Form and Feedback in Phenomenology

Link post

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

Now that we’ve covered phenomenology’s foundations and its methods, we’re almost ready to begin addressing AI alignment from a phenomenological perspective, but before we can start talking about AI or how to align it we need to build a clear understanding of what AI is and what aligning it would mean. To do this we need to understand noematology, which requires understanding consciousness and qualia, which requires understanding feedback, which requires understanding things and time. That’s a lot to tackle at once, so we’ll split the inferential chain and start where many philosophers have started — consciousness. Specifically, we’ll begin with a phenomenological reduction of self experience.

To start our reduction, let’s first bracket the experience of self as a phenomenon in a clear way that makes precise what we consider its subject, experience, and object. English grammar encourages us to phrase this phenomenon as “I experience myself”, but in normalized form we instead denote it {I, experience, I} to mean that the subject, I, has some experience of itself. This immediately invites many questions. What is the I? How do we know it’s both the subject and the object of this phenomenon? Does the I know it’s experiencing itself or is it only experiencing itself without knowing what it’s experiencing? We’ll need to address all of them, but since they all ask something about the I, we’ll begin by bracketing it and exploring its role as intentional subject and object.

Things

It’s tempting to start by asking what kind of thing I is, but this immediately exposes an assumption we need to suspend — that I is a thing. Epoche of this idea is often a jumping off point for considering the essential nature of (or rather the lack of essence of) the self, and that’s a topic we’ll cover, but to address the I now would be to put the cart before the horse because doing so assumes we have a solid understanding of what a thing is! That may sound silly if you aren’t used to philosophical discourse, but, as we’ll see, understanding just what we mean when we talk about a thing is the cornerstone of complex phenomenological analysis.

It’s a little hard to talk about what a thing is, though, because we naively think of things as ontologically basic. Look in a dictionary and you’ll find most definitions of “thing” are circular: they define things as objects, items, entities, articles, or artifacts. The best I could find is the circumlocution “that which can be described”, but this still leaves us with the trouble of how to specify what “that” is. Even the etymology of “thing”, while hermeneutically interesting, doesn’t much help us figure out what we mean when we talk about things. The epoche of “thing” seems to be our first real phenomenological challenge.

Rather than deal with things in the abstract, let’s begin by reducing something more concrete. Most people would agree that a cup is a thing. Why? Well, they might say, a cup is distinct from its surroundings. For example, you can pick a cup up off a table and the cup remains a cup whether it’s on the table on in your hand. But to say that the cup remains a cup is to suppose we already know what a cup is, so being a bit more skeptically cautious we step back and say only that we observe a cup on the table and continue to observe a cup when we pick it up and hold it in our hands. But even this is not cautious enough because we’ve still assumed that we can identify the cup, so we must also bracket the process of identification. Only then are we finally left with a question we can begin to answer: how do we see a cup from the pure sense data of experience?

One specific answer we might give is that our brains detect patterns based on the light that reflects off the cup and use those patterns to form a category. Then through a complex process of learning that we don’t fully understand, we develop an association between those visual patterns and the sound of someone saying “cup”, and then also the squiggles we use to write “cup”, so that we come to have a grouping of patterns we assign the label “cup”. This gives us a way of forming mental categories both for specific cups and cups in general, and by a similar process categories for other patterns we find in our experiences. Combined these categories create a logical structure we apply to the world as we observe it that we call ontology.

As with cups, so too with all things, thus we can say that things are ontological categories. But ontology does not seem to be a full accounting of thingness because it depends on the phenomenological subject. That is, ontology can only identify a thing via phenomena and gives the thing no existence independent of our experience of it, yet cups seem to continue to exist even when they are locked away in cupboards. Thus our naive sense of thingness seems to extend beyond ontological thingness, so it appears there must be some way in which things exist independent of conscious experiences of them.

Indeed, many of my philosophical forebearers agree. Kant called this sense in which things exist on their own the things-in-themselves or noumena as apart from phenomena, Heidegger referred to it as ontic being as opposed to ontological being, and Sartre distinguished between things and things as things. Each, I think, is looking for a way to talk about the natural organization of stuff that we observe as patterned prior to knowing that it’s patterned, especially since it seems that our individual, independently constructed ontological categories are highly correlated with each other in a way that suggests there is an external reality that exists independent of our knowledge of it. In other words, they are trying to address the intersubjectivity of thingness.

My own take on the metaphysics of the ontic nature of things, for what it’s worth, is as follows. The world is made up of stuff. That stuff might be particles, strings, fields, part of the universal wave function, or something else, but whatever it is this stuff interacts with other stuff and phenomenologically we call this interaction “experience”. Further, stuff experiences other stuff differentially, so stuff metaphorically clumps into clusters of stuff that have more, stronger experiences between the stuff in the cluster than the stuff outside the cluster. I call such clusters ontic things, though note that we only know ontic things as ontological things through our experience of them, so we necessarily lack perfect knowledge of the ontic given our assumption that our only source of knowledge is experience. Nonetheless we can often be confident enough in our assessment of the ontic and ontological to act as if ontological categories point to ontic clusters of stuff and as if both describe the same naively conceived thing. Taken together the ontic and ontological give us a sense of the forms of the world even though it is properly formless.

Using this sense of what a thing is — an ontological category describing an ontic clustering of stuff in the world — we can begin epistrophe of our earlier epoche of I as thing. To say that the I is a thing is to say that we observe some pattern in the world that we identify as being the source of experience and give this thing the name “I”. Thus when the I experiences itself, when {I, experience, I}, the I is experiencing itself as a thing it identifies as I. This addresses the first of our questions— “what is the I?” — but what about the second? How do we know that this same I is both the subject and the object of the phenomenon?

Time

Let’s suppose, for the sake of argument, that when we talk about the phenomenon {I, experience, I} we intentionally mean to say by using the same lexicographical term, “I”, for subject and object that the subject and object are exactly the same thing, viz. the I is experiencing exactly itself without any differences. This would imply that self experience cannot change the I since the I as subject cannot be modified by experiencing the I as object else the I would not be experiencing exactly itself, but this seems at odds with the evidence from our own lives where self experience not only can change us but is often a leading force for change, so the subject and object must not be exactly the same. What is their relationship, then, that we are led to think of self experience as {I, experience, I}?

A first approximation might be to say that whatever it is that we identify as the I is not the entirety of I. That is, perhaps there is some “core” part of the I that persists across experiences associated with some “outer” part of the I that changes so that [I, experience, I] is really shorthand for [core-I, experience, core-I and outer-I] and this phenomenon produces a modified outer-I. Such a core-I is essentially a soul, and it does a better job of fitting the evidence that the self changes even as it seems to remain connected to its past than supposing that the entirety of the I never changes, but to me it seems insufficiently parsimonious to suppose that the I has two parts, especially since this theory would suggest that every thing that changes, not just the I, would have a “soul” of some kind.

We could instead take the I to consist only of the outer, mutable part and for the core, apparently immutable part to be a pattern within the outer-I that remains stable over time. Then when we say that {I, experience, I} what we mean is something more like {I’, experience, I} where I’ is I with a a specific experience of itself added in. I, I’, and their successors may experience their precursors many times forming a long chain of things stretching into the past that we associate with I, so writing {I, experience, I} is a handy way to elide the many phenomena and iterations of the I that we are actually considering. This theory has fewer moving parts than supposing the I has a soul and does not imply pansoulism, so it seems a decent way of accounting for the identification of the object with the subject in self experience.

Only, notice that above we had to use a concept of time, both explicitly and implicitly, to give our description. Even when we supposed there was a soul we needed time to address issues with the outer self. If the I were immutable this would not be an issue, but to talk of precursor and successor Is is to talk about the past and the future. So if we are to accept that the I can change, we must have an understanding of what that change is over or through. After all, to persist is literally to stand through, and if time is to be that thing through which the self stands, we’ll want to consider what time is.

Husserl and Heidegger both gave considerable thought to the experience of time and found it to be not a thing but an aspect of intentionality. That is, time is not a dimension as it was often conceived of in Newtonian physics, but rather a consequence of the intentional nature of experience. That we can think of time as an objective medium or external reference frame through which we move is an ontological convenience we construct to make it easier to consider time as a phenomenological object, but in fact time is epiphenomenal and it’s only through phenomena that we come to experience the world as if it had time. This makes phenomenological theories of time examples of causal theories of time where causality creates a pattern we call time rather than metaphysically basic time creating causality, which is nice because modern physics seems to agree that we need a causal theory of time.

Understanding time as an effect rather than a cause has dramatic ramifications for our understanding of things. To talk as if a thing existed in the past or will exist in the future is really to say that phenomena contain patterns identified as a thing and that this pattern suggests it was also present in past phenomena. For that matter, the notions of “past phenomena” and “future phenomena” are themselves patterns in present experience that suggest the existence of phenomena that created the world as it is presently experienced and suggest the existence of phenomena that will be created based on the world as it now seems. Consequently I tend to think of time as an ordering relation that forms a poset over the set of phenomena, but we need not commit ourselves to any particular interpretation so long as we understand that things lack permanence and any attempt to think of things as having a persistent essence — even if we understand it as an observed pattern rather than an ideal form — is fundamentally unsupported by our assumption of the primacy of intentionality.

That being said, we can still talk about things as things because our naive idea of thingness is not useless, only fuzzy. So long as we remain clear that our demarcation of things is an ephemeral choice, albeit an ephemeral choice based on our observations of reality, we may talk about things and their existence over time. Thus, to return to our reduction of self experience, we can say that the same I is both the subject and object of the phenomenon of self experience so far as we understand the I to be the same thing before and after each particular experience. That the I may change so much over enough such experiences that we would no longer recognize the I as the same I given a long enough period of time is a matter of our choices about when a thing stops being itself and starts being something new — the same choices we must always make in building ontology.

Feedback

We now have only one question remaining to address with our reduction: is the I experiencing itself as itself or is it only experiencing itself? That is, when we say that {I, experience, I} do we mean that I experiences itself as an ontological thing or do we mean only that the I experiences itself as an ontic thing without subjective identification of the object as itself? Our experience of consciousness suggests that we mean the former, but understanding clearly its relationship to the latter will, I think, help pull together the concepts we’ve just discussed and prepare us for a deeper discussion of consciousness, qualia, and noematology.

To avoid confusion with consciousness, let’s begin by considering the self experience of something generally not considered conscious — a steam engine with a governor. If you’re not familiar with steam engines, the operation of one with a governor is straightforward: the boiler heats water to produce steam, the steam applies pressure to rotate the turbine, the turbine turns the drive shaft, part of the drive shaft’s mechanical energy is used to spin the governor, and the governor is connected to a throttle valve controlling the amount of steam that reaches the turbine. When the engine starts the throttle is open, so steam pressure increases, turning the turbine faster, spinning the governor faster, closing the throttle a little. This leads to decreasing steam pressure, turning the turbine slower, spinning the governor slower, opening the throttle a little, which causes increasing steam pressure, turning the turbine faster, spinning the governor faster, closing the throttle a little, and on and on until the engine runs out of heat or water, thus making sure the steam engine’s drive shaft always turns at a nearly constant speed. We call this a homeostatic or negative feedback process.

What’s phenomenologically interesting about this steam engine is that it has no idea what it’s doing. As far as we can tell it doesn’t know what it is or how it works, i.e. it has no ontological self awareness, nor does it encode the idea that the drive shaft should turn at a particular speed, yet because the steam engine exists as an ontic thing we can say that the steam engine experiences itself because the governor gives the engine a way to experience past experiences of its own ontic existence giving the steam engine the appearance of telos. The phenomena {slower steam engine, if too fast, steam engine} and {faster steam engine, if too slow, steam engine} arise because the governor connects the output of the steam engine to its input, and in doing so creates an ontic thing that, through our ontological lens, appears to persist, have self experience, and regulate its own action over time. This makes the governor less than conscious but more than inert: it makes it cybernetic.

The term “cybernetic” has been abused a lot, so to be clear when I say the governor is cybernetic I precisely mean that it’s an (ontic) thing that experiences its (ontic) self. Unfortunately I have to heap my own abuse onto “cybernetic” because, as we’ve defined it, everything worthy of thingness must be cybernetic. Consider a seemingly inert, non-cybernetic thing, like a rock. Rocks appear entirely inactive, yet rocks manage to stay together and be things despite the universal tendency towards entropy. The stuff of a rock is constantly interacting to maintain itself rather than merge into a soup of its surroundings, and so it is that rocks, after our understanding, must be experiencing themselves, and so must be cybernetic, and similarly all things, to the extent that they are things and differentiate themselves within the world, are cybernetic.

Cybernetics is still useful as a concept though because there is a sense in which a steam engine is more cybernetic than a rock. The self experience of the steam engine generates more thermodynamic entropy (in the form of exhaust heat) than the self experience of the rock (in the form of diffusion), and so the steam engine’s self experience can, though doesn’t necessarily have to, contain more information than the self experience of the rock. Indeed, the rock is perhaps a more efficient self experiencer than the steam engine since the rock’s self experience makes almost no information by also generating almost no entropy, but we are more interested in the steam engine because it produces a whole bit of information — whether to open or close the throttle — by creating lots of heat and mechanical energy. And cybernetic things only get more interesting as they produce more information.

Computers, for example, produce so much information that they create things out of information. I’m typing this on a computer, and as a result the computer is creating a thing we call a document that consists purely of information within the computer system. Indeed, this document has no physical representation of its own: it exists purely within the self experience of the computer due to the complex interactions of its many parts, and there’s no way to separate the document from the computer other than to produce another document in another computer or to print a representation of it on paper. And if cybernetic things can contain things within the information of their self experience, what happens if these information things start to experience themselves?

Well, for one it means that information things, like the document, have ontological existences separate from their ontic existences — you know, the map is not the territory — because we can create many instances of the “same” document yet they will all be made up of different information being produced by different computers. It also means that ontological things are manifested as ontic things, so ontological things can also experience themselves via the ontic, which is to say they are embodied, and when ontological things experience themselves we say that the ontic things they exist in are phenomenally conscious rather than merely cybernetic because they both experience themselves and experience themselves as themselves, as ontological things. Thus the I of self experience both experiences itself and experiences itself as itself, answering our final question.

At last our reduction is, for now, complete. Through it we’ve seen that:

  • Things exist ontologically as patterns within our experience of them.

  • Things exist ontically as clusters of stuff within the world.

  • Things exist ephemerally though chains of experiences creating the perception of time.

  • Through ephemeral existence over time, things can feed back experiences of themselves to themselves, making them cybernetic, and in so doing create information.

  • Things can exist within information, those things can experience themselves, and it’s from those information things that ontology, and thus consciousness, arises.

Next time, in our final installment before getting back to AI alignment, we’ll tackle questions about consciousness, qualia, and how the contents of qualia combine. You can read it here.