NB: Originally posted to Map and Territory on Medium, so some of the internal series links go there.

NB: I made some mistakes in this post about my use of academic philosophy terms in ways that I would no longer endorse. I've let it stand for historical purposes, but of particular import is to note that "existential phenomenology" fails to capture all of my perspective well, and my understanding of idealism in this post is rather poor, since my perspective is well described, in part, by idealism, specifically phenomenalism.

My introduction to axiological alignment is just that — an introduction. Out of necessity it covers the philosophical foundations of my thinking on AI alignment, but it doesn’t explore those foundations in much detail. I could just keep writing about axiological (now noematological) alignment and AI, but I suspect I would lose everyone, so first I’m going to go back and explicate the background theory as thoroughly as seems necessary. We’ll begin at the beginning, employ examples and metaphors, and explore, in order, the foundations of phenomenology, phenomenological methods, the phenomenological reduction, feedback, qualia, and finally noematology.

I think of phenomenology as the philosophy of the adept beginner. By this I mean it is the philosophy of the person skilled in all the techniques of philosophy from logic to observation to intuition who nevertheless chooses to approach philosophy as if they were a beginner so that, rather than trying to explain anything, they rest on the questions until they answer themselves. Phenomenologists are not the first to try this — Socrates, Zhuangzi, Nagarjuna, Descartes, and Hegel are among our predecessors — but we have the advantage of standing on their shoulders. Let’s see how far we can see!

Consider the fundamental question of philosophy, “why?”.

“Okay, why?”
“Why” what?
“Why is anything?”
What is this “anything” you refer to?
“The world. Why does the world exist?”
Does it? How do you know?
“I’m in it!”
Oh, so why are you in the world?
“I…I don’t know. I just am.”

It’s from not knowing why “I am” that we start because if we follow any line of questioning far enough we’ll eventually need to know what this “I” is to give a complete answer. Even if any particular answer doesn’t seem to depend on “I”, answers are given and understood by an “I”, and we can always ask of any answer “Why do I think that?”, so we have to address “I” sooner or later. Of course, we also have to deal with the whole world sooner or later, so choosing to start with “I” is also a pragmatic choice because it’s something we are each quite familiar with and have privileged knowledge of, and as we’ll see later there is something epistemically irreducible about “I” that makes it interesting.

Now depending on your background and education, you might be tempted to dismiss this idea of starting with “I” and jump straight to taking an objective approach. I understand the appeal, but remember we are working from a beginner’s perspective, and even the logic of science must begin with the subjective. If we don’t actually need “I” and it evaporates as we develop our understanding, then so be it, but beginning as naifs we must include the naive conception of “I” for now and only remove it later if we can fully account for it.

Continuing our line of questioning,

How do you know you are? Don’t think too hard; just give the obvious answer.
“I see and hear. I think. I remember. I experience.”

If we take up the kind of the radical philosophical skepticism introduced by the Ajnana school and Pyrrho, one of the few things we can claim to know is that we experience. It may not be clear who we are, what we experience, or how we experience it, but to know is to put the “I” in relation to the world and we call that relationship “experience”. Existential phenomenology is the philosophy that develops if you choose to assume experience is the only source of knowledge.

There are other options. Platonism, for example, chooses differently and permits the existence of direct knowledge without the need for it to be experienced. Some strains of Gnostic and Buddhist thought suppose there is only direct knowledge and experience is an illusion. And solipsism rejects nearly everything we think of as knowledge, granting only the existence of the “I”. All of these are possible ways to address “why am I?”, so why phenomenology?

Two reasons. First, modern physics makes it abundantly clear that there is probably no direct knowledge. We get no faster than light travel, no true spooky action at a distance, no nothing! We always pay for our lunch, even when it’s free, and if there is direct knowledge, then we’ve gone a suspiciously far way to understanding the physics of our universe without discovering it. At the same time, what physics does show us is that information only exists when it’s moving, which is to say when measurement happens, and thus appears tightly bound to the mechanisms of causality. So taken in whole, physics paints the picture of a world where the only way to know anything is through experience.

Second, parsimony. Reasonable people may disagree, but my take is that the phenomenological perspective — that experience is the only source of knowledge — is the least complex solution that is able to fully address the question of how we know of our own existence. If we try to get away with less complexity and say experience is not necessary, I think we fail to adequately explain why it appears to us that we have a shared, objective existence that extends beyond our own knowledge. And if we try to demand more complexity by, for example, assuming the existence of direct knowledge, I think we get a gear that doesn’t turn the machine and so can leave it out of our understanding without losing anything. Thus, by granting experience sole ownership of knowledge generation, we produce a “just enough” explanation that neither has anything missing nor anything extra.

I’ll have more to say on parsimony when we look closer at the phenomenological reduction, so for now let’s wrap up our dialogue.

So you experience, but what if you take away the “you”? Can there be unattached experience?
“…no? …no. No, there can’t.”

Experience is always experience of something, by something. That is, it is always that some subject experiences some object. In this way we say that experience is intentional because experience is directed and exists in tension between the subject and object the same way a rope may be held in tension between two posts. If we take away the subject or the object the experience falls away the same way the rope would go slack if we took away one of the posts, and if we take away the experience then there’s nothing connecting the subject and object so they would just be things, not subjects or objects, just as our two posts would be disconnected without the rope stretched between them. Taken together we call this relation of subject, experience, and object a phenomenon.

Suppose we take a specific phenomenon, like “dog barks at car”, and try to show it is not intentional by separating out its parts. Let’s start with the subject. If we take out the dog we have only “barks at car”, but even without the dog we must imagine something is doing the barking. This gives us with a pattern that can be matched by a subject, but then we are still describing the experience of something barking at a car, so we still have a complete phenomenon, albeit one with a thought as the subject rather than a thing. As we’ll see when we address qualia, the subject being a thought poses no problem because the subject, when experienced as the subject of a phenomenon, is always a thought anyway.

If we can’t fully remove the subject, what about the object? Taking out the car nominally gives us “dog barks” and, at least in English, this is a valid construction where the experience/verb has no object. But is there really no object? Consider what barking is: the dog’s bark happens when the dog uses its body to create compression waves in the air that we identify as barking, so “barks” has an implicit object like the air or the world into which the dog does its barking. Thus we can’t really remove the object either, although we can change our frame of reference to see the object differently.

Since we can’t remove the subject and we can’t remove the object, we obviously can’t remove both and get a pure experience of barking, but what if we remove the experience and attempt to hold the subject and object as things-in-themselves? This almost seems possible, but ask yourself how you know about the dog and the car. Since we assumed that experience is the only source of knowledge, it must be that something experiences the dog and car to know about them, and in this case that thing is you! Maybe the dog and car can exist without being known, but in existential phenomenology their existence is fundamentally unknowable if they are not experienced by a subject. So even if the dog is not related to the car by the dog’s barking, the dog and the car are objects of your own experience and thus part of some phenomena. If they were not they would be literally unknown.

Thus we see that phenomena, although they have parts, are not divisible, and all knowledge exists as phenomena. This is the keystone of the phenomenological perspective.

I see. You were secretly a phenomenologist all along!

Don’t be too surprised if you find yourself nodding along and thinking this all sounds obvious. It is! The trouble comes with the philosophical bullets you’ll be asked to bite when taking this radically naive view.

Consider the question of whether a thing exists in its own right independent of phenomena. That is, do things exist if they are not experienced? For example, suppose there is a star so far away that the light from it will never reach us. Since it is outside our light cone we will never experience evidence of it, so in what sense can this hypothetical star be said to exist? From the phenomenological perspective, we might say that the idea of such a star exists but that the star itself does not since we can’t interact with it.

Or consider particle physics. In what sense are quarks or strings or anything else “real”? Scientific realism holds that physical models, to the extent they are accurate, describe things as they really are — viz. if atomic theory correctly describes the world then the world really is made up of actual atoms. But this is to suppose a direct knowledge of the world which we can uncover through scientific inquiry. A phenomenologist might instead say we can investigate the ontic being of a thing, but only because we have knowledge of the ontological and through that may infer something of the metaphysical. This leaves phenomenology compatible with physicalism, but finds it decidedly opposed to realism and, in its existential form, idealism.

If none of that sounded too weird, when we get around to discussing qualia and consciousness we’ll find that existential phenomenology implies something like functionalism and definitely implies panpsychism. I won’t try to convince you of that now, but know that our humble choice to accept that all knowledge comes from experience will lead us to places far off the beaten path. It should be an exciting ride!


1 comments, sorted by Click to highlight new comments since: Today at 4:32 AM
New Comment

I've started reading the first few posts in this sequence and I'm very appreciative of your attention to detail concerning the origins of various terms, how others have used them, and how you are using them.

New to LessWrong?