Sunday, January 9, 2011

On the non-identity problem

A philosophical issue that often arises in situations whenever entities come into and less commonly go out of existence is known as the non-identity hypothesis i.e. that non-existent things lack states and in particular, that a category error has been made whenever a value judgement or moral weight is assigned to such non-existents. Within the context of antinatalism (which is what will be focused on), it's alleged that because one cannot meaningfully compare non-existence to existence, it cannot be either better or worse for the unborn to come into existence or to not. While superficially plausible, I shall advance two problems with this idea: one based on its unintuitive implications and one an attack on this hypothesis' theoretical grounding.

Firstly, and perhaps most pertinently, the non-identity hypothesis suggests that there can't exist any lives for which it's bad for them to have come into existence. Based on my observations, there seems to be a fairly broad intuition shared among many that this isn't the case; that there are indeed lives that are so bad that they aren't worth starting (in particular those with severe genetic disorders or destined to live in extreme external hardships) which immediately suggests that either most people secretly disbelieve the non-identity hypothesis or that they're unaware of its implications.

Of course, we could merely stop at those people who would have severe genetic disorders but we can go further. If it can never be bad to bring a person into existence then it can't be bad even if that person would experience an arbitrarily large amount of suffering. This means that one could effectively inflict an infinite amount of torture on someone else for absolutely no reason at all by having brought them into existence into a situation where that suffering is foreknown. This, I suspect, would be far more unintuitive to most than rejecting the non-identity hypothesis in spite of whatever intuitive support it's believed to have.

Note here that it's important that the suffering of the future individual is known ahead of time. One cannot use the non-identity hypothesis to justify any instance of torture, only those that are caused to occur to a being that doesn't concurrently exist. The reason for this is that as per the logic of the non-identity hypothesis, there are two relevant scenarios that the potential being could face; the first, either they come into existence and suffer, or they never come into existence; the second, they're already in existence and the choice is between them suffering and them not suffering. Supposedly, in the former scenario, the two cases are incomparable and hence neither is preferable while in the second, it's better for that individual not to suffer.

Using this reasoning, we can construct the following hypothetical scenario. An omnipotent killer robot is built that waits until everyone who is currently arrive is dead before carrying out an extermination of the people who came into existence after the creation of the robot. If the logic of the non-identity hypothesis holds, then we can informally show that the creation of this robot wasn't bad for anyone as follows:

1. The wrath of the killer robot is bad for someone.
2. If something is bad for someone, it would have been better for that something to not have happened to them.
3. The only way one could not have experienced the wrath of the killer robot was to never have come into existence, that wrath being a condition of their birth.
4. It would have been better for them not to have come into existence.
5. This contradicts the non-identity hypothesis.
6. The premise is false.
7. The wrath of the killer robot isn't bad for anyone.
8. The construction of the killer robot wasn't bad for anyone.

As far as I can tell, it's not possible to deny point two without undermining the non-identity hypothesis. Although the committed non-identity hypothesist would deny that it's bad to built such a genocidal robot, the above example serves as an exercise of how the hypothesis can be used to justify seemingly improbable situations.

Argument from zeroth order logic
Saving the best till last, the second objection to the non-identity hypothesis lies in an attack against the premise of the hypothesis itself; that one cannot assign states to those who don't exist. But straight away we can see that if someone doesn't exist, then they must be in the state of non-existence. If we want something more formal, we can consider the following example:

If one has hair, then one exists.
If one doesn't exist, then one doesn't have hair.

These statements are logically equivalent. If the first is true, then the second is necessarily true and vice versa. Consequently, given the truth of the first statement, we can infer the correctness of the second and furthermore deduce that a true statement has been made about something that doesn't exist. But the true statement in question applies a state to something that doesn't exist; in this case, the state of hairlessness. Thus the assertion that one can't apply states to something that doesn't exist is false by counterexample.

If one can apply states to something that doesn't exist, then there's nothing stopping one from meaningfully comparing existence to non-existence. A relevant example of this would be an instance of someone coming into existence and experiencing suffering. We could note that had they not come into existence, they wouldn't have suffered and thus as far as that instance of suffering went, non-existence would have be a preferable state in which to be.

All this argument required was a basic application of propositional calculus and subsequently, the only way to avoid the conclusions here would be to either claim that propositional calculus isn't a valid system of reasoning, or that propositional calculus doesn't apply to the situation regarding non-existing things.

Advancing the former has its obvious problems. Advancing the latter begs the question as to why such an arbitrary restriction is being imposed on logic, but more importantly, why that system of logic would be preferred given that it's less useful and less powerful than propositional calculus. All else equal, if one logic can meaningfully talk about something that the other can't, then surely the former is to be preferred.

Though this essay presents two arguments against the idea that states cannot be applied to non-existent things, the denial of the non-identity hypothesis has few implications other than that the positions of pro or antinatalism are not necessarily incoherent nor is it necessarily incoherent to speak of rights of the non-existent.


  1. I just glanced over your post because I'm pressed for time (will read it thoroughly later). The whole non-identity problem so often ignores the passage of time (which your robot scenario shows very well).

    As a student of history, I instantly grasp how incomplete it is to say "a non-existent person can't be harmed, therefore we aren't doing harm by bringing them into existence". That would be true for people who NEVER WILL come into existence. However, there "are" people who don't yet exist but WILL come into existence eventually. Imagine no further than a maternity ward five years from now.

    Therefore, although these potential people do not YET exist, it's definitely legitimate to treat them on an "as if" basis -- as if they do exist, precisely because they eventually WILL exist. This is the very basis legitimizing the common sentiment "think of our grandchildren".

  2. Another example: building a bomb and setting it up to blow up a pre-school --- six years from now.

    If the non-identity problem were a problem, this would be perfectly fine.

  3. Not quite because a) the bomb could kill adults who work with the children, b) the death or injury of the children could be of harm to their parents and other who have an interest in their survival, c) the destruction of property is harmful to the owners of the pre-school, d) the potential to incite fear in the masses as a result of a terrorist-like act and e) etcetera etcetera. As long as there are people alive when the bomb is set, this act of bomb setting cannot be ensured benign.

    I was careful in my scenario to make sure that everyone who is currently alive is dead before the killer robot starts acting, though perhaps I should have ensured that all the animals were dead too for more rigour.

  4. The Plague DoctorJune 7, 2011 at 2:01 AM

    "The construction of the killer robot wasn't bad for anyone."

    The construction itself wasn't bad for anyone; it is the combination of the birth and the construction that is bad. If you inform the parents in advance (before their child is born) of the construction of the robot (or time bomb), then the creator of the robot is not morally responsible, as the parents willingly subject their children to their future doom.

    I just thought of this, and this appears to be an additional novel (deontological) justification for omnicide, provided that everyone in the world is informed prior to contructing such a robot (or time bomb).

  5. The Plague DoctorJune 9, 2011 at 9:09 AM

    Now that I thought about it, the proviso that everyone is informed is not even necessary for this argument to hold.

  6. What about assigning states to fictional characters or objects? Wouldn't that also count as assigning states to things that don't really exist?

    (if I say, for example, that the story of Santa Claus is charming, I am assigning a state to something that does exist, the story; but if I say that Santa Claus himself is jolly, surely I am assigning a state to something that has never existed)