Identity is a complex thing. Philosophers have been thinking about it for millenia, it lies at the heart of such fields as semantics and ontology, and I’d argue that everyone has, at some point, thought about it.
In computer science, and logic at large, identity is a fundamental topic, and yet it is not something that my peers often talk about—but maybe I’m just hanging out with the wrong folks.
Today, I want to spend some time capturing some embryonic observations on identity. Maybe I will succeed in making them interesting to technologists, or maybe it will remain idle thought. Maybe I will flesh them out at some point, or maybe I won’t. I’m content with either.
What I cannot do, and will not attempt, is capture the essence of identity as a concept. I’m not a philosopher, and even if I were, giving definitive answers is not something I am very good at.
The illusion of identity
It seems simple to have an intuitive understanding of identity, especially in the natural world: the pot on my stove is an entity, distinct from all other pots. The countertop next to it is a bit worn, and I can identify it by looking at all the individual stains and imperfections. But what if I were to take the lid of the pot, or refurbish the countertop, removing all the stains and planing it? They would still be the same pot and the same countertop.
This is true for living organisms as well. If I dye my hair and decide to become a carpenter, many of my features would change, but I would still be myself.
In some programming languages, values are defined by where in memory they lie. They are mutable. A complex object stays the same object, even if a property changes. For smaller objects this might not be as clear: if I change a coordinate of a two-dimensional vector, even if it is in place, is it still the same vector? This might be a clash of the conceptul vector, and the vector object as it exists in the “real”, or in this case virtual, world, or it might be an indication that something about my definition of identity is not quite adequate1.
Other languages define it by equality: things being equal are identical. But this might fall short in certain areas: two file objects might not compare the actual file descriptors, but only the name of the opened file, and the mode in which they were accessed. Is their identity the same, since they are considered equal?
The multiplexity of identity
Then, perhaps, it might make more sense to think of identity as not one fixed lens, but a collection of classifications? Maybe all BILLY shelves from IKEA are the same to me, or I only care whether this Option contains a value or not, ignoring what the contained value actually is.
These might be considered classifications or groupings, or they might serve as our identities in various contexts.
When I say “I identify as a mostly heterosexual cis male, a father, a husband” this might not give you what you think of as a complete identity. It does not tell you anything about me as a person. But it might be enough for you in a certain environment.
Not all identities must be groupings. I can identify the number 5 by describing it as “the number that comes after 4”2 or “the next prime number after 3”, among other things. Both are exact, and they only overlap in the fact that they describe the number 5.
As such, identity is dictated by context, and can change as it changes. While I might only care that an Option is non-empty in one function, the next function I pass it to might need to inspect the value it contains.
The contextuality of identity
Some identity is meaningful mostly through context. A key might not be unique—different keys for different locks might overlap—, but its identity is described by the lock it was designed to open, even if, incidentally, it is able to open all kinds of other locks3.
The same is true for some APIs: they were written to be used by one consumer. That they ended up all over your codebase, polluting it with leaky abstractions that make no sense, is an artifact of its identity, its use case.
This is often described as the model, the interface, and the environment. Some systems thinkers think about the model and the interface, and argue that the environment should not matter. But the model does not exist in a vaccuum, and making that assumption leads to brittle design. It is, in essence, sloppy thought.
The siren’s call of identity
As with all highly abstract, fundamental ideas it is easy to get lost in them. “Zen and the Art of Motorcycle Maintenance” stylizes this as “the high country of thought”, and the book tells a vivid story of what can go wrong if you lose your way.
Even if you are not a philosopher thinking about the fundamentals of human existence all the time, dipping in and out of that high country, taking in what you can, and trying to learn from the millenia of thought contained there, can be rewarding and actually useful, contrary to popular belief4.
Or maybe it’s just that I have a soft spot for working through opaque texts and the sense of accomplishment when you finally “get” it.
Footnotes
1. Another exmple to drive the point home: think of an collection of objects in memory. If I need to change the location of the collection because I needed to grow it in size and find a new contiguous block of memory for it, does the identity of the elements it contains change? Their location did.
2. If you want to be mathematically precise about it, you would probably say something like “the natural number that is the successor to 4”, but that would likely mean less to the majority of people you encounter in your life.
3. Locking mechanisms are weird and wonderful that way. If you dig deep enough, you might find keys that open all sorts of things.
4. My personal ventures into that high country have at times be enlightening, at times frustrating. And I’ve only ever scratched the surface.