In Changing Your Meta-ethics, Eliezer Yudkowsky says that it’s okay to change your meta-ethics; that changing your meta-ethics is fine because you can keep your object level ethics the same as you change your meta-ethics. In his example, if you see a child on some train tracks, you can always choose to drag the child off the tracks, regardless of what your meta-ethics say about the moral status of that child.

But is that actually true? I agree that you always have the option to drag the child off the tracks, just as you always have the option to leave the child on the tracks and let him or her get hit by an onrushing train. However, that’s not a very interesting thing to say – in any given scenario, you always have the option of doing many different things, and your options are usually not constrained by your meta-ethics, but rather by your environment and the resources available to you. What’s important, then, isn’t what you can do, but rather what you choose to do.

Meta-ethics, by design, informs what you choose to do. You can see a child on train tracks, and choose not to help that child because that child belongs to the wrong ethnic group. Or because you think the child has the wrong skin color. Or because the child is of a different religion. Or because their clan has a blood feud against yours. Or maybe you choose to help because the child is a relative of yours. Or because you believe in your common humanity. Your choice, regardless of what your choice is, is informed by a meta-ethical system. If you haven’t explicitly considered a meta-ethical system, then you default to the meta-ethical system of your society.

Going farther, Eliezer says in the introduction, that most people agree that killing is wrong, regardless of their meta-ethical systems. That is to say, varying meta-ethical system produce the same outcome even though the justifications for that outcome are different. But that’s not true either. Most notably, Eliezer conflates killing and murder. Prohibitions against murder are a human universal. Prohibitions against killing are not. This is because murder is not the mere act of killing, but also includes moral judgments about the intent of the killer, the relative social and economic statuses of the killer and the killed, and the circumstances in which the killing occurred. All of these judgments are informed by meta-ethics.

More concretely, historically, the first step in any kind of mass killing has been the altering of the perpetrators’ meta-ethical systems to remove moral value from the victims. The purpose of dehumanization is to reduce victims to a morally subhuman state, thus allowing perpetrators to carry out their acts with a clean conscience. This is especially evident in organized mass killings, all of whom are preceded by months or even years of propaganda aiming to portray the victims as not deserving humanity. Whether we’re talking about The Protocols of the Elders of Zion, or the Radio Télévision Libre des Mille Collines, organized mass killings have always been preceded by extensive propaganda efforts which alter the perpetrator’s meta-ethical systems in such a fashion that the victims are no longer regarded as human.

Eliezer says, quite glibly in my opinion, says that one can hold their will in place, preserving their object level ethical decisions while changing their meta-ethics. It’s not clear to me that that’s an easy thing to do. The entire purpose of changing your meta-ethical system is to introduce doubt into your ethical decisions. It’s to make you question whether what you’re doing is right. It strikes me as extraordinarily naive to think that you can do something this disruptive while preserving your object-level ethics. As a result, I don’t have same sanguine attitude towards changing one’s meta-ethics. Changing your meta-ethics is something that should almost never happen, and if it does, it should be a life-changing event.