only the secondary predictions it makes (which is typically what scientists look for, as it is impractical and impossible to directly observe all the things they would like to find out about). He misses out on the fact that all statements make a set of primary predictions, too, and therefore two contradictory theories predicting the same evidence we have are not both equally real. Either one or the other is, and though we may not always be able to tell the difference, I believe we ca,n in most cases, judge more theory more likely than the other(s).
Jason Rosenhouse, while acknowledging a difference between theories which have accurate predictive power and theories which are true, gives only a brief and very weak defense of his position:
As a practical matter, it is almost impossible to avoid assuming that our best scientific theories track the world as it actually is. What’s the alternative? What else could they be doing? Realism is axiomatic. It’s not something you try to defend on the basis of simpler principles. Is there some risk that actual harm will come from being wrong on this point? If not, then why should I go to so much intellectual effort to reject an assumption that seems so obvious and natural?
But how do we tell the difference between two theories which predict the same evidence equally? After all, for just about any theory we have that predicts the evidence we have, we can come up with at least one alternative, perhaps even an outlandish and unbelievable alternative (see the evidence-planting alien and deceptive-fossil-planter-God examples above). Without simply using our gut, how can we tell which theory is true and which theory is false?
First of all, crazy theories such as the ones I have mentioned make propositions which are highly improbable prior to considering their evidential value. Second, they make multiple assumptions (which lowers the theory’s prior probability of being true, since each new assumption we add to theory has a chance of being incorrect). Last but not least, a theory with few adjustable parameters has less wiggle room concerning what evidence it can predict (and how probable it can make the evidence) than theories with numerous adjustable parameters, which often means that the former (if it accomodates the evidence we have without any ad hoc assumptions) predicts the evidence we have with higher probability, which in turn makes it more probable than the complex theory (all else held equal). On these points, see this paper and/or Richard Carrier’s recent book Proving History: Bayes’s Theorem and the Quest for the Historical Jesus.
Conclusion: Determining which scientific theory is real is not merely a matter of looking at what it predicts, though that is a key part of the process. Nor is it something that must be done with “just common sense” or “intuition” – it can and should be done objectively, and this can be done with an analysis of Bayesian prior probabilities.
