Reviewing "What Is This Thing Called Science?" by Alan Chalmers

What exactly is science? Is there anything particularly special about scientific knowledge that distinguishes it from other forms of knowledge - something that justifies our inherent trust in "scientists" and "scientific" results? Moreover, is there a "structure" or "logic"to how science has been done? Many of us feel that over the past few centuries, science has in some sense progressed - theories are constantly being revised and overthrown in favour of new ones - however, is what is the criterion by which we can judge a scientific theory to have been an "improvement" over its predecessors? If such a logic can be formalised, may it provide a guide to the conduct of science today?

Such are some of the questions raised by philosopher of science Alan Chalmers in this landmark introduction to the field. His work is essentially an exploration of the question of what defines scientific knowledge and the practice of science. He achieves this by scrutinising some of the key ideas that have dominated the philosophy of science over the past century, some of which can be associated with particular figures such as Karl Popper or Thomas Kuhn.

Themes that appear prominently in the book are the proper relationship between theory and experiment, and whether the objects of scientific theories, such as quarks and electric fields, can be considered "real", or are merely convenient tools. His exploration begins with the idea that theories can be derived directly from observed facts.

I. Induction

The most powerful affirmation of truth is obtained via logical deduction, the process by which the truth of certain statements follows from the truth of other statements. However, this process cannot be used for the construction of scientific theories because such theories tend to speak of general laws, while all that can be observed are specific cases. For example, I can see that a swan is white, but that cannot be generalised via deduction to the rule that all swans are white.

An alternative would be the process of "induction"*, where a sufficient body of data can provide the means to arrive at general laws. If I see many swans at many instances, over a wide range of locations, and they are all white, maybe I can conclude that all swans are white.

*Chalmers never mentions him, but Francis Bacon (1561-1626) was an influential early proponent of "inductionism".

However, as Chalmers points out, this runs into two major issues. One is that it is non-trivial to construct a specific of criteria by which an inductive generalisation is valid. How many swans would I have to observe? In how broad a range of locations would I have to observe them? The other problem is potentially more profound and it relates to the notion that what is "fact" is not fully determined by sensory input, but is also influenced by internal facts. This complicates the assumption that we can directly know what is in the world through observation.

One trivial example is in so-called "ambiguous" images. In the example below, you may be able to see either a duck or a rabbit (your perception may flip between both in a "Gestalt switch"). However, in either of the cases, the sensory input is exactly the same - the image never changes. In a more scientifically significant example, when Galileo turned his telescope skyward and recorded to observation of moons orbiting around Jupiter*, he was forced to contend with the criticism that the spots he was observing were actually artefacts produced by the telescope. This was despite the fact that Galileo and his critics were looking at the same thing.

*Interestingly, Galileo made observations of a "star" near Jupiter, the positions of which match with what we now know is the planet Neptune.

How could scientific theories be unambiguously induced from the data if there is a potential to disagree on what the data actually represents? Indeed, this problem may become more acute in cases where different underlying theories can produce the same or very similar results. For example, Albert Einstein made the point that despite the "essentially different bases" underlying the Theory of Relativity and Newtonian physics, each theory "in its consequences leads to a large measure of agreement with experience". Thus any attempt to logically derive mechanical laws from the "ultimate data of experience" is "doomed to failure", according to him.

II. Karl Popper's "Falsificationism"

These criticisms of inductionism undermine the case that theories can be induced "from" the data. So how else may they be generated? Austrian-British philosopher Karl Popper and others in his school of thought asserted that theories just emerged via the mysterious intuitions and creativity of human beings. Their formation does not form part of the scientific method. Hence, the ultimate method of weaning out inferior from superior theories was to test them against evidence. If a theory is "falsified" by observations inconsistent with its predictions, it should be discarded.

However, one issue with this interpretation is that it does not accord with the history of science, where, upon detecting "anomalous" observations that conflict with a theory, scientists will tend to persist with the theory and attempt to demonstrate how the anomalous result was in fact consistent with the theory to begin with. As Chalmers explains, this is because the predictions of a theory are a product not only of the theory itself, but also of the "background knowledge" that describes the state of the world under which a prediction is made.

For example, the Newtonian theory produced predictions of the orbital motions of the planets around the sun that were satisfactorily consonant with observation - that is, except for the cases of Mercury and Uranus. The anomalous orbit of Mercury was not resolved until the arrival of the Theory of Relativity; however, the case of Uranus was dealt with in a way that eventually demonstrated its consistency with Newtonian mechanics, despite initial appearances to the contrary. The "background knowledge" of the time asserted the existence of all the planets up to Uranus. However, French astronomer Urbain Le Verrier asserted that Uranus' orbit would be consistent with Newtonian mechanics if it was somehow being gravitationally influenced by an eighth planet, yet unobserved. He made calculations that predicted where Neptune should be, and lo and behold, in September 1846, German astronomer Johann Galle observed the planet. Rather than modifying Newtonian mechanics in the face of anomalous observations, the scientists of the day modified their background knowledge instead.

III. Thomas Kuhn's "Paradigms"

American philosopher of science Thomas Kuhn attempted to formalise this idea by posing a structure to the way in which theories are thrown out in favour of new ones. He published his ideas in an influential book titled The Structure of Scientific Revolutions. He maintained that science is conducted by way of paradigms - these are theories, as well as their associated methods and philosophies. Ptolematic astronomy, Dalton's atomic theory, Newtonian physics and the Theory of Relativity are all paradigms.

In Kuhn's framework, scientific progress takes the form of abrupt shifts from one paradigm to another. During a period when a given paradigm is dominant, "normal science" aims primarily to articulate the ways in which the paradigm can be applied to a range of problems - for example, adapting Newton's laws to fluids, or applying Schrödinger's equation to systems such as the hydrogen atom.

Eventually, anomalies appear such as the orbits of Uranus and Mercury (or phenomena such as blackbody radiation). Once the weight of the anomalies becomes acute, the scientific community enters a "crisis period". Now, radical attempts are made to remedy the anomalies. Some scientists will attempt to address the issue by formulating new paradigms that overturn the fundamental assumptions underlying the existing paradigm. In the end, the community as a whole will jump ship and adopt a new paradigm, whereupon "normal science" will resume again.

One significant contribution of Kuhn's approach is that it recognises the existence of a scientific community, which serves to shift the focus away from abstract principles and elucidate how science is actually done. However, one major problem (as Chalmers sees it) is that Kuhn's framework is essentially "relativist" in the sense that there is no overarching criterion that deems one theory superior to another - not even to one that is has superseded.

This is because Kuhn emphasised that paradigms are generally "incommensurable" in the sense that scientists that subscribe to different paradigms essentially live in different worlds; they may see the same thing, but they perceive different things. He memorably compares Antoine Lavoisier, who elucidated the role of oxygen in combustion, and Joseph Priestley, who defended the idea that a substance "phlogiston", was responsible for combustion - phlogiston was present in combustible materials, and was released by the process of combustion itself. Kuhn maintains that while Lavoisier saw oxygen, Priestley saw dephlogisticated air - they lived in different worlds.

If each paradigm was associated with its own "world", its own standards, practices and theories, then how was one paradigm to be compared objectively to its competitors? Kuhn resorts to the authority of the scientific community as a whole, noting that consensus will be the judge.

IV. Imre Lakatos' "Research Programs"

Hungarian philosopher Imre Lakatos attempted to recover the sense of "progress" apparently lost in Kuhn's account with his own framework of "research programs". This was intended as a way of reconciling Popper's falsificationism, which emphasised the rejection of theories based on contradictory evidence, and Kuhn's ideas, which highlight that attempts frequently are (and should be) made to resolve apparent contradictions within an existing paradigm before rejecting it.

Lakatos' programs are similar to Kuhn's paradigms. They each are constituted by a "hard core" on which the theory is built. For example, the core of Newtonian mechanics is Newton's three laws of motion, plus his law of gravity. However, predictions can only be made when this core is augmented by a "protective belt" of auxiliary hypotheses, including background knowledge and the methods by which the laws specified by the core should be applied.

When predictions are observed to be false, then either the theory or the auxiliary hypothesis must be false. While Popper sought to pin the blame on the theory, Lakatos made clear that in order to preserve the coherence of science, it is the belt that must be modified in order to protect the core in the light of anomalous evidence. The belt must be modified in such a way that the evidence is accounted for. However, it should not be modified in an ad hoc way such that the new hypotheses are only testable by reference to the observations that they were invoked to explain. For example, the existence of Neptune was not an ad hoc modification insofar as its existence was testable in a manner independent of the original rationale for the hypothesis (i.e., the orbit of Uranus).

Indeed, this example demonstrates the ways in which anomalous results can be progressive in the sense that the attempt to modify the protective belt and save the theory gives rise to new scientific knowledge. However, if observations can be saved only with resort to ad hoc assumptions, then the research program can be said to be degenerative. The conduct of (normal) science within this program does not lead to any new, independently testable knowledge, and it is high time for a scientific revolution.

Chalmers credits Lakatos' framework with resolving the question of whether the theory or the background knowledge should be criticised in the face of anomalous evidence, a problem found in Popper's original methodology. The core should be preserved as much as possible, and only rejected in favour of a new theory if attempts to save it become degenerative.

However, he criticises the lack of any prescriptive standard for progress in the framework. It is true that in hindsight, it is easy to see that a paradigm had become degenerative, and so it had to be replaced with a superior one. However, in the current moment, may be impossible to predict whether a theory should be replaced, or whether the continued modification of its protective belt will eventually lead to its rescue. In one sense, it is correct to persevere, since the theory, hitherto so successful, might be rescued after all. Lakatos does not specify exactly when it should be abandoned. Therefore, a certain ambiguity is present as to what should be done.

V. Progress

Chalmers sought to push back against the relativist position inherent in Kuhn's account, and the ambiguity found in that of Lakatos. He agrees that there might not be a universal method that can prescribe criteria and methods for science across all fields at all points in time. However, he notes that a theory can be accepted by a scientific community according to a set of standards present at that point in time, shared between proponents of rival theories. For example, astronomers in the Ptolematic and Copernican traditions, who envision sun- and earth-centred universes respectively, both sought to describe in a simple and general way the motion of heavenly bodies. Despite the fact that their theories were so different, they were able to agree on shared goals and principles, and over an extended period, the Copernican worldview came to dominate, given the some observations provided by figures such as Galileo. Therefore, there is a sense in which science progresses, even if something like a prescriptive universal method remains unfeasible.

In this sense, he does not replace the ideas of Kuhn and Lakatos with anything better, rather, he is admitting that their shared goal of building a universal account of science is futile. In its place, he offers an account that invokes a local sense of progress. Theories can progress from one to another according to contemporary standards - however, it is futile to prescribe standards that should encompass science across all fields and at all times.

VI. Reality

Throughout the book, Chalmers describes several other points of view, including the Bayesian approach to evidence-supported scientific theories, and approaches that emphasise the theory-agnostic acquisition of experimental knowledge.

However, the aspect I found the most intriguing was Chalmers' discussion of the "realist" position, which views scientific theories as describing things that are actually there in the world. Chalmers' own position is that there are things out there in the world that we can describe with our theories, and whose existence we can confirm via observation. For example, the bending of light from a star around the sun is something that we can observe. However, it is simultaneously true that in our attempt to confirm theories by way of observational evidence, we must always be very careful to know what it is that could be verified by our evidence.

For example, the bending of light around the sun provides evidence for Einstein's law of gravitation, as opposed to the prevailing Newtonian law. However, it does not provide support for his Theory of Relativity as a whole, of which the law of gravitation and the phenomenon of gravitational lensing is only a part. Hence, a theory must be "partitioned" into those parts that are experimentally verified and those that are not.

This idea of "partitioning", attributed to Deborah Mayo, can serve to provide a solid basis of evidence that we know to be true and that would not be overruled by future science. This serves to insulate scientific knowledge from the "anti-realist" charge that in the history of science, theories have very often been overthrown, so what guarantee do we have that our current scientific knowledge will persist?

In response, Chalmers notes that those aspects of previous theories that have been overthrown were precisely those that were not experimentally verified. In this vein, he contrasts two entities, the electron and the ether, both of which are unobservable by the human eye, but which have suffered different fates. While the ether concept has been rejected, the electron remains a vital components of modern particle physics.

Despite its current rejection, the ether was in fact a fundamental basis adopted by James Clerk Maxwell to construct his electromagnetic theory. Electric and magnetic fields were characterised as mechanical states of stress and tension in the ether, and so on. His theory was to be later confirmed by experiments such as the artificial production of electromagnetic waves using accelerating charges, and it still remains an integral part of modern physics. However, it so happened that the picture of an ether, on which he based his theory, was completely redundant to all the mathematics that emerged from the theory. The mathematical aspect was experimentally confirmed, while the ether concept was not. Maxwell's theory was "partitioned" in such a sense.

On the other hand, the existence of the electron - as a charged particle with a certain mass and spin etc. - has been confirmed to a degree that distinguishes its case from that of the ether, which received no independent experimental confirmation of its existence. In this sense, the electron can be said to be "real", while the same privilege could never be accorded to the ether.

Chalmers does leave open the possibility that those "confirmed" aspects will be revised and modified, but he maintains that they would always remain as "limiting cases" of subsequent theories. For example, Newtonian mechanics was subject to such severe tests that his law of gravity and laws of motion could be said to be real. However, as we know, they were eventually shown to be inadequate when large masses or speeds are involved. Despite this, Newtonian mechanics is regularly used to predict the trajectories of satellites and comets due to their approximate correctness in a certain regime.

I must admit, I am somewhat uncomfortable with this notion of "reality" as accepting those laws that approximate the "truth", as does Newton's laws. Surely, in a strict sense, Newton physics is plain wrong despite its usefulness in everyday regimes of mass and speed? In a mathematical sense, it is an approximation of Einstein's physics; however, in a conceptual sense, it is utterly different. There is no equivalence of mass and energy, no time dilation, no length contraction, or warped spacetime in Newtonian physics. The two theories are totally incommensurable, so in a conceptual sense, I do not think it is correct to say that Newtonian physics is an approximation of Einstein's physics. It so happens that in certain regimes, they agree mathematically to a certain degree.

At the same time, I agree with the "partitioning" approach; though I think we have to be careful about what we are confirming via experiment and observation. Are we confirming Newton's law of gravity as an ultimate rule, or Newton's law of gravity as a viable approximation at certain regimes? We cannot make that judgement in a theory-independent manner - in other words, in a manner that is resistant to future scientific progress.

Will we ever arrive at a description of reality as it actually is? I don't think there is any guarantee that we will. The human science-forming capacity is powerful yet limited by biological constraints, an argument made by the so-called "new mysterians" such as Thomas Nagel, Colin McGinn and Noam Chomsky. These thinkers speculate that there are some mysteries that can never be solved by the human mind given these constraints. Perhaps the ultimate laws of physics constitute one of these mysteries.

VII: Laws

Another question of fundamental importance discussed by Chalmers is why the world should obey scientific "laws". He first attempts to characterise what a "law" exactly is, then tackles the question of why the world should obey laws.

He rejects the idea that fundamental laws constitute "regularities" in events, for two reasons. First, there are those regularities that are always present, yet which presumably no physicist would label as a physical law. He cites the example, first given by Karl Popper, that "no moa [an extinct flightless bird] lives beyond fifty years" (p.198). Even if this was true in the case of every moa that ever lived, it is surely not a fundamental law in any sense. Moreover, it is also true a physical law, described by, say, an equation of motion, is often in operation when that equation is in fact not followed. For example, Galileo's law of fall, stating that objects all with equal acceleration g regardless of their weight, tends not to describe the motion of a feather. However, that does not mean that the gravitational phenomenon described by Galileo does not operate in that case. It does, but it does so in conjunction with winds and air resistance etc. The law of fall would only capture the motion of feather when it falls in a vacuum, which rarely ever happens.

He instead adopts the position that a law is something ascribes "powers", "capacities" or "dispositions" to entities in the physical world. For example, the law of Coulomb force ascribes some attractive and repulsive capacities to particles based on a property called their "electric charge". Chalmers contends that this notion of "law" applies for almost every law in physics, but also that it runs into difficulties in the case of laws such as the first and second laws of thermodynamics, and some conservation laws, such as the conservation of energy, momentum and electric charge. In these cases, what "entities" are ascribed powers and capacities, by the law?

As for the question of why laws, such as the "capacity" laws and conservation laws, should in fact describe the world in which we live. His answer is largely negative: "I don't know. They just do. I am not entirely comfortable with this situation, but I don't see how it can be avoided" (p.208).

I think the question of why the world is governed by fundamental laws is in a sense pointless. Given that the universe does indeed exist, it has to be governed by some laws, including, in the extreme sense, a law that simply describes every event that happens and every entity that exists, without any kind of generalisability whatsoever.

The mystery is perhaps why these laws should adhere to a form that is general and that appears comprehensible to a specifically human mind. According to the new mysterians, our science-forming faculty is limited by our biology (I think this is reasonable), and it seems remarkable that the laws that we have so far constructed using our innate mathematical abilities do track will with reality, a state of affairs that the physicist Eugene Wigner famously called the "unreasonable effectiveness of mathematics in the natural sciences". Wigner states that the appropriateness of mathematics to describe the laws of physics is a "wonderful gift which we neither understand or deserve" and that we should "hope that it will remain valid in future research and that it will extend, for better or for worse, to our pleasure, even though perhaps also to our bafflement, to wide branches of learning".

VIII. Reflections

Chalmers' book is an excellent introduction to the philosophy of science. Reading it, I was familiar with some of the ideas, such as inductionism and Kuhn's paradigms; however, there were some I was not familiar with, such as Feyerabend's "anarchism" and Lakatos' "research programs", and I felt that I could grasp the essential elements after reading the book.

I also felt that Chalmers was even-handed in his critique of opposing positions, whether they be theory- and experiment-centric approaches, or the realist vs anti-realist ontology. In most cases, he took a middle-road position between extremes.

Perhaps inevitably for an introductory text, the ideas seem somewhat disjointed and there is little historical context for their emergence, except for the remarks on how Lakatos, as a student of Popper, intended to modify his mentor's views to account for Kuhn's framework. Indeed, this was not an aim of Chalmers in writing his book, and nevertheless, he does provide some historical passages, as well as references to many other works that would probably focus more on this aspect.