The Simulation Argument
Published:
Could our world be a simulation?
The idea is ancient in spirit.
Plato imagined prisoners seeing shadows.
Descartes imagined a deceiving demon.
Modern computing adds a new possibility:
a technologically advanced civilization might simulate conscious beings.
But the most famous modern case is not simply:
“We are probably simulated.”
It is Nick Bostrom’s Simulation Argument.
Argument, Not Hypothesis
This distinction matters.
The simulation hypothesis says:
we live in a simulation.
The simulation argument is a probabilistic trilemma about future civilizations and simulated observers.
They are not identical.
Bostrom’s Three Propositions
In simplified form, Bostrom argues that at least one of these must be true:
- Almost all civilizations at our stage go extinct before becoming technologically mature.
- Almost no technologically mature civilizations run large numbers of ancestor simulations.
- We are very likely living in a simulation.
The force comes from counting observers.
Ancestor Simulation
An ancestor simulation is a detailed simulation of beings like earlier biological ancestors.
The simulated agents would have experiences comparable to ours.
This assumption is crucial.
Why Observer Counting Matters
Suppose one advanced civilization creates:
billions of conscious simulated humans.
Then simulated observers could vastly outnumber biological original observers.
If you regard yourself as a random observer from the relevant class, simulation becomes statistically likely.
Simple Ratio
Suppose:
real observers = (N_R)
simulated observers = (N_S)
Then a naive self-sampling probability might be:
[ P(\text{simulated}) = \frac{N_S}{N_R+N_S} ]
If (N_S \gg N_R), the ratio approaches 1.
The Argument Is Conditional
This conclusion depends on major assumptions.
If simulated consciousness is impossible, then:
(N_S=0)
for genuine conscious simulated observers.
The argument collapses.
Substrate Independence
Bostrom’s argument assumes something like:
consciousness can be implemented computationally.
If consciousness requires biological substrate, ancestor simulations may contain behavioral replicas without subjects.
Computational Feasibility
Advanced civilizations would need enormous computational resources.
Could they simulate entire conscious populations with sufficient fidelity?
Maybe.
But this is not established.
Full-Universe Simulation Not Required
The argument does not require simulating every particle in the universe.
A simulation might render only what affects observers.
This lowers computational demands.
Level of Detail
A simulated world could use:
- approximations,
- adaptive resolution.
Video games already do this crudely.
But conscious agents may require complex internal simulation.
Civilization Survival
The first branch says technologically capable civilizations may usually disappear before reaching a mature simulation stage.
Possible causes include:
- self-destruction,
- resource limits.
The argument does not identify which.
Great Filter Connection
This resembles the Great Filter idea:
some step toward long-lived advanced civilization may be extremely difficult.
If the filter lies ahead, few civilizations reach simulation capacity.
Civilizational Choice
The second branch says advanced civilizations may choose not to run many ancestor simulations.
Reasons could include:
- ethics,
- law,
- lack of interest.
Capability does not imply use.
Ethical Cost
If simulated beings can suffer, ancestor simulation could create enormous moral harm.
Advanced civilizations might prohibit it.
This weakens the observer-count premise.
Resource Allocation
Even abundant computing may have better uses.
The number of simulations depends on priorities.
Assuming civilizations maximize ancestor simulations is unwarranted.
Reference Class Problem
The probability calculation depends on:
Which observers count as relevant?
This is the reference class problem.
Do we count:
- all humans,
- all conscious beings,
- all human-like observers?
Different classes change the ratio.
Anthropic Reasoning
The argument uses anthropic reasoning:
we reason from the fact that we are observers with certain experiences.
Anthropic reasoning is powerful but controversial.
Self-Sampling Assumption
One principle says:
reason as if you are a random sample from the set of relevant observers.
But why should this principle hold?
There is no universal agreement.
Self-Indication Assumption
Another principle gives greater weight to hypotheses containing more observers.
This can produce very different conclusions.
Anthropic probability is philosophically unstable.
Observer-Moment Counting
Should we count:
persons
or:
individual conscious moments?
A long-lived observer may contribute many more observer-moments.
This changes probabilities further.
Duplicate Observers
Suppose one simulation copies the same person one million times.
Should that make you one million times more likely to be simulated?
Many people find the implication strange.
Measure Problem
When infinitely many observers may exist, simple counting fails.
We need a measure over observers.
Cosmology already faces analogous measure problems.
Simulation Stack
A simulated civilization could create its own simulations.
Then reality contains:
simulation within simulation.
Observer numbers might grow rapidly down the stack.
Resource Limits
Each level has finite resources.
Nested simulations may have lower fidelity or fewer observers.
Still, layering complicates inference.
Base Reality
The universe containing the original computing substrate is sometimes called:
base reality.
But even if simulations exist, simulated worlds can still contain real internal objects and experiences.
Simulated Does Not Mean Unreal
If you live in a simulation:
- pain still hurts,
- relationships still matter.
Your environment would be implemented differently.
It would not become nonexistent.
Software Analogy
A virtual machine is real as a computational process.
Its files are real within the system.
Implementation level and functional reality can coexist.
Ontological Dependence
A simulated world would be grounded in another physical system.
Dependence does not imply illusion.
This repeats the whole–part lesson.
The Matrix Intuition
Popular culture imagines simulation as:
fake world vs real world.
Philosophically, the distinction is subtler.
A simulated world can be real at its own level.
Epistemic Challenge
How could we tell?
If the simulation perfectly reproduces all observations, internal evidence may be unable to distinguish it from base reality.
This resembles brain-in-a-vat skepticism.
Empirical Tests?
People have proposed searching for:
- computational artifacts,
- lattice effects,
- resource limits.
No accepted evidence currently demonstrates that our universe is simulated.
Discretization
If spacetime were discrete, would that imply simulation?
No.
A fundamental physical theory could itself be discrete.
Discreteness is not a signature of software.
Numerical Limits
If constants appeared computationally convenient, would that prove simulation?
No.
Many natural theories produce simple constants.
Pattern matching is not enough.
Glitches
Stories about:
“glitches in reality”
are anecdotal.
Memory error, coincidence
are generally more plausible explanations.
Extraordinary claims require stronger evidence.
Rendering-On-Demand Analogy
Some imagine quantum measurement as:
the universe renders only when observed.
This is an analogy, not a scientific deduction.
Quantum mechanics does not imply computer graphics.
Speed of Light as Processor Limit?
Another popular analogy treats the speed of light as:
simulation processing speed.
Physics already explains it as a structural invariant of spacetime.
The computational metaphor adds no evidence by itself.
Planck Scale as Pixel Size?
Likewise, treating Planck length as:
universe pixel size
is misleading.
The Planck scale marks where quantum-gravity effects may become important.
It is not known to be a grid resolution.
Falsifiability
If every possible observation can be explained as:
the simulator programmed it,
the hypothesis becomes unfalsifiable.
This weakens its scientific status.
Scientific vs Philosophical Hypothesis
A simulation proposal could become scientific if it predicted:
specific detectable deviations.
Without such predictions, it remains primarily metaphysical.
Prior Probability
Even without direct evidence, one can ask:
What prior probability should simulation receive?
This is difficult.
We have no empirical frequency of base universes.
Bayesian Framing
Posterior simulation probability would depend on:
[ P(\text{simulation}|\text{evidence}) \propto P(\text{evidence}|\text{simulation}) P(\text{simulation}) ]
The prior is deeply uncertain.
The likelihood may also be broad.
Too-Flexible Hypothesis
A simulator could choose almost any world.
Therefore:
[ P(E|\text{simulation}) ]
is hard to specify.
A hypothesis predicting everything predicts little.
Simplicity
Is:
we live in ordinary physical reality
simpler than:
we live in a computed world created by another civilization?
The latter adds:
- external universe,
- simulators,
- motives.
Ockham’s razor may favor ordinary realism absent evidence.
But Simulation Could Compress Laws
A defender might say:
simple computational rules could generate our complex universe.
Then simulation is not necessarily descriptively complex.
The comparison remains difficult.
Infinite Regress
If our simulators are simulated, who simulates them?
The chain could continue.
But infinite regress is not logically required.
There could be a base level.
Creator Analogy
Simulation arguments resemble theological creation arguments in some respects.
Both introduce:
- creators outside our observable world.
But simulators would be finite technological agents, not necessarily divine.
Design Evidence
Apparent fine-tuning could be interpreted as:
designer choice.
But the same underdetermination remains:
- multiverse,
- selection effects.
Simulation is one possible explanation among many.
Evil Simulator
Could simulators deceive us deliberately?
Yes in principle.
Then skepticism becomes extreme.
But possibility is not probability.
Simulation and Free Will
If our world is simulated, could we still have free will?
The answer depends on:
- determinism,
- compatibilism.
Implementation in software does not automatically remove agency.
Deterministic Simulation
A deterministic simulated agent could still be free under compatibilism.
The metaphysical issue is unchanged.
Consciousness
The entire argument hinges on whether simulated agents can be conscious.
If functionalism is true:
likely possible in principle.
If biological naturalism is true:
perhaps not.
Moral Status
If simulations contain conscious beings, creating them creates moral obligations.
Turning off a world might kill conscious populations.
The ethical implications are enormous.
Simulation Ethics
Advanced societies might therefore avoid ancestor simulations precisely because they are morally dangerous.
This loops back to proposition two.
Historical Simulation
Suppose a simulation recreates historical suffering.
If the simulated agents truly suffer, the project is ethically closer to recreating harm than depicting it.
Copies
If one conscious person is copied a million times, are there a million persons?
Under ordinary numerical identity:
yes after branching.
Simulation makes duplication ethically significant.
Probability and Identity
Anthropic reasoning becomes more difficult when copies share identical memories.
Which one should expect to be “you”?
Personal identity and probability collide.
Bostrom’s Actual Conclusion
The simulation argument does not prove:
we are simulated.
It says we should accept at least one of three broad propositions.
The argument is a trilemma.
Why This Matters
Popular summaries often skip the first two branches.
That transforms a careful conditional argument into a sensational claim.
Could All Three Be Partly True?
Yes.
Civilizations might:
- often fail,
- sometimes abstain,
- occasionally simulate.
The original trilemma concerns limiting proportions, not simplistic exclusivity.
Empirical Ignorance
We do not know:
- probability of technological maturity,
- computational capacity,
- ethics of advanced civilizations.
Therefore numerical confidence is premature.
Anthropic Uncertainty
We also lack consensus about:
- reference classes,
- observer weighting.
These are not minor technical details.
They drive the conclusion.
The Simulation Argument as Philosophy
Its value may lie less in proving simulation and more in forcing questions about:
- consciousness,
- computation,
- identity,
- probability.
It connects many themes of this series.
Reality Inside a Simulation
Even if simulated, our experienced regularities would still form:
our physics.
Science would remain valid at our level.
The ontology beneath it would change.
The Philosophical Lesson
The simulation argument is not evidence that we live in a simulation.
It is a conditional anthropic argument:
if advanced civilizations create huge numbers of conscious simulations, simulated observers may dominate the reference class.
Its force depends on uncertain assumptions.
The correct conclusion is not certainty.
It is conceptual caution.
The Next Question
The simulation argument asks what probability we should assign.
A more direct question remains:
Could we actually be living in a simulation?
What evidence could distinguish that possibility from ordinary reality?
The next essay asks:
Could We Be Living in a Simulation?