The Case Against Reality Review: When Fitness Beats Truth
Donald Hoffman’s The Case Against Reality begins with a surprisingly difficult question: why should evolution make us see reality as it really is? Natural selection tracks reproductive success, not philosophical accuracy. From that modest starting point Hoffman eventually arrives at an extraordinary conclusion: space, time and physical objects may be no more like objective reality than the icons on a computer desktop are like the electronics underneath.
The first claim is easy to take seriously. The difficulty is working out how much of the rest actually follows from it. Hoffman makes a strong case that evolution gives us no guarantee of a perfectly truthful picture of the world. He makes a much more ambitious claim when he argues that our perceptual interface may bear little structural resemblance to fundamental reality, and a larger leap still when he proposes that what lies behind the interface is not matter at all, but networks of conscious agents.
That progression is what makes the book so enjoyable and, at the same time, frustrating. The Case Against Reality moves through evolutionary biology, game theory, genetics, perception, consciousness, quantum mechanics and metaphysics. Even when I found myself increasingly unconvinced by where Hoffman ended up, the route there was fascinating.
Evolution Does Not Owe Us the Truth
Hoffman’s strongest argument begins with evolution. We tend to assume that accurate perception must be advantageous: an animal that sees the world correctly should outperform one living inside a distorted representation of it. But natural selection has no independent objective of producing truth. Traits spread when, in a particular environment, they contribute to reproductive success.
Suppose an organism is looking for food containing some resource. A perceptual system trying to represent the world accurately might distinguish quantities of 10, 20, 30, 40, 50 and 60. But perhaps the resource is beneficial only in an intermediate range: too little is useless and too much is toxic. Another perceptual system might divide the same objective range into something closer to bad — good — bad. It contains less information about the world while providing exactly the information the organism needs. The organism does not need to know everything that is there. It needs to know what to do.
Hoffman and his collaborators have pushed this intuition much further through evolutionary game theory. Their Fitness-Beats-Truth theorem compares, within a formal model, a strategy that estimates objective world states with one that goes directly after expected fitness payoff. Under the theorem’s assumptions, the fitness-only strategy dominates increasingly strongly as the perceptual space grows. Whatever one thinks of the extrapolation, this part of the case is not merely rhetorical: it is built around explicit evolutionary games and a mathematical theorem.
The basic intuition is easy to accept. Human perception clearly does not provide an exhaustive description of reality. We see only a small part of the electromagnetic spectrum. We do not perceive molecules, atoms or quarks directly. Other animals navigate through ultraviolet sensitivity, echolocation, chemical signals and forms of sensory information that barely enter our experience. Even something as apparently straightforward as colour is not simply sitting on an object waiting for a mind to photograph it.
Hoffman goes considerably further. In the Interface Theory of Perception, he compares perception to a graphical interface. A blue folder on a desktop does not resemble transistors, voltages, memory addresses or the physical arrangement of data inside a computer. That is precisely why it is useful. The interface hides most of the machinery and presents something adapted to our purposes.
Perhaps the apple on my desk works the same way. Perhaps space and time are part of the desktop. Perhaps physical objects are icons. That is where a compelling evolutionary argument begins turning into a much more radical theory of reality.
A Map Is Not the Territory — but It Can Still Be a Map
There is an enormous difference between saying that perception is incomplete and saying that it preserves almost none of the structure of whatever lies underneath it.
Consider a cliff. My perception of it leaves out almost everything about the underlying physical system. I do not perceive its atomic structure, electromagnetic fields, geological history or microscopic irregularities. Yet I may nevertheless represent several important relationships correctly: there is an obstacle here, it has an edge there, it lies approximately this far away, and crossing that edge will cause me to fall. Those may be genuine structural facts about the world even if the conscious image through which I encounter them is extraordinarily compressed.
Evolution could therefore give us something like:
reality → lossy compression → useful model
rather than:
reality → fitness interface whose structure need not resemble objective reality
A map leaves out nearly everything about the territory. It replaces buildings, trees, people and landscapes with lines and symbols; it can flatten a three-dimensional world onto a sheet of paper and distort scale for convenience. But the road represented on the map really does connect the two towns. The map is not the territory, yet it can still preserve relationships that are true of the territory.
This is one of the places where Hoffman’s conclusion remains disputed. In a 2026 critique of The Case Against Reality, Josef Mattes argues that applying the Fitness-Beats-Truth theorem to real biology requires assumptions that are far from innocuous. The theorem allows arbitrary non-negative fitness functions, including many that may make little biological sense, while simulations in which fitness demands change or co-evolve with organisms can leave considerably more room for approximately veridical perception.
None of that proves that our perception is truthful. It shows something narrower and more important for evaluating the book: mathematically valid results do not automatically establish how strongly the model maps onto biological evolution. I ran into a similar problem while reviewing The Doomsday Calculation. There too, clean mathematical reasoning can generate a startling conclusion while the real dispute concerns whether the assumptions built into the calculation are justified in the world to which we apply it.
Red, Qualia and the Strange Unity of the Self
Before the book gets anywhere near fundamental physics, ordinary consciousness is already sufficiently mysterious. Take red. The subjective experience of redness is one of those examples that seems to return endlessly in discussions of consciousness. Perhaps philosophers simply have an unusual obsession with red, but the example works because the problem is so immediate: how do I know that your experience of red is anything like mine?
We can agree that a strawberry is red. We can measure wavelengths and compare neural activity. We can both stop at the same traffic light. None of that gives me access to the actual experience occurring inside your mind. Perhaps what you experience when looking at something I call red would, if I could somehow experience it, resemble what I call blue. There is not even an obvious vocabulary with which to make the comparison, because every descriptive word ultimately refers back to subjective experiences of our own. It is the kind of question one can stumble upon as a child before discovering that philosophers have spent centuries trying to work out what exactly the problem is.
This is the territory of qualia: not the wavelength of red light or the neural mechanisms of colour discrimination, but what redness actually feels like. It leads naturally to David Chalmers’ famous hard problem of consciousness. We may eventually explain in extraordinary detail how a brain identifies a face, stores a memory, directs attention or generates language. Yet another question remains: why should any of those processes be accompanied by subjective experience? Why is there something it is like to be the system doing them?
Split-brain patients make the question stranger still. In the classic cases, the corpus callosum connecting the cerebral hemispheres was severed as a treatment for severe epilepsy. The traditional interpretation of some experiments suggested something remarkably close to two partially independent centres of awareness operating inside one skull. The modern picture is less clean. A major review of the split-brain evidence found extensive functional disconnection but also processes, including aspects of action control, that can remain unified. Whether consciousness itself literally divides is therefore much less settled than the popular “two minds” account suggests.
That ambiguity may be more interesting than a clean answer. What exactly makes several neural processes into one consciousness? Why does the person experiencing the world feel so indivisibly like one person?
The same problem appears across time. I instinctively regard myself as the same person I was ten years ago, even though my memories, beliefs, body and personality have all changed. Selfhood presents continuity where the underlying biological and psychological process is full of change. As I have explored elsewhere when asking what physical structure a mind actually requires, even the apparently obvious boundary between the person and the biological machinery sustaining that person becomes difficult to draw once we start replacing or removing parts.
Perhaps the unified self is itself a kind of interface: an extraordinarily useful stable object constructed from a much messier process underneath. That possibility is considerably easier to accept than Hoffman’s claim that physical reality itself works the same way, but the two ideas have an obvious family resemblance.
Conscious Realism: What Lies Behind the Interface?
Eventually Hoffman makes his largest move. He does not merely argue that perception gives us an imperfect picture of physical reality. He argues that spacetime and physical objects are themselves components of our species-specific interface.
What lies underneath, in his theory of conscious realism, is a network of what he calls conscious agents. Space, time, particles, bodies and brains are representations generated through interactions among these agents rather than the fundamental ingredients from which consciousness somehow emerges.
This reverses the conventional problem. Physicalism asks how matter can produce consciousness. Hoffman’s answer is that it does not: consciousness is fundamental, while what we call matter emerges from it. That bypasses one version of the hard problem, because subjective experience no longer has to materialise out of something wholly non-conscious. But it does not explain why fundamental reality has a conscious character in the first place. The mystery has moved into the foundations.
That is not automatically a defect. Any foundational explanation that does not continue indefinitely eventually treats something as basic. Physics may reach a field, mathematical structure or set of laws that is not explained by something deeper. Max Tegmark’s Mathematical Universe Hypothesis pushes the stopping point towards mathematics. Hoffman places it in consciousness. The relevant question is not whether his theory eventually stops explaining. It is why consciousness should be the right place to stop.
Hoffman’s universe is also extremely difficult to imagine, although that difficulty is partly the point. The instinctive picture is of conscious agents floating somewhere underneath our physical universe. But “somewhere” already assumes space. Asking what those agents were doing before spacetime existed assumes time. If spacetime really is part of the interface, neither question is properly formulated within the theory.
Hoffman is therefore not proposing another hidden physical universe underneath ours. The network of conscious agents is supposed to be fundamental reality. In their 2024 paper Interfacing consciousness, Hoffman and Robert Prentner take the simulation analogy surprisingly far: the world we experience can be understood as a simulation instantiated not on external computer hardware, but in consciousness itself.
In ordinary simulation theory the hierarchy looks roughly like this:
physical hardware → computation → simulated universe
Hoffman instead proposes something closer to:
conscious agents → interactions → experienced spacetime universe
The comparison also exposes a difficult question about time. Natural selection as we understand it involves variation, inheritance and differential reproductive success across generations. That does not by itself refute Hoffman: evolutionary history could be part of the emergent spacetime description rather than the fundamental level. But a deeper theory still needs some structure capable of grounding ordered transitions, and Hoffman’s conscious agents do have transition rules between experiences and actions.
So something doing part of the conceptual work of time survives underneath ordinary time, even if it is not time measured by clocks or embedded in a spacetime manifold. “Spacetime is not fundamental” is therefore much more precise than the looser claim that time is simply unreal.
Can Science Escape the Interface?
This is where I find the strongest objection to Hoffman’s more radical conclusion. Even if our senses are an interface, why should that prevent us from learning about what lies underneath it?
We do not perceive quarks directly. We infer them through scattering experiments, detector readings, mathematical relationships and a web of theories that make successful predictions. Human beings repeatedly use perception to discover that perception itself is misleading. The Sun appears to cross the sky around us. Solid objects appear completely solid. Time appears to pass identically for everyone. Scientific investigation has forced us to abandon all of those intuitive pictures.
The interface therefore does not seem to imprison us. We can use it to reverse-engineer itself.
Perhaps our theories will never tell us what reality “looks like in itself.” That phrase may not even make sense: looking is something perceptual organisms do. Yet we can discover relationships. If A occurs, B follows. Changing X produces Y. These quantities obey this symmetry. This equation predicts an event never previously observed.
Even if electrons, fields and spacetime turn out to be abstractions rather than the universe’s ultimate furniture, the mathematical relationships between them may still correspond to genuine structure in reality. This is close to the philosophical position known as structural realism: science may not reveal the intrinsic essence of whatever exists, but it can progressively reveal relationships and constraints that are objectively there.
There may still be limits that no instrument can cross. Even a hypothetical final theory capable of perfectly predicting every physically possible experiment could leave us asking why those equations hold, why the constants have those values, why this mathematical structure exists rather than another, or why there is anything at all. There may always be another explanation underneath; we may eventually reach a brute fact; or some distinctions may be inaccessible in principle from inside our universe.
None of this would mean that science had failed. Perfectly describing what reality does may not be identical to knowing why reality ultimately is what it is. That is very different from concluding that fitness-driven perception prevents objective knowledge.
When Does a Theory of Reality Become Science?
The evidential steps in The Case Against Reality are not equally strong. Natural selection favours traits that increase reproductive fitness rather than independently rewarding truthful representation. From this it follows that perception need not be exhaustive or literally copy the world. Hoffman then argues that perception is better understood as an interface, that spacetime and physical objects need not resemble fundamental reality, and finally that fundamental reality consists of conscious agents.
The further we move along that chain, the more additional argument is required. Hoffman’s evolutionary models make formal claims that can be analysed, criticised and modified. Conscious-agent theory has a much larger burden. If it is to become a fundamental scientific account rather than an alternative ontology layered beneath existing physics, it needs to recover the empirical successes of physics and ideally produce some consequence that distinguishes it from competing accounts.
That problem reminds me of Everett’s Many-Worlds interpretation of quantum mechanics, discussed extensively in Sean Carroll’s Something Deeply Hidden. The point of comparison is not that Many Worlds becomes meaningless unless one laboratory experiment can uniquely falsify it. The harder issue is evidential discrimination. If several different ontologies reproduce the same observable predictions, experiments confirm the mathematical structure they share much more directly than they establish which ontology is correct.
The same standard should apply to Hoffman. Deriving quantum mechanics or spacetime from conscious-agent mathematics would show that the framework can reproduce something remarkably rich. Predicting a phenomenon that competing approaches had not already predicted, followed by experimental confirmation, would provide a much stronger reason to think the proposed ontology is doing more than redescribing the same observations.
Until then there remains a gap between a mathematically interesting possible ontology and an empirically established description of reality. Elegant mathematics can establish consequences of a set of assumptions. It cannot by itself tell us that the universe chose those assumptions.
A Book Worth Disagreeing With
In some ways, much of The Case Against Reality is an enormously sophisticated version of the old question: if a tree falls in a forest and nobody is there to hear it, does it make a sound? Pressure waves propagate through the air whether anyone listens or not, but sound as a subjective experience requires an auditory system. Hoffman asks us to extend that distinction much further. What if colour works that way? Shape? Objects? Space and time themselves?
The tree does not disappear when nobody looks at it. Hoffman is not arguing for solipsism or claiming that nothing exists outside an individual mind. His claim is stranger: whatever objectively exists may not be a tree at all in the form in which our interface presents it. The tree may be the icon.
I can accept that some aspects of perception might work surprisingly like this. Evolution filters and compresses information. It produces useful categories rather than exhaustive scientific models, and it may construct features of experience that have no simple counterpart outside the perceptual system. What I do not think Hoffman has established is that the interface therefore fails to preserve meaningful structure, still less that the reality behind it consists specifically of conscious agents.
I am therefore not convinced by The Case Against Reality, but I liked it enormously. The book succeeds because its questions continue running around in your head after its argument has ended. Why should evolution favour truth? How could I ever know whether your red resembles mine? What makes many neural processes into one mind? What makes the person I am today the same person I remember being ten years ago? Could spacetime itself be emergent? Can science discover reality even if perception was never designed to show it to us? And at what point does a mathematically elegant ontology become a scientific theory of the world?
Hoffman’s strongest contribution may not be proving that reality is nothing like what we perceive. I do not think he has proved that. It is forcing us to notice how casually we assume the opposite.
We look around and instinctively treat the world as though it must basically be what it looks like. Evolution gives us no such promise. Yet the remarkable part of the story is what happened next. The same species whose senses evolved to survive rather than conduct physics learned to distrust those senses, build instruments that extend them, develop mathematics that contradicts intuition and formulate theories about structures no human being could ever perceive directly.
Perhaps the world we experience really is partly an interface. Perhaps some of its icons have almost nothing in common with whatever lies underneath them. But that does not mean reality is forever inaccessible.
The interface may be the place from which we started rather than the place where understanding has to end.
Comments
Post a Comment