Although there have been signs of improvement since the late twentieth century, one legacy of empiricism that continues to cause confusion, and has opened the door to a great deal of bad philosophy, is the idea that it is possible to split a scientific theory into its predictive rules of thumb on the one hand and its assertions about reality (sometimes known as its ‘interpretation’) on the other. This does not make sense, because – as with conjuring tricks – without an explanation it is impossible to recognize the circumstances under which a rule of thumb is supposed to apply. And it especially does not make sense in fundamental physics, because the predicted outcome of an observation is itself an unobserved physical process.
Many sciences have so far avoided this split, including most branches of physics – though relativity may have had a narrow escape, as I mentioned. Hence in, say, palaeontology, we do not speak of the existence of dinosaurs millions of years ago as being ‘an interpretation of our best theory of fossils’: we claim that it is the explanation of fossils. And, in any case, the theory of evolution is not primarily about fossils or even dinosaurs, but about their genes, of which not even fossils exist. We claim that there really were dinosaurs, and that they had genes whose chemistry we know, even though there is an infinity of possible rival ‘interpretations’ of the same data which make all the same predictions and yet say that neither the dinosaurs nor their genes were ever there.
One of them is the ‘interpretation’ that dinosaurs are only a manner of speaking about certain sensations that palaeontologists have when they gaze at fossils. The sensations are real, but the dinosaurs were not. Or, if they were, we can never know of them. The latter is one of many tangles that one gets into via the justified-true-belief theory of knowledge – for in reality here we are, knowing of them. Then there is the ‘interpretation’ that the fossils themselves come into existence only when they are extracted from the rock in a manner chosen by the palaeontologist and experienced in a way that can be communicated to other palaeontologists. In that case, fossils are certainly no older than the human species. And they are evidence not of dinosaurs, but only of those acts of observation. Or one can say that dinosaurs are real, but not as animals, only as a set of relationships between different people’s experiences of fossils. One can then infer that there is no sharp distinction between dinosaurs and palaeontologists, and that ‘classical language’, though unavoidable, cannot express the ineffable relationship between them. None of those ‘interpretations’ is empirically distinguishable from the rational explanation of fossils. But they are ruled out for being bad explanations: all of them are general-purpose means of denying anything. One can even use them to deny that Schrödinger’s equation is true.
Since explanationless prediction is actually impossible, the methodology of excluding explanation from a science is just a way of holding one’s explanations immune from criticism. Let me give an example from a distant field: psychology.
I have mentioned behaviourism, which is instrumentalism applied to psychology. It became the prevailing interpretation in that field for several decades, and, although it is now largely repudiated, research in psychology continues to downplay explanation in favour of stimulus-response rules of thumb. Thus, for instance, it is considered good science to conduct behaviouristic experiments to measure the extent to which a human psychological state such as, say, loneliness or happiness is genetically coded (like eye colour) or not (such as date of birth). Now, there are some fundamental problems with such a study from an explanatory point of view. First, how can we measure whether different people’s ratings of their own psychological state are commensurable? That is to say, some proportion of the people claiming to have happiness level 8 might be quite unhappy but also so pessimistic that they cannot imagine anything much better. And some of the people who claim only level 3 might in fact be happier than most, but have succumbed to a craze that promises extreme future happiness to those who can learn to chant in a certain way. And, second, if we were to find that people with a particular gene tend to rate themselves happier than people without it, how can we tell whether the gene is coding for happiness? Perhaps it is coding for less reluctance to quantify one’s happiness. Perhaps the gene in question does not affect the brain at all, but only how a person looks, and perhaps better-looking people are happier on average because they are treated better by others. There is an infinity of possible explanations. But the study is not seeking explanations.
It would make no difference if the experimenters tried to eliminate the subjective self-assessment and instead observed happy and unhappy behaviour (such as facial expressions, or how often a person whistles a happy tune). The connection with happiness would still involve comparing subjective interpretations which there is no way of calibrating to a common standard; but in addition there would be an extra level of interpretation: some people believe that behaving in ‘happy’ ways is a remedy for unhappiness, so, for those people, such behaviours might be a proxy for unhappiness.
For these reasons, no behavioural study can detect whether happiness is inborn or not. Science simply cannot resolve that issue until we have explanatory theories about what objective attributes people are referring to when they speak of their happiness, and also about what physical chain of events connects genes to those attributes.
So how does explanation-free science address the issue? First, one explains that one is not measuring happiness directly, but only a proxy such as the behaviour of marking checkboxes on a scale called ‘happiness’. All scientific measurements use chains of proxies. But, as I explained in Chapters 2 and 3, each link in the chain is an additional source of error, and we can avoid fooling ourselves only by criticizing the theory of each link – which is impossible unless an explanatory theory links the proxies to the quantities of interest. That is why, in genuine science, one can claim to have measured a quantity only when one has an explanatory theory of how and why the measurement procedure should reveal its value, and with what accuracy.
There are circumstances under which there is a good explanation linking the measurable proxy such as marking checkboxes with a quantity of interest, and in such cases there need be nothing unscientific about the study. For example, political opinion surveys may ask whether respondents are ‘happy’ with a given politician facing re-election, under the theory that this gives information about which checkbox the respondents will choose in the election itself. That theory is then tested at the election. There is no analogue of such a test in the case of happiness: there is no independent way of measuring it. Another example of bona-fide science would be a clinical trial to test a drug purported to alleviate (particular identifiable types of) unhappiness. In that case, the objective of the study is, again, to determine whether the drug causes behaviour such as saying that one is happier (without also experiencing adverse side effects). If a drug passes that test, the issue of whether it really makes the patients happier, or merely alters their personality to have lower standards or something of that sort, is inaccessible to science until such time as there is a testable explanatory theory of what happiness is
In explanationless science, one may acknowledge that actual happiness and the proxy one is measuring are not necessarily equal. But one nevertheless calls the proxy ‘happiness’ and moves on. One chooses a large number of people, ostensibly at random (though in real life one is restricted to small minorities such as university students, in a particular country, seeking additional income), and one excludes those who have detectable extrinsic reasons for happiness or unhappiness (such as recent lottery wins or bereavement). So one’s subjects are just ‘typical people’ – though in fact one cannot tell whether they are statistically representative without an explanatory theory. Next, one defines the ‘heritability’ of a trait as its degree of statistical correlation with how genetically related the people are. Again, that is a non-explanatory definition: according to it, whether one was a slave or not was once a highly ‘heritable’ trait in America: it ran in families. M
ore generally, one acknowledges that statistical correlations do not imply anything about what causes what. But one adds the inductivist equivocation that ‘they can be suggestive, though.’
Then one does the study and finds that ‘happiness’ is, say, 50 per cent ‘heritable’. This asserts nothing about happiness itself, until the relevant explanatory theories are discovered (at some time in the future – perhaps after consciousness is understood and AIs are commonplace technology). Yet people find the result interesting, because they interpret it via everyday meanings of the words ‘happiness’ and ‘heritable’. Under that interpretation – which the authors of the study, if they are scrupulous, will nowhere have endorsed – the result is a profound contribution to a wide class of philosophical and scientific debates about the nature of the human mind. Press reports of the discovery will reflect this. The headline will say, ‘New Study Shows Happiness 50% Genetically Determined’ – without quotation marks around the technical terms.
So will subsequent bad philosophy. For, suppose that someone now does dare to seek explanatory theories about the cause of human happiness. Happiness is a state of continually solving one’s problems, they conjecture. Unhappiness is caused by being chronically baulked in one’s attempts to do that. And solving problems itself depends on knowing how; so, external factors aside, unhappiness is caused by not knowing how. (Readers may recognize this as a special case of the principle of optimism.)
Interpreters of the study say that it has refuted that theory of happiness. At most 50 per cent of unhappiness can be caused by not knowing how, they say. The other 50 per cent is beyond our control: genetically determined, and hence independent of what we know or believe, pending the relevant genetic engineering. (Using the same logic on the slavery example, one could have concluded in 1860 that, say, 95 per cent of slavery is genetically determined and therefore beyond the power of political action to remedy.)
At this point – taking the step from ‘heritable’ to ‘genetically determined’ – the explanationless psychological study has transformed its correct but uninteresting result into something very exciting. For it has weighed in on a substantive philosophical issue (optimism) and a scientific issue about how the brain gives rise to mental states such as qualia. But it has done so without knowing anything about them.
But wait, say the interpreters. Admittedly we can’t tell whether any genes code for happiness (or part of it). But who cares how the genes cause the effect – whether by conferring good looks or otherwise? The effect itself is real.
The effect is real, but the experiment cannot detect how much of it one can alter without genetic engineering, just by knowing how. That is because the way in which those genes affect happiness may itself depend on knowledge. For instance, a cultural change may affect what people deem to be ‘good looks’, and that would then change whether people tend to be made happier by virtue of having particular genes. Nothing in the study can detect whether such a change is about to happen. Similarly, it cannot detect whether a book will be written one day which will persuade some proportion of the population that all evils are due to lack of knowledge, and that knowledge is created by seeking good explanations. If some of those people consequently create more knowledge than they otherwise would have, and become happier than they otherwise would have been, then part of the 50 per cent of happiness that was ‘genetically determined’ in all previous studies will no longer be so.
The interpreters of the study may respond that it has proved that there can be no such book! Certainly none of them will write such a book, or arrive at such a thesis. And so the bad philosophy will have caused bad science, which will have stifled the growth of knowledge. Notice that this is a form of bad science that may well have conformed to all the best practices of scientific method – proper randomizing, proper controls, proper statistical analysis. All the formal rules of ‘how to keep from fooling ourselves’ may have been followed. And yet no progress could possibly be made, because it was not being sought: explanationless theories can do no more than entrench existing, bad explanations.
It is no accident that, in the imaginary study I have described, the outcome appeared to support a pessimistic theory. A theory that predicts how happy people will (probably) be cannot possibly take account of the effects of knowledge-creation. So, to whatever extent knowledge-creation is involved, the theory is prophecy, and will therefore be biased towards pessimism.
Behaviouristic studies of human psychology must, by their nature, lead to dehumanizing theories of the human condition. For refusing to theorize about the mind as a causative agent is the equivalent of regarding it as a non-creative automaton.
The behaviourist approach is equally futile when applied to the issue of whether an entity has a mind. I have already criticized it in Chapter 7, in regard to the Turing test. The same holds in regard to the controversy about animal minds – such as whether the hunting or farming of animals should be legal – which stems from philosophical disputes about whether animals experience qualia analogous to those of humans when in fear and pain, and, if so, which animals do. Now, science has little to say on this matter at present, because there is as yet no explanatory theory of qualia, and hence no way of detecting them experimentally. But this does not stop governments from trying to pass the political hot potato to the supposedly objective jurisdiction of experimental science. So, for instance, in 1997 the zoologists Patrick Bateson and Elizabeth Bradshaw were commissioned by the National Trust to determine whether stags suffer when hunted. They reported that they do, because the hunt is ‘grossly stressful . . . exhausting and agonizing’. However, that assumes that the measurable quantities denoted there by the words ‘stress’ and ‘agony’ (such as enzyme levels in the bloodstream) signify the presence of qualia of the same names – which is precisely what the press and public assumed that the study was supposed to discover. The following year, the Countryside Alliance commissioned a study of the same issue, led by the veterinary physiologist Roger Harris, who concluded that the levels of those quantities are similar to those of a human who is not suffering but enjoying a sport such as football. Bateson responded – accurately – that nothing in Harris’s report contradicted his own. But that is because neither study had any bearing on the issue in question.
This form of explanationless science is just bad philosophy disguised as science. Its effect is to suppress the philosophical debate about how animals should be treated, by pretending that the issue has been settled scientifically. In reality, science has, and will have, no access to this issue until explanatory knowledge about qualia has been discovered.
Another way in which explanationless science inhibits progress is that it amplifies errors. Let me give a rather whimsical example. Suppose you have been commissioned to measure the average number of people who visit the City Museum each day. It is a large building with many entrances. Admission is free, so visitors are not normally counted. You engage some assistants. They will not need any special knowledge or competence; in fact, as will become clear, the less competent they are, the better your results are going to be.
Each morning your assistants take up their stations at the doors. They mark a sheet of paper whenever someone enters through their door. After the museum closes, they count all their marks, and you add together all their counts. You do this every day for a specified period, take the average, and that is the number that you report to your client.
However, in order to claim that your count equals the number of visitors to the museum, you need some explanatory theories. For instance, you are assuming that the doors you are observing are precisely the entrances to the museum, and that they lead only to the museum. If one of them leads to the cafeteria or the museum shop as well, you might be making a large error if your client does not consider people who go only there to be ‘visitors to the museum’. There is also the issue of museum staff – do they count as visitors? And there are visitors who leave and come back on the same day, and so on. So you need quite a sop
histicated explanatory theory of what the client means by ‘visitors to the museum’ before you can devise a strategy for counting them.
Suppose you count the number of people coming out as well. If you have an explanatory theory saying that the museum is always empty at night, and that no one enters or leaves other than through the doors, and that visitors are never created, destroyed, split or merge, and so on, then one possible use for the outgoing count is to check the ingoing one: you would predict that they should be the same. Then, if they are not the same, you will have an estimate of the accuracy of your count. That is good science. In fact reporting your result without also making an accuracy estimate makes your report strictly meaningless. But unless you have an explanatory theory of the interior of the museum – which you never see – you cannot use the outgoing count, or anything else, to estimate your error.
Now, suppose you are doing your study using explanationless science instead – which really means science with unstated, uncriticized explanations, just as the Copenhagen interpretation really assumed that there was only one unobserved history connecting successive observations. Then you might analyse the results as follows. For each day, subtract the count of people entering from the count of those leaving. If the difference is not zero, then – and this is the key step in the study – call that difference the ‘spontaneous-human-creation count’ if it is positive, or the ‘spontaneous-human-destruction count’ if it is negative. If it is exactly zero, call it ‘consistent with conventional physics’.
The less competent your counting and tabulating are, the more often you will find those ‘inconsistencies with conventional physics’. Next, prove that non-zero results (the spontaneous creation or destruction of human beings) are inconsistent with conventional physics. Include this proof in your report, but also include a concession that extraterrestrial visitors would probably be able to harness physical phenomena of which we are unaware. Also, that teleportation to or from another location would be mistaken for ‘destruction’ (without trace) and ‘creation’ (out of thin air) in your experiment and that therefore this cannot be ruled out as a possible cause of the anomalies.
The Beginning of Infinity Page 38