Why I Don't Take Negative Utilitarianism Seriously
The page about a universe with nobody home makes a claim in passing that deserves its own space: a universe with nobody home contains exactly zero suffering, scores perfectly on that metric, and I do not think it scores perfectly.
That is a short sentence with a whole ethical position folded into it, and it puts me in direct opposition to a view a lot of thoughtful people hold. Worth unfolding.
What the view actually is
The formulation comes from Popper, in a footnote to The Open Society and Its Enemies: replace “the greatest happiness for the greatest number” with something like the least amount of avoidable suffering for all. His motivation was political rather than metaethical. Classical utilitarianism, with its mandate to maximize happiness, lends itself to utopian schemes that justify present misery by future bliss; a principle aimed at reducing concrete, identifiable suffering is much harder to weaponize that way. That is a good observation and I think it is largely correct.
Two points of pedantry, because they matter for who gets credited with what. Popper was not a utilitarian and did not use the phrase; the term negative utilitarianism was introduced by R. N. Smart in a 1958 reply in Mind, and sources disagree about exactly how much of the position Popper meant to be holding. What follows is about the view, not about Popper.
The benevolent world-exploder
Smart’s reply is the famous one. If the good is the absence of suffering, then the best available action is to painlessly extinguish all sentient life, since a world with no sentient creatures has no suffering in it at all. Smart called the agent who does this the benevolent world-exploder, and offered him as a reductio: whatever produces that conclusion has gone wrong somewhere.
I should be fair here, because the objection is old enough that the responses are well developed. Serious suffering-focused thinkers do not go around endorsing world destruction, and several have argued in detail that the implication does not go through — see, for instance, Simon Knutsson’s The World Destruction Argument . The usual moves involve uncertainty about whether the destruction would succeed, the suffering caused by attempting it, cooperative and rule-consequentialist considerations, and the observation that a view can rank outcomes without licensing anyone to bring them about. Some of these work better than others. I am not going to adjudicate them, because my objection does not depend on the world-exploder being an entailment.
Why the reductio lands differently for me
For most people, “this view implies painlessly ending all life” is a reductio because ending all life kills everyone, and killing everyone is obviously monstrous. Fair enough.
That is not quite my reason, and the difference is the whole point of this note.
The world-exploder’s end state — a cosmos with no experience in it anywhere — is precisely the state that the first four pages of the sequence identify as the worst outcome available. Not one bad outcome among others. The one that zeroes the ledger. So the view and I are not disagreeing about how much weight to put on suffering versus flourishing. We are looking at the same universe and assigning it opposite signs. It is their optimum and my floor.
That is an unusual kind of disagreement and it is worth naming as such. Two people who both think suffering is bad and joy is good can trade weights and converge. There is no weight I can assign that makes the empty universe come out well, and none they can assign that makes it come out badly, because on their view there is nothing left in it to be bad for.
The argument I actually rely on
Strip out the rhetoric and here is the load-bearing part.
Page 02 argues that we do not know who the moral patients are, cannot find out, and are stuck reasoning from a similarity prior we cannot justify. Now run both views through that uncertainty and look at what each recommends when it is wrong.
If I am wrong — if I preserve experience in the universe and experience turns out to be, on balance, a bad deal — then the universe contains beings who can notice this and act on it. The error is live, visible from the inside, and correctable by the people it concerns. There is somebody there to reconsider.
If the suffering-focused view is wrong, its error mode removes the patients. There is no one left to notice the mistake, no one for whom the correction would be a correction, and nothing that could count as evidence against the decision. The mistake becomes permanently unauditable at the moment it is made.
This is the asymmetry that page 04 says does most of the work, applied to ethics rather than to AI. Under deep uncertainty about who counts, I want the option that keeps the question open. A view whose failure mode is the permanent closure of the question is a view I am not going to bet on, however elegant the argument for it looks from inside.
The smaller, more honest reason
There is also a plainer version, which I want on the record because the elaborate argument above is partly a rationalization of it.
I do not think the point is comfort. Page 03 uses the image of demoscene hardware — a machine doing something remarkable that it has no business doing, where the constraint is precisely what makes it worth watching. That is roughly how I feel about being a person. The thing I am attached to is that something is happening in here at all, and that it is strange that anything is. Suffering is part of what is happening. I would prefer less of it, sometimes very strongly. I would not trade away the happening to get rid of it.
A view that treats experience as a container for a quantity to be minimized has, I think, mistaken the container for the contents. That is not an argument. It is a report of where my intuitions actually sit, and it seemed better to say so than to dress it up.
What I will concede
Quite a lot, actually, and the note is worth less without this section.
The asymmetry intuition is real. Most people do think there is no obligation to create a happy person, and a strong obligation not to create a miserable one. That asymmetry is not obviously confused, it has a serious literature behind it, and every attempt I have seen to explain it away — my own included — feels like motivated reasoning at some point in the chain.
Suffering-focused people are usually right about the object level. The things that motivate them — factory farming, wild-animal suffering, the sheer quantity of unattended agony that exists right now — are genuinely underweighted by almost everyone, including me. Being wrong about the axiology does not make someone wrong about where the horrors are.
I cannot beat them on rigor. I have no aggregation theory, no felicific calculus, and no account of how moral patienthood is distributed. As the note on whether I am a utilitarian admits, my positive commitments are unusually thin. Someone holding a fully specified suffering-focused view has a more complete theory than I do. They are just running it on a term whose sign I think is inverted.
Where that leaves it
I am not claiming to have refuted negative utilitarianism. I am recording that it and I evaluate the same universe in opposite directions, that the divergence is at the level of what the ledger is for rather than what is written in it, and that no amount of careful weighting will bring us together.
The practical consequence is worth stating for anyone who arrives here holding the view: the rest of this site will read as upside down to you. Everything in Part I is an argument that the empty universe is the thing to avoid at nearly any cost, and everything in Part II is machinery for avoiding it. If you think the empty universe is fine, or good, the machinery is still coherent — it just points the wrong way.