The simulation hypothesis has become one of the most popular philosophical ideas of the last two decades. The modern formulation by AI analyst Nick Bostrom¹ is simple: perhaps the universe we experience is not fundamental reality at all. It is a computer simulation running inside another, more fundamental world. Our world is not then fundamental reality, but rather a simulated reality running inside a greater world that serves as its foundation.

The idea owes much of its popularity to a seemingly straightforward observation. Every generation of computers produces simulations that are more detailed and more realistic than the last. If technology continues advancing, perhaps one day we will simulate entire worlds populated by intelligent beings. If we could eventually build such simulations, perhaps someone else has already done so. Perhaps we are living inside one of them.

Many philosophers and scientists have already examined and questioned the simulation hypothesis, arguing that it is either unlikely, unsupported by evidence, or problematic under certain assumptions about consciousness and computation^2–5. These objections and the critiques they have raised are important but they generally challenge the plausibility of the hypothesis rather than the underlying logic of the argument itself. As a result, they often leave open the possibility that, under different assumptions about reality or consciousness, the simulation hypothesis could still remain viable.

Here is a different approach

Rather than debating whether simulations are technologically possible, whether simulated consciousness can exist, or whether advanced civilizations would actually create simulations, we should examine the logical foundations of the argument itself: Does the simulation argument provide valid evidence for its own conclusion?

By analyzing the relationship between the premise and the conclusion, we will show that the argument undermines the very evidence it uses to support itself. In this sense, the issue is not whether a simulated universe is possible, but whether the simulation hypothesis can ever be justified from within the world it claims may itself be a simulation.

As with any philosophical argument, appearances can be deceiving. The best way to evaluate the simulation hypothesis is to set out its reasoning step by step. By separating the premises from the conclusion, we can ask not only whether the conclusion follows, but whether accepting it undermines the very evidence on which the argument depends.

Let us begin by stating the premise (P):

P: The fundamental world is capable of building computer simulations because we ourselves can design increasingly sophisticated simulations.

From this premise, the conclusion (C) is drawn:

C: Therefore, our own world may itself be a computer simulation.

At first glance, this appears perfectly reasonable. The existence of simulations in our world is taken as evidence that simulations are possible in reality, making it plausible that we ourselves are simulated.

However, as soon as we accept this conclusion, a problem appears.

The Self-Undermining Premise

To see why the simulation hypothesis undermines its own evidence, let us suppose that C is true — that our world is merely a simulation.

If C is true, then everything we observe — including computers, programmers, and every simulation we have ever built — belongs only to the simulated world. Furthermore, every scientific theory we have developed describes only the behavior of the simulated world we inhabit. None of those theories tells us how the underlying, fundamental world operates. In particular, they cannot establish whether simulations are physically possible in that world. Thus, if we accept C as true, then our original premise P is no longer justified.

The premise is not merely the claim that “simulations exist.” The premise is that the fundamental world is capable of producing simulations. But once we accept the conclusion, we have no observations whatsoever about the fundamental world. We have observations only about a simulated world. Consequently, our evidence no longer supports the premise that began the argument.

In other words, our evidence no longer establishes “The fundamental world can build simulations.” It only establishes “Our simulated world appears capable of building simulations.” Those are completely different statements.

The simulation argument quietly replaces one proposition with the other. It begins with evidence about the world we observe and ends with a conclusion about a world we can no longer observe.This is a classic example of self-undermining reasoning.

One might object that if C is true and we indeed live in a simulation, then the fundamental world must necessarily be capable of generating simulations like ours. If so, doesn’t that imply that premise P is true after all? The problem is that this makes the argument circular. We are using P to justify C, while simultaneously using C to justify P. Neither proposition is independently established; each depends on the other.

Another objection is that the premise should be understood only conditionally: if our world is fundamental, then our observations suggest that worlds like ours can produce simulations. But this does not rescue the argument either. The evidence supports the conclusion only under the very assumption that the conclusion places in doubt — namely, that our world is fundamental. Once the conclusion is accepted, the observational basis for the premise disappears. The argument therefore remains self-undermining.

Either way, the simulation argument fails. If the conclusion is used to justify the premise, the reasoning becomes circular. If the premise is treated as an empirical claim supported by our observations, accepting the conclusion removes the very evidence that justified it. Under neither interpretation does the argument provide an independent logical justification for its conclusion.

The Recursion Problem

There is another problem with the simulation hypothesis that receives much less attention.

If a simulated world can itself create another simulated world, then simulations could, in principle, continue indefinitely.

Let us define the simulation recursion depth by N.

• N = 0: the fundamental world.
• N = 1: simulations running directly inside the fundamental world.
• N = 2: simulations running inside the N = 1 simulations.
• …
• N = k: the kth level of recursive simulations running inside the N = k − 1 simulations.

By considering this recursive structure, one thing is certain: the chain must have a beginning. At least one fundamental world must exist where the very first simulation is executed. Thus, N = 0 is necessary.

But what about every other level?

Suppose our universe were at recursion level N = 8. Would we expect to observe anything different than if we were at N = 7? Could we predict how physics, biology, mathematics, or cosmology would differ between different recursion levels? Would there be any observable feature whatsoever that distinguishes one recursion level from another?

The answer is no. We have no observation that distinguishes one recursion level from another. More importantly, the hypothesis offers no prediction about what such differences should be. Whether we occupy the first simulation, the hundredth simulation, or any level in between makes no observable difference whatsoever.

Consequently, every recursion level beyond N = 0 becomes an unobservable property. Every level greater than N = 0 is empirically indistinguishable from every other level. The hypothesis predicts exactly the same observations regardless of whether we inhabit the first simulation, the tenth, or the thousandth.

A Scientific Principle

Science does not merely tolerate unobservable entities — it demands that they do explanatory work. Atoms explain chemistry, genes explain inheritance, black holes explain astronomical observations. Each introduces new entities because those entities make successful predictions about the world.

Recursive simulation levels do none of these things. They predict nothing. They explain no physical phenomenon. They do not alter any observation we can make. Once a hypothesis introduces entities that have no observable consequences, there is little reason to multiply them. By the principle of parsimony, the simplest explanation is to stop at N = 0: one fundamental world, with no additional recursive layers unless independent evidence requires them.

The simulation hypothesis therefore faces two distinct challenges. First, its most common argument is self-undermining because accepting the conclusion removes the justification for the premise. Second, its recursive structure introduces infinitely many possible worlds that make no observable predictions and explain no natural phenomena.

Whether simulations can ever become sophisticated enough to imitate reality is an interesting engineering question. Whether we ourselves inhabit such a simulation is a philosophical one. But unless the hypothesis can produce evidence that survives its own conclusion—and predictions that distinguish one simulation level from another—it offers no scientific advantage over the far simpler assumption that we inhabit the fundamental world.

References

[1] N. Bostrom, “Are We Living in a Computer Simulation?,” The Philosophical Quarterly, vol. 53.211, pp. 243-255, 2003.

[2] J. Birch, “On the ‘Simulation Argument’ and Selective Scepticism,” Erkenntnis, vol. 78, pp. 95-107, 2013.

[3] S. Hossenfelder, “The Simulation Hypothesis is Pseudoscience,” YouTube, 2021.

[4] M. Egnor, “Of Course You Aren’t Living in a Computer Simulation. Here’s Why,” Science & Culture, 2017.

[5] M. Gleiser, “Reality is not a Simulation and Why it Matters,” IAI News, 2023.

Dr. Georgios Mappouras studied Electrical and Computer Engineering (ECE) at the National Technical University of Athens (NTUA), Greece. After graduating in 2014, he moved to the United States to pursue a Ph.D. in Computer Architecture at Duke University, which he completed in 2020. Since then, he has worked in the technology industry in Silicon Valley.