The simulation hypothesis: are we in a computer?

This file does not begin on a night. There is no light in the sky, no radar return, no witness calling the police. It begins with a question, set down in a philosophy journal in 2003.

Scene · reconstructed from the paperworkPicture it. Not a scene, really; a page. In Philosophical Quarterly, volume 53, issue 211[bostrom-2003], the philosopher Nick Bostrom published a paper whose title is itself a question: Are We Living in a Computer Simulation? The file doesn't say where he wrote it or how long it took. It says what the paper argues, and the paper is still freely available from the author's own site[bostrom-2003].

The question sounds, at first, like late-night talk. Could the world around us be a calculation running somewhere, with us inside it? Films had played with that for years. What the paper does is different. It does not ask whether the idea feels true. It asks what would follow, logically, if certain things about the future of civilisations were true. And it sets out the answer as a choice between three possibilities.

I want to state the one thing in this file that gets lost most often, before anything else. Bostrom does not claim we are simulated[bostrom-2003]. His argument says only that at least one of three options must hold. He does not say which. The headline version, that a philosopher proved we live in a computer, is not what the paper says. This archive keeps that line in its margin for a reason.

So what kind of case is this? It is the strangest kind we keep. There is no event, no object, no document from a government. The grade is claimed and the Finding is speculation (the record[bostrom-2003]), because what we have is an argument, not an observation. The evidence stamp is the lowest we give. That is not a judgement of the argument's quality. It is a statement about what kind of evidence exists.

And yet it has a paper trail of its own, and it is a real one:

• the original paper, with its journal reference and DOI (the record[bostrom-2003]);
• a critical reply in the same journal the same year (the record[weatherson-2003]);
• a physics preprint asking whether the idea could leave traces (the record[beane-2012]);
• Bostrom explaining the argument on camera (the record[bostrom-yt]).

The cover of this file is the Hubble eXtreme Deep Field, thousands of galaxies in a tiny patch of sky. It is there to show the scale of what the question is about. It is not evidence of anything in this case, and the file says so.

So the night of this file is a page in a journal. The next thing to do is read what is on it.

The heart of Bostrom's 2003 paper[bostrom-2003] is a trilemma. As the dossier summarises it, at least one of these is true:

• One. Civilisations like ours almost never reach the point where they can simulate minds. Perhaps they die out first (the record[bostrom-2003]).
• Two. Civilisations that do reach that point almost never choose to run such simulations. Perhaps they lose interest (the record[bostrom-2003]).
• Three. We are almost certainly living in a simulation ourselves (the record[bostrom-2003]).

The reasoning, as Bostrom explains it on camera[bostrom-yt], runs roughly like this. If advanced civilisations can run simulations of minds, and if they actually run a great many of them, then simulated minds would vastly outnumber original ones. A mind picked at random would then most likely be a simulated one. Block that conclusion and you have to accept one of the first two options instead.

Read that way, the paper is not a claim about our world. It is a claim about the shape of the choice. It closes off the comfortable middle: you cannot, according to the argument[bostrom-2003], believe that our descendants will run vast numbers of ancestor simulations and that we are not in one.

The paper did not go unanswered. In the same journal and the same year, Brian Weatherson published Are You a Sim?[weatherson-2003], a critical reply. As the dossier describes it, Weatherson argues that the indifference reasoning behind the argument, the step from "most minds are simulated" to "I am probably simulated", does not force the conclusion. That is a philosopher's objection to a philosopher's step, and it is the first critique in this file's timeline.

For nearly a decade, that is where the record stays: in philosophy. Then in October 2012 three physicists, Silas R. Beane, Zohreh Davoudi and Martin J. Savage, posted Constraints on the Universe as a Numerical Simulation[beane-2012] to the preprint server arXiv. It appeared in the European Physical Journal A in 2014.

Their field already simulates small pieces of nature on a grid, a lattice, inside large computers. Their question, per the paper's summary[beane-2012], was practical: if our universe were computed on a lattice like that, would the grid leave fingerprints?

So the documents, in order:

• 2003: the argument (Bostrom[bostrom-2003]);
• 2003: the first reply (Weatherson[weatherson-2003]);
• 2012: the preprint (Beane, Davoudi and Savage[beane-2012]);
• 2014: its publication (the record[beane-2012]).

That third document is where the file changes character. It is the moment a thought experiment was asked to produce a measurement.

The turn in this file is the moment philosophy handed the question to physics, and physics looked for pixels.

Beane, Davoudi and Savage[beane-2012] proposed that a lattice-based simulation could leave signatures in what we observe. Their example was the cosmic rays with the very highest energies. On a grid, those particles might not arrive evenly from every direction. There could be a subtle preference, an anisotropy, lined up with the structure underneath (the record[beane-2012]).

I find this the most interesting move in the whole case. It turns a puzzle about probability into something you could, in principle, go and look for. It does not need anyone to believe the hypothesis. It only asks: if it were so, what would we see?

And here the file has to be plain. The preprint proposes a test; it does not report a result, and this archive holds no observation that supports the hypothesis. The analysis[bostrom-2003] puts it in two lines: what is documented is the argument and proposals to test it; what is not documented is any evidence for it.

Now the fair other side. The critique in this file is Weatherson's reply[weatherson-2003], and it attacks one joint: the indifference reasoning, the step from "most minds are simulated" to "I am probably simulated". If that step fails, the conclusion is not forced, and much of the force of the trilemma goes with it.

So where did the story bend? Not in the paper. Bostrom's paper[bostrom-2003] is careful; it offers three doors and does not say which one is open. The bend came as the idea travelled. "One of three must be true" became "we are probably in a computer", and then "a philosopher proved it". The paper's own margin, kept in this file, says otherwise: only that one of three options holds[bostrom-2003].

That leaves a question that may not have an answer at all.

This file's status is open, and I want to be honest about why. Not because new evidence is on its way, but because the central question may never be closed.

The open question is simple: is the simulation hypothesis testable at all?

Here is what the documents do not settle:

• Testability. Beane, Davoudi and Savage[beane-2012] proposed one way to look, through the arrival directions of the highest-energy cosmic rays. That is a proposal, not a result. Whether a sufficiently good simulation would leave any trace at all is not something the file can answer.
• The first option. Whether civilisations survive to simulation-capable technology is not something the argument[bostrom-2003] can tell us. It only tells us what follows either way.
• The second option. Whether such civilisations would choose to run simulations is a question about minds we have never met (the record[bostrom-2003]).
• The reasoning step. Whether the indifference principle really forces the conclusion, which Weatherson[weatherson-2003] disputes.

And here is what is not open. A serious philosophical argument for the idea exists, published and debated since 2003 (the record[bostrom-2003]). That is what the file shows. What it does not prove is that we live in a simulation, and the file's Finding, speculation, will not be made more exciting than that.

I keep coming back to the archivist's note at the top of this file: a theory whose evidence would, by design, be hard to find (the record[bostrom-2003]). That is not a dismissal. Plenty of good questions are hard. But it is a warning about how to read headlines. When you see one, ask which of the three doors it is talking about.

If the hypothesis is testable, this file holds no result either way. If it is not, it may stay where it started: a careful argument, with three doors, and no way to look behind them. You can hear Bostrom set out the argument himself[bostrom-yt], in his own words, in the video below.

The paper, the reply, the physics preprint and the timeline are below. Start with the original; it is shorter and more careful than its reputation.