Stephen Hawking’s Philosophy: Model-Dependent Realism

“Traditionally these are questions for philosophy, but philosophy is dead. Philosophy has not kept up with modern developments in science, particularly physics. Scientists have become the bearers of the torch of discovery in our quest for knowledge.” — Stephen Hawking, The Grand Design (2010)

i) Introduction ii) Why Model-dependent Realism? iii) The Brain’s Models iv) True Reality and Models v) Pluralism vi) Hawking’s Constructive Empiricism? vii) The Aesthetics of Models viii) Conclusion

The words “model-dependent realism” (MDR) were first used in Stephen Hawking and Leonard Mlodinow’s book The Grand Design, which was published in 2010. Before that, Hawking had of course already talked about the importance of models in physics. (As a physicist, it would be hard not to stress their importance.)

For example, here’s Hawking from 1994 — some 16 years before the The Grand Design:

“If what we regards as real depends on our theory, how can we make reality the basis of our philosophy? But we cannot distinguish what is real about the universe without a theory. I therefore take the view, which has been described as simple-minded or naïve, that a theory of physics is just a mathematical model that we use to describe the results of observations… Beyond that it makes no sense to ask if it corresponds to reality, because we do not know what reality is independent of theory.”

Still, at that time Hawking hadn’t turned that emphasis on models into a philosophical position with its very own title: namely, model-dependent realism.

In the following, Hawking expresses what model-dependent realism is in a strong and clear manner:

“There is no picture- or theory-independent concept of reality. Instead we will adopt a view that we will call model-dependent realism: the idea that a physical theory or world picture is a model (generally of a mathematical nature) and a set of rules that connect the elements of the model to observations. This provides a framework with which to interpret modern science.”

Thus without theories and models we would be bombarded by a possible “infinite variety of facts” (Karl Popper). Reality, therefore, is “seen” through the lenses supplied by our (sometimes conflicting) theories or models.

One may wonder why Hawking stresses scientific models rather than scientific theories. (In fact, he sometimes appears to use the two words as virtual synonyms — perhaps for convenience’s sake.) So let the scientist and professor John Holland describe what he takes models to be:

“Although model building is not usually considered critical in the construction of scientific theory, I would claim that it is. Every time a scientist constructs a set of equations to describe the world, such as Newton’s or Maxwell’s equations, he or she is constructing a model.”

Now many would regard that as a fairly loose description of a model. Or, more accurately, a description that’s certainly at odds with our everyday use of the word model. Yet in Hawking’s case too, theories can take the form of a “mathematical models”. More specifically, mathematical models are used “to describe the results of observations”.

Why Model-Dependent Realism?

One question which needs to be asked here is the following:

Why did Stephen Hawking use the word “realism” to characterise his position?

At an intuitive level, model-dependent realism (MDR) appears to be an anti-realist position, rather than a realist one. After all, if one is stressing models, theories and “mental concepts” (see later section), then isn’t that also to stress some kind of anti-realist position? Perhaps the wording doesn’t matter.

Hawking wouldn’t have read that much contemporary philosophy of science; so this kind of conceptual clarification might have annoyed him. (Many physicists get annoyed by philosophical “conceptual analysis” because they believe it implies “conceptual conservatism”.) In any case, perhaps because MDR claims that all we have is models, theories and mental concepts, then why can’t we be realist about these things instead? That is, not realist about noumena, “objective reality” or “the Real” — but realist about our models, theories and mental concepts instead.

As already hinted at, Hawking didn’t have much time for philosophy and on more than one occasion he said that it was “dead”. Yet it’s very odd that he didn’t realise he was doing philosophy when he wrote these parts of The Grand Design. What’s more, Hawking made the following incredible claim:

“Model-dependent realism short-circuits all this argument and discussion between the realist and anti-realist schools of thought.”

So perhaps it wasn’t philosophy simpliciter that Hawking was against; but only philosophy “which has not kept up with modern developments in science”. (Many philosophers themselves have said the same about their fellow philosophers.) Thus one must now assume that model-dependent realists have kept up with modern science. However, even philosophy which has kept up with science still remains philosophy. And that’s also true of Hawking’s own model-dependent realism.

The Brain’s Models

Hawking wrote about how the human brain models the world and how scientific models do the very same thing. Indeed he seemed to have taken the brain’s models as the basis of (or inspiration for) scientific models . Hawking writes:

“According to the idea of model-dependent realism … our brains interpret the input from our sensory organs by making a model of the outside world. We form mental concepts of our home, trees, other people, the electricity that flows from wall sockets, atoms, molecules, and other universes. These mental concepts are the only reality we can know. There is no model-independent test of reality. It follows that a well-constructed model creates a reality of its own.”

The problem here is that the brain’s models are hardwired and (as it were) given — unlike scientific models. Or, I should say, there is a hardwired element on top of which we apply contingent concepts. Thus those hardwired “interpretations” take a similar role to Kant’s categories. That is, they are transcendental (in Kant’s sense).

This means that we can agree with Hawking on the necessity and importance of scientific models; though also reject his analogy with hardwired brain “pictures”.

The basic gist of the brain’s models is that they aren’t the same as what it is they model. Indeed it may even follow that they don’t even resemble what it is they model. And if this is the case as far as the brain’s models are concerned, then the situation is even trickier when it comes to scientific models.

As some philosophers have put it, we don’t have (true) representations. (Though philosophers like Jerry Fodor use the word “representation” in a very technical way.) What’s more, it’s counterproductive to believe that we do. (This was a line put forward by Richard Rorty and other pragmatists.) In any case, Hawking goes into a little detail to demonstrate his case:

“In vision, one’s brain receives a series of signals down the optic nerve. Those signals do not constitute the sort of image you would accept on your television.”

And then there’s Hawking’s philosophical conclusion. Namely:

“The brain, in other words, builds a mental picture or model.”

One problem that can be broached here is that just because the journey from the world’s causal forces to the final “mental picture or model” is both circuitous and complex, that doesn’t in itself mean that the picture (or model) can’t be a faithful (or even truthful) representation of those external causal forces. Then again, we can’t deny the contrary either.

Hawking then states something which can be taken as showing that the brain’s models are faithful or even truthful. He cites the following case:

“The brain is so good at model building that if people are fitted with glasses that turn the images in their eyes upside down, their brains, after a time, change the model so that they again see the world the right way up.”

True Reality and Models

Model-dependent realism asserts that all we can know about reality is that it

“consists of networks of world pictures that explain observations by connecting them by rules to concepts defined in models”.

Does that therefore mean (to jump forward a little) that an ultimate “theory of everything” will never be found? Hawking/Mlodinow suggest that this is unclear:

“In the history of science we have discovered a sequence of better and better theories or models, from Plato to the classical theory of Newton to modern quantum theories.”

Hawking then concludes in the following manner:

“It is natural to ask: Will this sequence eventually reach an end point, an ultimate theory of the universe, that will include all forces and predict every observation we can make, or will we continue forever finding better theories, but never one that cannot be improved upon? We do not yet have a definitive answer to this question.”

Hawking puts the well-known pessimistic meta-induction position in the above passage. This position is applied to previous theories and models and it’s argued that this provides us with a very good reason to see a model as being precisely that — a model. Thus if it’s (only) a model, then it can’t be a piece of reality. (Though models must also belong to reality — what else can they belong to?)

Ontologically, “true reality” is true reality, not a model of true reality.

Can a model simply replicate a piece of reality?

Yes; though not in literally every respect. And even if it could model everything about any given x, then what would be the point of such modelling? That’s not what scientific models are meant to do. (“A true model of the universe would be the universe itself.”) However, in everyday parlance things are complicated by the fact that that’s exactly what models are — replicas of given bits of reality (i.e., as in model aircraft, model car, model soldier, etc.).

Thus if we must model “true reality”, then it can’t actually be that true reality. And if it’s not that pure reality, then we have something less than that pure reality. What’s more, that’s a good thing because (as stated) there’s little to be gained from any kind of exact replication — at least not in theoretical physics.

Yet it is also said that those who accept the model-dependent realist position also accept that there is indeed a “reality-as-it-is-in-itself”.

Yes, “reality-as-it-is-in-itself” clearly exists. But that doesn’t mean that we can access (or describe) it as it is in itself. After all, contingent brains (with their contingent “pictures”) and persons are doing the assessing and describing. And precisely because of that, a perfect model or theory will always be out of the question.

Now what about Hawking’s “mental concepts”?

Hawking seemed to reduce scientific models to what he called “mental concepts”. And, as he put it, these “mental concepts are the only reality we can know”. However, does Hawking’s own conclusion actually follow? He boldly states:

“It follows that a well-constructed model creates a reality of its own.”

[Even an] anti-realist wouldn’t be too keen on this type of language. As it stands, it’s partly true and partly false. It’s partly true in that we gain access to “reality” through models, theories and “mental concepts”. Thus, in that basic sense, if we use different models, theories and mental concepts about the same given x, then we’ll get different “realities”.

The problem is that causal forces are completely written out of this picture. We can accept that, as the philosopher Donald Davidson argued, causal forces don’t “come under a [given] description”. Nonetheless, they are still the same causal forces which are described by different models. So such causal forces must constitute some kind of reality which can’t be dispensed with.

In addition, models are part of reality in the sense that they can hardly be anything else. And even if models are seen as abstract objects, models are still part of reality. But does a model also “create a reality of its own”? Again, both yes and no.


Hawking’s model-dependent realism is similar to the “poetic naturalism” of theoretical physicist Sean Carroll.

Firstly, here’s Hawking writing about how “different theories” can exist side by side:

“It might be that to describe the universe, we have to employ different theories in different situations. Each theory may have its own version of reality, but according to model-dependent realism, that is acceptable so long as the theories agree in their predictions whenever they overlap, that is, whenever they can both be applied.”

Sean Carroll, on the other hand, states:

“[T]here is only one, unified, physical world, but many different ways of talking about it, each of which captures an element of reality.”

The obvious question to ask both Carroll and model-dependent realists is:

How do we know if it’s the same “element” described by different models or a different element entirely?

In other words, is Carroll correct to argue that “different ontologies” can liv[e] happily alongside “the same underlying reality”?

In addition, it isn’t true that “different vocabularies” imply (or entail) “different ontologies”. At a very crude level, if someone uses the name the ‘Morning Star’ and another person uses the name the ‘Evening Star’, and both persons know that both names refer to the planet Venus, then we don’t have different ontologies on our hands here. Instead, different words, senses, technical terms, etc. can be seen to have the same ontology and therefore they posit the “same underlying reality”.

Both Carroll and Hawking hold that reality should be interpreted and that interpretation takes the form of scientific modelling. Thus if the pluralism of “poetic” elements has it that various models don’t coincide in describing the same particular x, then does that mean that many realities exist? (Hawking came close to saying this.) Again, if the models are very different, then how do we know that models/theories have the same x (or phenomenon) in their focus?

It’s also argued that

“[w]here several models are found for the same phenomena, no single model is preferable to the others within that domain of overlap”.

Yet just because different models can model any given x, that doesn’t also mean that “no single model is preferable”. It’s true that no single model need offer us what Hawking calls the “absolute truth”. However, that doesn’t also mean that a theory/model can’t be superior to its rivals. Of course spelling out a model/theory’s superiority may be a tricky business (see the last section); though scientists are already well aware of that.

To clarify this way of thinking about pluralism we can see what Hawking has to say about what physicists call “dualities”. These are “situations in which two very different theories describe the same phenomenon”. So Model 1 and Model 2 may deal with the same causal forces in different ways. That is, the descriptions are different even when the casual forces are the same. (Though how could we know that?)

Hawking complicates this by hinting that M1 and M2 deal with different “properties”, rather than saying that M1 and M2 say different things about the same properties. To be faithful to Hawking, he writes:

“Each theory can describe and explain certain properties, and neither theory can be said to be better or more real than the other.”

As stated, it’s hard to work out if Hawking meant different models of the same properties; or different models about different properties. It can be argued that properties can only come care of models or theories. Thus different models — by definition — will generate different properties. And if that’s the case, then M1 and M2 can’t be dealing with the same properties in their different ways.

Constructive Empiricism?

Immediately after making a comment about model-dependent realism short-circuiting philosophy, Hawking indulges in some philosophy of his own. Not only that: the philosophy he does indulge in is very similar to Bas van Fraassen’s constructive empiricism. Take Hawking’s following words:

“According to model-dependent realism, it is pointless to ask whether a model is real, only whether it agrees with observation. If there are two models that both agree with observation … then one cannot say that one is more real than another. One can use whichever model is more convenient in the situation under consideration.”

The words above advance the idea of empirical adequacy, which is an element of van Fraassen’s constructive empiricism.

This isn’t to say that there isn’t something real which underpins observations and models. It’s just that “the Real” (to use Slavoj Žižek’s favourite term) is basically a Kantian noumenon.

Yet the claim that different models/theories “predict the same phenomena” (i.e., they’re all “empirically adequate”) is a little loose as it stands.

Take the philosophical position of panpsychism.

Nothing in panpsychist theory is contradicted by any phenomena or observations. The same applies for many other metaphysical theories. This point is shown when A.J. Ayer commented on monism and pluralism. He said that if monism were true, then the world would be exactly the same as that posited by a pluralist.

Hawking then cites the quark as an example of a model. He wrote:

“[A]ccording to model-dependent realism, quarks exist in a model that agrees with our observations of how subnuclear particles behave.”

This means that all we have is the quark model itself and our observations. We don’t need anything “real” beneath, between or behind the model and our observations. In addition, the empirical adequacy position and MDR are basically instrumentalist in nature. That is, instrumentalists believe that models, concepts and theories should be evaluated solely in terms of their agreement with observations; as well as in terms of their explanatory and predictive power.

Hawking himself appeared to move one step beyond merely emphasising the importance — and indeed necessity — of models: he also stressed what he called “mental concepts”.

Hawking certainly didn’t believe that mental concepts and nature/reality are as one. Yet he once wrote that “mental concepts are the only reality we can know”. Furthermore, he stated: “There is no model-independent test of reality.” This seems to mean that Hawking went further than simply saying that mathematics describes (or perfectly models) nature. After all, he stressed the importance of mental concepts. However, it can still be said that the models of physics are mathematically accurate. Thus even if we require mental concepts to get at these mathematical models, the models can still perfectly capture “reality”.

The Aesthetics of Models

What constitutes a good model?

It’s here when certain aesthetic criteria are brought into play. Hawking happily admits this. So here’s Hawking’s own list of what constitutes a good model:

1) It must be elegant. 2) It must contains few arbitrary or adjustable elements. 3) It must agrees with — and explain — all existing observations. 4) It must make detailed predictions about future observations that can disprove or falsify the model if they are not borne out.

As for the first criterion: the recognition of elegance is partly subjective affair. Nonetheless, saying that the notion of elegance is subjective isn’t necessarily a criticism. It depends.

Elegance can be cashed out in very scientific (or third-person) ways. However, there’ll always be a aesthetic remainder. That is in the sense that the scientific (or third-person) criteria of an aesthetic criterion will always leave out any mention of the aesthetic term itself.

Even the criterion that models should contain “few arbitrary or adjustable elements” is not itself scientific. The word “arbitrary” will need a definition and explanation because — as it stands — it’s quite loaded. Nonetheless, that’s not to say that it can’t be defined or that arbitrary elements can’t be specified.

The final two criteria aren’t aesthetic in nature; though they can still be problematic.

Thus Hawking was correct to state that no model meets all these aesthetic criteria. (This quandary has been well-discussed in the philosophy of science.) For example, a theory which scores well when it comes to “elegance” may score poorly when it comes to “[a]gree[ing] with and explain[ing] all existing observations”.


As already stated, it’s very odd that Stephen Hawking said that “philosophy is dead” on more than one occasion. What did he think his model-dependent realism is? Above and beyond that, it’s difficult to decipher why Hawking chose the term “realism” in the first place.

As for the philosophical position of model-dependent realism itself: it has a strong pragmatic — rather than a strictly empiricist — appeal. And, precisely because of that, metaphysical realists and scientific realists will have serious problems with Hawking’s philosophical position.

23 views0 comments