The Unnatural Nature of Science Page 9
In stark contrast to the claim for the scientists’ imaginative-artist approach to creativity is that of the Nobel laureate in economics Herbert Simon and his colleagues. They believe that scientific creativity can be carried out by a computer program: that there is thus no real difference between the work of the ‘genius’ scientist and that of those of lesser ability, and so the idea of high creativity is a myth. For them, the process of discovery can be described and modelled.
Their central hypothesis is that the mechanisms of scientific inquiry are not peculiar to that activity but can be analysed as special cases of the general mechanisms of problem-solving. They do recognize science as a social process and also, since its goals when beginning to tackle a problem are usually not clearly defined, that it differs from ordinary problem-solving: finding problems and formulating them in a precise form is an integral part of science. In contrast, problem-solving, it is suggested, can be considered within the framework of cognitive psychology in terms of creating a symbolic representation of the problem and using operators on this. The search for a solution is not random trial and error but is guided by rule of thumb – by heuristics. For example, there are 50 billion billion (50 × 1018 ) possible settings of ten dials on a safe if each is numbered from 0 to 99, but a click at the correct setting for each reduces the number of trials to open the safe to about 500. Good scientists merely have better heuristics and do not require ‘intuition’.
Simon and his colleagues’ major claim is to have developed computer programs which, using their problem-solving approach, can make discoveries over a wide range of topics. ‘Discoveries’ is not really the right word, for they have not discovered anything new; rather they have shown how the computer could have discovered universal gravitation from the information available to Newton, or Planck’s constant – a fundamental quantity in quantum mechanics – given the information available to Planck, so all their demonstrations have involved the invaluable wisdom of hindsight.
In their programs, the criterion for the proposed solution is that the law found should fit the data ‘well enough’ – not worse than 3 per cent error. They claim that the generalization that the computer has found to fit the data will never be unique. Their approach is to ignore the small error and catch the ‘rabbit’ first. For them, the function of verification procedures is not to provide scientists with unattainable certainty or uniqueness for their discoveries but to inform them about the risks they are running in committing themselves to hypotheses that have been formulated and to provide guidance that may enhance their chances of making relatively durable discoveries.
While their computer programs may be successful, there have been criticisms, not the least being the amount built into the programs since the programmers do know the answer themselves. Their programs have made no new discoveries. But probably a more serious criticism is that scientific research involves more than just problem-solving: there is also data-gathering, description, explanation and theory-testing. The invention of new instruments, for example, does not fall within their computer programs: they are concerned only with ‘the induction of descriptive and explanatory theories from data’. While they do recognize the importance of correctly formulating a problem, they claim that this is nothing more than a variety of problem-solving, a claim which is strongly disputed. One only need recall that Einstein’s discovery of the theory of relativity was influenced by his posing the following problem: what would be the consequences of running alongside and then catching up with a point on a light wave? Computers couldn’t ‘think’ like this.
For the choice of problem is crucial. As the Nobel laureate Peter Medawar put it, science is the ‘art of the soluble’, and part of that art is choosing a problem which will turn out to be soluble. Francis Crick, for example, makes much of this point in relation to protein structure. The early heady days of molecular biology led to the sequence hypothesis, namely that the sequence of amino acids in proteins, which was specified by the DNA, completely determined the properties of the protein (Chapter 1). It was known that, although proteins were synthesized as a linear chain of amino acids, this chain then folded up spontaneously into complex shapes. The three-dimensional shape adopted by the chain was fundamental to the protein functioning properly, for the special properties of proteins are due to the shape of the folded chain which gives each protein a unique configuration and determines its function. Now, although the sequence of amino acids was assumed to be sufficient to determine the folded structure, this had not been formally established, nor was it possible to predict the structure from the sequence. Crick and his colleagues decided not to tackle the protein-folding problem, although it was in many ways the obvious next step in their research. How right they were, since thirty years later this still has not been fully solved – it is an exceptionally difficult problem.
Another aspect of problem-solving which is beyond current computer programs is knowing when to approximate, which comes only from experience. Approximation involves making simplifying assumptions which will make a problem tractable – at the risk, of course, of oversimplifying and thus making the solution of less value. And perhaps of no less importance is to know when to stop working on a problem or to abandon a particular line of investigation. It can be painful to give up much past investment of effort and take a new line.
Whether or not Simon’s problem-solving approach is correct – and I doubt that it is – it nevertheless contains an important idea, namely that at least part of scientific thinking is a kind of problem-solving of a very structured kind which can be simulated by computers. This emphasizes again the unnatural nature of scientific thinking, for computer programs of the type Simon and his colleagues use are quite unlike common-sense thinking. In general, computers are very bad at simulating common sense and such human activities as recognizing handwriting, and when they are successful in such activities they use a quite different mechanism to that used by our brains. But whatever talents computers may have, genius is not one of them.
Genius is always fascinating, raising some scientists to demigod status in the eyes of other scientists, but its nature remains an almost total mystery. Genius is usually judged with hindsight, but the scientific genius exerts a massive effect on both contemporaries and posterity. But no matter whether in any particular case the accolade of genius is applied or deserved, anyone who works in science quickly recognizes the leaders and truly creative workers in the field – they are often faster, more hardworking, more imaginative, cleverer, know more, understand more, speak better, calculate faster or possess at least some of these attributes.
Scientific genius can be recognized by the leadership the scientist gives and, more important, by the enormous influence on both contemporaries and posterity. Newton, Darwin and Einstein clearly qualify. Genius is ultimately ascribed for enduring eminence or reputation, which ought to reflect a contribution which has illuminated the nature of world. There may be some justification (Chapter 6) for regarding ‘genius’ as a social construct – social forces acting to establish who should be rewarded – but, while genius may include a social component, it is also about objective achievement. And while we may dispute whether someone is or is not a genius, we usually have little difficulty in identifying the outstanding scientists in any field.
Scientific genius also is quite different from that in the arts. One fundamental aspect that makes it different from genius in any other field is that, because science is a communal effort, in the long run the existence of scientific geniuses may be irrelevant: given time, resources and a sufficiently large trained community committed to science, all discoveries will probably be made. (In fact, the Ortega hypothesis, discussed later, claims that experimental science has progressed largely through the work of those of mediocre talent.) Discoveries and progress in science need not depend on any single person: if not Newton, then many others; if not Einstein, then X, Y and Z. The pathway of progress may be different, but in the end the result will most likely be the same. Not so
for Hamlet or Così fan tutte – there are no replacements for Shakespeare and Mozart. Again, the description by François Jacob of scientific genius emphasizes further differences:
In science the great man is, first of all, the one who knows how to spot the right problems at the right moment, while there is a chance of solving them. He is the one who knows how to surround himself with the right collaboration, to find among his pupils those capable of becoming his successors, and of developing the theories he has set forth …
Hardly an apt recipe for a Shakespeare, a Mozart or a Picasso.
The psychologist Howard Gruber has considered the problem of how early gifts and talent are transformed into exceptional creativity. It seemed to him that Thomas Henry Huxley was more brilliant and versatile than Darwin, his contemporary, and that any committee looking at their early research plans – Huxley’s for the voyage of the Rattlesnake and Darwin’s for the voyage of the Beagle – would have favoured Huxley. And when Huxley heard of Darwin’s theory he exclaimed, ‘Why didn’t I think of that?’ Why indeed? And the same is probably true for many others too. Gruber’s answer to Huxley’s question is: openness – young Darwin’s vague and open receptiveness was more successful than Huxley’s hard-edged analytic approach. But this is not necessarily a universal formula for success. How are we to understand the difference between the gifted and the extraordinary?
Studies on intellectually and academically gifted children show them to be highly efficient in the use of both short- and long-term memory processes. However, a more important feature is that they are capable of what are called metacognition and metamemory. Metacognition refers to a person’s awareness of his or her thought processes and is assumed to be crucial for the selection and implementation of complex problem-solving strategies. Metamemory is a subcomponent of metacognition and again refers to self-knowledge, in this case about the person’s memory system. Gifted children were found to be particularly aware of the strategies they used for remembering, such as interest in the topic and how they linked up thoughts. Successful scientists are similarly self-aware – it is one of their defining characteristics.
Studies on skill at chess may help to illuminate some aspects of skill at science. When grandmasters were compared with experts, there appeared to be no difference in the approach to chess problems and in their skill in solving them. The difference is due to motivation, character and knowledge. The grandmaster has a richer knowledge-base to draw on, due to thousands of hours of play and study. One needs passion and discipline to devote such time and energy, and that is a matter of character. There is a strong sense of truth in this when applied to gifted scientists: they have stamina, devotion, psychic courage and ‘character’, and they work very hard at problems.
The cry of ‘Eureka!’ may be rarer than popularly supposed, but, even so, the cry does ring out over the centuries. But the cry is often misleading, for it suggests that the solution to a scientific problem comes in a moment of divine, or ablutional, inspiration; it neglects the slow and often painful process from the formulation of the problem, through false turns, to that lovely moment of solution.
These points are encapsulated by Newton’s reply to the question of how he had discovered the law of gravity: ‘By thinking on it continually.’ Gruber makes the point that science, and creativity in general, has a long time-scale – the ideas of Newton, Darwin and Einstein took many years to develop. If there is a blinding flash, a ‘Eureka!’, we should not forget the ‘years’ that had previously been spent in thinking about the problem. Are they really less important? It is a characteristic feature – almost a defining feature – of science that it takes a long time to solve a problem. This is partly due to the difficulty inherent in the problem and partly because science is social and it is necessary to learn what other scientists have done in order to assimilate current knowledge.
Among the important characteristics of the great scientist is the ability, already referred to in relation to Crick, to recognize which problems to solve. It is also important to recognize which evidence and ideas to accept or discard. To challenge well-established beliefs can be remarkably difficult intellectually. One has also to make hard judgements about the available experimental data: as Francis Crick has pointed out, a theory that fits all the facts is bound to be wrong, as some of the facts are themselves bound to be in error (see Chapter 5).
The detailed study by Gruber of the origin of Darwin’s ideas on evolution provides a valuable case-study of scientific creativity. One of the first important ideas that Darwin developed was in relation not to evolution of animals but to a geological problem, the formation of coral reefs. While on the west coast of South America in 1835, Darwin put forward the idea that coral reefs were formed by the upward growth of the coral during the sinking of the land. This was, in a way, an evolutionary theory, in that it required the limitation of growth of the corals – corals do not grow beyond a limiting distance above the water – and the theory explains why the variation in forms of coral islands is continuous.
We share, however, Gruber’s surprise that Darwin’s first theory of animal evolution was along somewhat strange lines. In order to account for species changing and yet being adapted to their environment, and yet also for the number of species remaining approximately constant, Darwin invoked the emergence of simple life-forms, called monads, by spontaneous generation. The monads, he suggested, evolved as the result of direct environmental influence but had only a limited lifespan, so the species to which the monads gave rise eventually died and became extinct.
It is only possible to understand this seemingly ludicrous idea in terms of the concept of species present at that time. Each species was thought to contain its own specific essence, and thus it was impossible to imagine that it could either change or evolve. The ideas of the geologist Charles Lyell, who greatly influenced Darwin, illustrate this very clearly, since Lyell could not conceive that one species could be converted into another. And, if Lyell, who was so close to evolutionary thinking, could not conceive of this, then it was even less acceptable for his predecessors such as Lamarck. Lyell’s criticism of Lamarck was severe, particularly in respect of Lamarck’s ideas that there was a progression towards perfection. Lyell realized that species were the key and pondered about how they arose and became extinct without making real progress. Unlike Lamarck, he believed that species could become extinct either through physical factors or – and this was significant – through competition with other species. But as to the origin of new species he could merely say, ‘Species may have been created in succession at such times and at such places as to enable them to multiply and endure for an appointed period …’ This was a doctrine of special creation, and Darwin’s monads are in a similar line of thought, though later, in the Origin of Species, Darwin gave much attention to rejecting special creation.
It is all very well to write computer programs that will arrive at the idea of evolution by natural selection – hindsight gives wonderful wisdom – but such programs take no account of the paradigms or ideas of the time, such as the idea that species are immutable. It is often difficult to recognize how hard it was to break with current concepts.
The very title of the notebook in which Darwin wrote down his first evolutionary theory in July 1837 is in itself revealing – Transmutation of Species. From the beginning, he writes, ‘Each species changes … The simple cannot help becoming more complicated.’
In his new monad theory there was a crucial innovation: the idea of branching evolution, the tree of life. ‘Organized beings represent a tree, irregularly branched.’ But the branching model of the monad theory required the simultaneous extinction of many species, which is implausible. Darwin thus began to consider the possibility that monads have a variable lifespan; but he recognized the weakness of the idea, and by September 1837 the monads had died. He had in one way abandoned the problem of the origin of life, and this had the enormous virtue of simplifying what was already a very difficult problem. In the
summer of 1838 this is made explicit, since he enjoins himself to not to try and go too far back ‘for if so it will be necessary to show how the first eye is formed – how one nerve becomes sensitive to light … which is impossible.’
In the monad theory, with its branching tree, variation arose from an inherent tendency to progress; but this, Darwin realized, provided no explanation of its mechanism. In fact, the origin of variation was forever a problem for Darwin, for genetics was still to be developed. Why there should be variation in animal size and shape was simply unknown. He was very conscious of this gap in his theory, and he showed great intellectual courage – Gruber calls it heroic – in basing his theory on a mechanism he was unable to explain. He did, however, put forward a theory along Lamarckian lines in which acquired characters could be in principle inherited.
Here there is an important similarity with Newton, who stuck to gravity without having an explanation of its underlying mechanism. Both Newton and Darwin were driven by the data and were forced to recognize that they couldn’t explain everything. It may be a characteristic of great scientists to know what to accept and what to leave out.
In a famous passage, dated sometime in 1836, Darwin at last doubts the stability of species: ‘When I see the Islands in sight of each other, and possessed of but a scanty stock of animals, tenanted by these birds, but slightly differing in structure and filling the same place in Nature, I must suspect they are only varieties … such facts would undermine the stability of Species.’ A similar idea opens the first Transmutation of Species notebook: ‘According to this view animals on separate islands ought to become different when kept long enough apart …’ It was not until March 1837 that Darwin fully appreciated the significance of the island fauna, when the ornithologist John Gould told him of the distinctness of the hummingbirds in his collection which came from three different Galapagos islands.