Axio › Volume 5 › Agent-Binding

Agent-Binding

From naked preferences to empirical truths

This chapter is a draft — it is readable but still changing.

“Lying is wrong.”

Is that true? Most people say yes without hesitation, and most of meta-ethics is a long quarrel about what could possibly make them right. The realist says a moral fact makes them right. The error theorist says nothing could, so they are wrong. The fictionalist says they are usefully wrong. The relativist says “right for them,” which is a way of abandoning the question while appearing to answer it.

I say the sentence, as it stands, is not the kind of thing that can be true or false at all — and that seeing exactly why dissolves the quarrel. “Lying is wrong” fails the same way “It’s raining” fails. The weather sentence is not mysterious; it is incomplete. Bind the missing variables — raining where, when — and it snaps into a proposition a glance out the window can settle. The moral sentence has a missing variable too, and it is not a place or a time. It is a valuer. Bind the claim to a specified agent — “given agent A’s values, lying is wrong” — and what was a naked preference becomes an empirically evaluable statement. I call this move agent-binding, and this chapter is about what it buys: a way through moral realism’s wreckage that is neither error theory, nor fictionalism, and — emphatically — not relativism.

Hume’s Gap

Start where the trouble starts. Hume argued that moral claims — statements about what ought to be — cannot logically follow from purely descriptive claims about what is. No inventory of facts about the world, however complete, contains a verdict. You can describe the lie, the liar, the consequences, the neurochemistry of the betrayed party’s distress, and at no point does the description entail “and therefore it was wrong.” The ought has to come from somewhere else, and Hume’s diagnosis of where — from our sentiments, our preferences — has never been answered, only resented.

I align with Hume’s skepticism entirely. Preferences alone — what I call naked preferences — cannot hold objective, universal, factual status. All value is subjective in exactly the sense I defended in the myth of objective value: value exists only in relation to agents capable of wanting and choosing, and a moral claim floated free of any agent is a preference with delusions of grandeur.

But notice what Hume’s gap does not license. It does not license the conclusion that moral claims are meaningless, or all false, or beyond rational evaluation. It licenses only the conclusion that they cannot be evaluated as stated — that something is missing. The gap is not a void. It is an unbound variable.

Binding the Agent

The general machinery here comes from Conditionalism. All truth is conditional: every truth claim depends on background conditions, and a statement earns a truth value only when those conditions are specified. In when statements fail I catalogued the ways a grammatical, confident-sounding sentence can nonetheless say nothing — and one of the failure modes was the standardless evaluation: “this is fair,” “the system is secure,” evaluations awaiting a standard, neither true nor false until the standard arrives. The cure in every case is the same: condition-binding, making the hidden variables explicit until the pseudo-proposition becomes a claim.

Agent-binding is condition-binding applied to the domain of value. A moral claim is a standardless evaluation, and in the value domain the missing standard is always a vantage — some agent’s values and goals. So bind it:

The fully bound form is what I mean by conditional objectivity. The foundation is subjective — the values belong to an agent, and nothing in the fabric of the universe endorses them. But the judgment made on that foundation is objective: given the vantage, the verdict follows or it doesn’t, and anyone can check. “X is morally wrong,” said by me, translates precisely to: given my clearly defined values and preferences, X objectively conflicts with those. That statement is true or false. It respects Hume’s gap completely — no ought has been conjured from an is. It simply relocates the moral claim to where it can be evaluated: inside the conditional, bound to the valuer whose ought it always was.

Neither Error nor Fiction

The serious modern heirs of Hume’s skepticism run the diagnosis differently, and agent-binding carves out a middle ground between them.

J.L. Mackie’s error theory holds that moral claims are systematically false: they presuppose objective values — intrinsically motivating, metaphysically “queer” properties unlike anything else in the natural world — and no such properties exist, so every moral assertion fails. Agent-binding grants Mackie the whole of his skepticism about the properties. There are no queer values woven into reality; the realist’s metaphysics is as empty as he says. But the wholesale verdict of falsity overshoots, and the taxonomy of statement failure shows exactly how. “Lying is wrong,” unbound, is not false — false claims have done the honest work of exposing themselves to refutation, and an unbound evaluation never gets that far. It is underdetermined: neither true nor false until the standard is bound. Mackie read underdetermination as falsehood. That misdiagnosis matters, because a false claim is finished, while an unbound claim is repairable — and once repaired, once bound to an agent, a great many moral claims come out plainly, checkably true. Error theory sees the missing metaphysics and declares morality a ruin. Agent-binding sees the missing variable and supplies it.

Richard Joyce’s moral fictionalism takes the other exit: moral claims are not objective truths but beneficial social fictions — morality as a useful myth we keep telling because it lubricates cooperation, despite lacking literal truth. This is more congenial than it sounds hostile; Joyce is right that moral practice earns its keep pragmatically. But fictionalism concedes too much and settles for too little. A morality maintained as a known fiction is a fragile foundation — myths hold only as long as nobody looks down — and no fiction is required. Agent-bound moral statements are not useful pretenses; they are literally true, in the same workaday sense that any conditional empirical claim is true. Where Joyce offers utility without truth, agent-binding delivers both: the pragmatic function and genuine, evaluable, conditional objectivity underneath it.

So against the realist: no universal moral facts. Against Mackie: no universal moral error either — only unbound claims awaiting their binding. Against Joyce: no fiction — bound moral claims need no myth to stand on.

Emphatically Not Relativism

Now the accusation this position always attracts. Values are subjective; moral truth is relative to an agent’s vantage — isn’t that just relativism in a lab coat?

No, and the difference is the whole point. David Deutsch has charged that physical and moral relativism deny that truth exists — often with bad intent and always with bad consequences — and he is right about relativism. Moral relativism asserts that morality is merely a product of cultural norms or personal whim, with no deeper standard by which claims can be judged or compared across contexts. On the relativist account, the slaveholding culture’s standards are as valid for it as ours are for us, and nothing more can be said. This undermines coherent moral discourse, rational accountability, and any possibility of ethical progress; in practice it is weaponized to evade critique, since no perspective can be challenged from outside its own arbitrary framework. Relativism does not humble moral judgment. It abolishes it.

Agent-binding subjectivism is built on the opposite instinct. The critical difference lies in what happens to the evaluative standard:

Because the standard is explicit, claims remain criticizable, and on multiple fronts. An agent can be meaningfully criticized when her actions fail to cohere with her own declared values — hypocrisy becomes a demonstrable error, not a difference of perspective. Her values themselves can be examined for internal contradiction, for unstated dependencies, for consequences she would herself reject if she traced them. And judgment across agents survives too, exactly as Conditionalism promised it would: where agents share conditions — and agents who want to live among other agents share far more than the relativist admits — bound moral claims are debatable between them on common ground. The subjectivist does not lose the ability to condemn the slaveholder. He states the vantage the condemnation issues from, shows that the slaveholder’s own commitments cannot survive articulation, and makes the argument in the open, which is more than the realist’s oracle ever did.

This is why the position deserves its full name: agent-binding subjectivism, resting on two principles that sound opposed and are not. Values are inherently subjective, arising from individual agents’ preferences and goals. And moral claims are objective precisely when bound to the vantage of particular agents, which supplies rigorous criteria for their evaluation. The relativist denies truth. The subjectivist clarifies truth’s locus. Deutsch’s fear — the erosion of accountability and coherent ethical discourse — is answered not by resurrecting moral realism’s metaphysics but by binding the vantage and keeping the arguments honest.

The Device in Hand

Moral judgments become rigorous, objective, and evaluable exactly when we articulate the vantage from which they are made. That is the whole device, and it is small enough to carry everywhere. It concedes Hume’s gap and builds on it rather than pretending to bridge it. It keeps Mackie’s metaphysical austerity without his verdict of universal error. It keeps Joyce’s pragmatism without his fiction. It keeps everything Deutsch is defending — truth, accountability, the possibility of being wrong — without the agent-independent values that were never there to do the defending.

What it demands in exchange is honesty about vantage. Every moral claim in the rest of this volume is bound, implicitly or explicitly, to the agents whose values give it truth conditions — and when I argue that some values are better than others, that too will be a bound claim, made from a stated vantage, criticizable in the open. Naked preferences persuade no one and settle nothing. Bound ones are the load-bearing structure of everything that follows.