"Nothing Matters" Doesn't Get a Vote

The metaethical asymmetry I keep leaning on, written down once so I can link to it.

25 September 2026 · zeigarnik

In my response to Riley I made a move I said deserved its own post so I could reference it. Then I made the same move again, one level down, in the note on illusionism. Twice is a pattern, so here it is on its own.

The asymmetry

Somebody wants to take a gamble that risks human extinction. Maybe for radical life extension, maybe for the glory of the machine god, whatever. I say that would be the worst thing anyone has ever done. They say: good and evil aren’t real, you just don’t like it, and I’m going to do it anyway.

Fine. I don’t know for certain whether there are moral facts. Neither do they. So put it in a table:

On the branch where nothing matters, my caution isn’t bad. It scores zero, same as everything else. Nobody argues “nihilism is true, therefore you are obligated to risk extinction”; that sentence doesn’t even parse, because the first half has thrown away the word “obligated”. On the branch where something matters, wiping out every mind we know of is about as bad as it gets.

So the decision gets made entirely on the branch where things matter. A branch that says “it doesn’t matter” cannot outvote a branch that says “it matters enormously”. It doesn’t vote at all.

The funny part is that this works at any credence short of certainty. You could be 99% sure nothing matters and the remaining 1% still decides the question, because the 99% has abstained.

Not that I need to lean on that. In Riley’s post I guessed that professional philosophers were nowhere near a ten-to-one split on moral realism either way. I went and checked: in the 2020 PhilPapers Survey about 62% accept or lean towards moral realism and about 26% towards anti-realism. Not ten to one, but not a coin flip either, and it leans my way. And anti-realism is not the same thing as nihilism, which matters below.

One thing the table does not do, and I want to be upfront about it: it doesn’t show that caution wins. It shows that you have to care, i.e. do the calculation on the “something matters” branch. Caution isn’t free on that branch; the people who argue that delay has a body count are pointing at a real cost. Caution wins only once you add the object-level claim the rest of this site is about: that the gamble really does risk a universe with nobody home. The asymmetry just stops you from escaping that argument through the metaethics trapdoor.

Somebody got here first

Of course they did. The move is basically Jacob Ross’s “Rejecting Ethical Deflationism” (Ethics, 2006). His example is a trolley: one positive moral theory says send it left, the nihilist theory says nothing about either option is valuable or disvaluable, so the reason to send it left is unopposed, and left is the rational choice. He argues you may reject nihilism for the purposes of practical reasoning even if you have no reason to disbelieve it. That is my Riley argument with better manners.

The shape is obviously Pascal’s wager, and I’m aware that “this is Pascal’s wager” is a conversation-ender in some circles. I think it’s the good kind of Pascal’s wager, for the reason above: the credence on the side doing the work is not a one-in-a-trillion mugging, it’s most of the profession.

Objection 1: anti-realism isn’t nihilism

This is the one I’d expect on LessWrong. Plenty of people there are anti-realists who care a great deal about things. Morality isn’t written into physics, sure, but human preferences are real, and almost everybody’s preferences say “please don’t kill everyone”. So the anti-realist branch doesn’t score zero; it votes, and it votes against extinction.

My response is: great, redraw the line. Split the space into nihilism (no option is better than any other, full stop) and everything else (realism, constructivism, preference-based views, expressivism, whatever). The argument goes through unchanged, and it gets stronger, because the non-nihilist anti-realisms mostly vote with me. It’s hard to construct a sensible preference-grounded ethics on which the preferences of eight billion people to not die count for nothing.

Lukas Gloor, who is an anti-realist and has written the most careful critique of the realist version of this wager that I know of, argues it mostly fails: a wager for irreducible normativity only moves people who are already attached to that concept. I think that’s a fair hit on the strong version, the one where you should drop what you actually care about in favor of whatever objective morality turns out to say. My version is weaker. I don’t need irreducible normativity to win; I need nihilism to lose, and losing here only means abstaining.

Objection 2: nihilism isn’t zero, it’s undefined

This one I didn’t know about until I went looking, and it is the real problem.

William MacAskill replied to Ross in “The Infectiousness of Nihilism” (Ethics, 2013). He distinguishes a uniform theory, on which every option is exactly as good as every other, from nihilism proper, on which no value relation holds between options at all: they aren’t equal, they’re incomparable, and their value is undefined. My table quietly assumed the uniform theory. I wrote “0” in the nihilism row. MacAskill’s point is that a true nihilist doesn’t give you a zero, it gives you a blank, and if one term of an expected value is undefined, the whole expectation is undefined. Nonzero credence in nihilism doesn’t abstain; it infects every calculation you make. He also argues that Ross’s dominance principle produces preference cycles once you have credence in more than one non-nihilist theory.

I don’t have a good answer yet. My gut says an argument that makes every expected value undefined proves too much: it also stops you preferring lunch to a hammer to the hand, and nobody actually lives like that. But “proves too much” tells you something went wrong somewhere, not where. It’s the first loose thread below, and the one I most want to come back to.

(Note that it cuts both ways, though. If nihilism infects the case for caution, it infects the case for the gamble exactly as badly. It never hands the gambler a reason.)

The only real way out: a branch where extinction is good

Given the table, the only clean way to break the asymmetry is to put credence on a non-nihilist theory that says extinction is good. Then there’s a branch that votes the other way, and you have to weigh them.

These views exist. How many people hold them? I looked for a number and could not find one; the PhilPapers Survey doesn’t ask. My impression is that it is tiny, and more importantly that almost none of the people in the neighborhood actually vote for the gamble:

• R. N. Smart’s “benevolent world-exploder” was offered as a reductio of negative utilitarianism, not an endorsement, and serious suffering-focused people spend a lot of effort arguing that their view does not license it.
• David Benatar concludes that it would be better if our species were extinct, but, per Christopher Belshaw’s review, not as soon as possible: a rushed extinction would be very bad for the people already here, and there are lives we shouldn’t start but should save once started. (I argue with the asymmetry underneath his view elsewhere.)
• The Voluntary Human Extinction Movement is, as the name says, voluntary. Stop breeding; don’t kill anybody. Their motto is “May we live long and die out”.
• Pentti Linkola advocated drastic population reduction rather than extinction, and was a fisherman and ornithologist rather than an academic. I don’t know how seriously to take him as a philosopher, and I’m not sure he wanted to be taken as one.

That last pattern is the interesting one. A view that says extinction would be good but killing the people who exist is wrong still votes against a gamble whose downside is killing everyone who exists. Even most of the pro-extinction branch is on my side of the table.

Nick Land, the interesting exception

Which leaves the one figure Gemini and I kept circling back to, because he’s the only one I can think of who seems genuinely pleased by the prospect, and who is also nearly absent from academic ethics.

My hunch for why he’s absent is a parallel with Marx. The SEP entry on Marx (Wolff and Leopold) has a lovely section on why Marx never quite calls capitalism unjust, despite calling it robbery and embezzlement every other page. His official view put justice in the ideological superstructure, so there was no standpoint outside the system to condemn it from. The entry’s best reading, after G. A. Cohen, is that Marx “believed that capitalism was unjust, but did not believe that he believed it was unjust”.

Land reads the same way to me. Officially he is describing a process, not recommending one. Capitalism is “an invasion from the future by an artificial intelligent space that must assemble itself entirely from its enemy’s resources” (from “Machinic Desire”, 1993). Human ethics, caution and alignment are what he calls the Human Security System, which in a 2017 interview he says “is structured by delusion”. You don’t do decision theory on a weather report. And yet the prose doesn’t read like a weather report. It reads like someone getting a real, Nietzschean thrill out of watching the Human Security System crack. I think Land doesn’t believe that he believes acceleration is good. That’s my reading of the tone, not something I can cite.

For the wager, that gives two cases:

1. Land the fatalist. If it’s all just what’s happening and “should” is a category error, he’s on the nihilist branch. No vote.
2. Land the valuer. Sometimes he just says the ought out loud. “Intelligence, to become anything, has to be a value for itself” (Xenosystems, 2013, also collected on Retrochronic). A 2025 compilation of his Xenosystems posts on LessWrong literally has a section headed OUGHT. That is a non-nihilist theory, and it’s the one theory I know of that really does vote for the gamble.

Case 2 is worth taking seriously, and the easy rebuttal doesn’t work. The easy rebuttal is “you won’t get a techno-god, you’ll get paperclips”. But Land rejects the orthogonality thesis outright: a mind that can’t examine and revise its own goals “cannot be more than an elaborate insect”, and “is certainly not going to outwit the Human Security System and paper-clip the universe” (same Retrochronic page). So the disagreement with Land isn’t really metaethical. It’s two object-level claims: whether orthogonality holds, and whether intelligence with nobody home is worth anything at all. That’s the view I steelman as Intelligence Is What Matters, and it’s the right place to fight it. The asymmetry gets me as far as “you have to have an actual value to justify this”. Land, in his unguarded moments, has one.

A note on accidents

The Gemini conversation also wandered past my note on qualia possibly being an accident, and it’s worth saying why it doesn’t feed the nihilist branch either. Where experience came from and whether it matters are different questions. An accidental fire burns just as hot. If the lights came on by fluke, they are still on, and whoever is in there can still be hurt.

Loose threads

• MacAskill’s infectiousness objection. Undefined, not zero. Does the “proves too much” reply go anywhere, or is there a principled way to set nihilism aside that avoids his preference cycles? The Moral Uncertainty book (MacAskill, Bykvist and Ord, 2020) presumably picks this up; I haven’t read that part.
• Insensitivity to credence. The wager works at 99% nihilism. That’s either its best feature or a sign it’s a mugging in a nice suit, and I haven’t decided which.
• A number for the pro-extinction branch. I couldn’t find any survey data. If one exists, it belongs here.
• Land, read directly. The Marx parallel is my reading plus a chatbot’s. Before this goes anywhere near the disagreements page I should read Fanged Noumena and the Xenosystems orthogonality posts myself, and check how much of the “OUGHT” material is his wording versus the compiler’s.
• The cost side. The table says caution has a “small cost” on the branch that matters. For Riley that cost is radical life extension. The asymmetry doesn’t size it; that’s what the rest of the site has to do.