I’m still working on my thesis on CEM, but for now I felt like talking about personhood. And to do that we need to establish some ideas about the act of making a definition.
When we’re trying to define something, in math or linguistics or anything, we always have in mind some objects we want the definition to capture and some we don’t. These are often the clearest examples (and non-examples) of the thing we’re trying to define. For example, if I were to try to define what a living organism is, I would obviously want humans to satisfy my definition and rocks to not satisfy my definition. There might be (and often are) some edge cases like viruses to debate about later, but from the beginning if my definition says humans aren’t alive or rocks are alive, we know something has gone terribly wrong and we have to revise our definition. There is furthermore always some notion of the particular usefulness we want the definition to have for us. For a contrived example we could define life by creating an explicit list of every organism in the universe, but such a schema would be both hard to use and not give us much insight about life. Convoluted definitions are sometimes necessary, especially in more mathematical contexts, but by and large the simpler we can make them the better, because simple definitions are easier to use and, done properly, capture no more or less information than their convoluted counterparts (i.e. the definitions are logically equivalent).
So three conditions we want a good definition of an object O to meet are the following:
We want our prototypical examples of O to satisfy our definition and our prototypical non-examples of O to not satisfy our definition.
We want our definition to be simple enough to be easy to use.
We want our definition to provide us with some sort of insight into the fundamental nature of O.
I’d conjecture* that in some cases it might actually be mathematically impossible to satisfy conditions 2 and 3, and so we will have three categories of definitions to work with. Good definitions we’ve already established will satisfy all three conditions. Satisfactory definitions will at least satisfy condition 1. Bad definitions don’t even satisfy condition 1.
Note that the reason we want a definition to satisfy condition 3 is more than simply intellectual or aesthetic. If we have insight into the nature of O we can more effectively confront novel edge cases where we aren’t quite sure if an object is O.
So we’ve established what sorts of characteristics a good definition should have. Let’s think about how we ought to go about defining personhood.
We want our definition to exclude all rocks but include all humans. We want our definition to be simple and easy to work with. We want our definition to give us some insight into personhood as a concept.
I conjecture that any such definition of personhood will either overshoot, undershoot, or provide us with no useful insight.
As a trivial example, if we define everything in existence to be a person, we will certainly capture all the things we want to be people under that definition, but we will also capture all the rocks and sand (and I don’t like sand, it’s course and rough and irritating and it gets everywhere). This is what I mean by overshooting. Some people may wish to bite the philosophical bullet and claim that, in fact, yes, everything under the sun is a person. I hope that strikes you as odd and unconvincing, but it is a viable position.
On the other hand, for another trivial example, if we tie personhood to some ability like higher cognition, we will end up excluding some disabled and comatose people. This is what I mean by undershooting, and I hope this suggestion strikes you as cruel and callous.
Now, we can capture exactly and only all humans under the umbrella of personhood by defining X as having personhood if and only if X is a living human. But this gives us little useful insight and creates tension with edge cases. I tend to think, for example, we’d license calling a sentient, peaceful, humanoid alien a person. This would be only a satisfactory definition, not a good one, under our three conditions. And that rings true to me; it’s a definition that will in almost all cases work but doesn’t really tell us all that much.
Back to my conjecture. In general suppose we define personhood as “X is a person if and only if X has C” for some characteristic C. If there exists some living human without characteristic C, we have undershot (and we might be Nazis). If there exists some non-human, un-sentient object with characteristic C, it is clear we have overshot; almost nobody wants to license calling rocks people, even very special rocks. If we have neither undershot nor overshot, then our defininition says at the very least that all humans are people and nothing non-human that isn’t sentient is a person. What I have to show to prove my conjecture is that such a definition would provide us with no especially useful insight about personhood, but would rather be a trivial sort of definition (i.e., that if C neither undershoots nor overshoots, then C is just the characteristic of being human).
Suppose our definition doesn’t undershoot and C is not the characteristic of being human. I aim to show that our definition then overshoots, and it will be QED. For any such characteristic C we can conceive of an entity X that is non-human, non-sentient and has C, and we are done; since our entity would automatically be a person under our definition, our definition overshoots.
You could object that I can’t always conceive of such an entity X, and in that case I would invite readers to think of a counterexample, some characteristic C (other than being human, or some characteristic logically equivalent to being human) such that I could not imagine even in principle a non-sentient, non-human object having C. It’s kind of silly, but you can imagine an object I call the Characteristic Rock, which is a rock handed down from God himself that has the ability to take on any attribute you like1 while remaining un-sentient and non-human.
There is one other possible objection here that comes to mind. I sort of illegitimately smuggled into the proof the assumption that C was not only not just the characteristic of being human, but also that it wasn’t the characteristic of being sentient. If we suppose that C is the characteristic of being sentient, then we have a workable model: our definition neither undershoots nor overshoots, and I would argue it gives us useful insight for talking about edge cases. I would agree with this objection and it seems to me the only possible good definition for personhood—and recall that we already established a satisfactory one in the trivial, human-centric way—must at the least involve sentience and may in fact just be sentience.
*I haven’t worked out the formal details, but it seems like this should be provable through information theory; we can contrive some absurdly complicated theorem T in set theory and establish a set of objects P={Proofs of T} where the minimum number of steps of a proof of T is absurdly large. I claim that any simple definition of P will deny us insight into the nature of P, and that some critical information will be lost when we compress such a set into something simple and easy to work with.
Any attribute that doesn’t supervene on consciousness, at any rate.