Intentionality

Here I'm discussing a cluster of concepts/terms including (but possibly not limited to) intentionality, reference, interpretation, registration, aboutness, meaning, representation. Among those, I see intentionality as the most basic one.

(The following thoughts are somewhat strongly influenced by Brian Cantwell Smith's On the Origin of Objects.)

General goal: Find/construct a crane for intentionality.

Intentionality and related concepts (in ~philosophy)

Intentionality

Intentionality is the philosophical concept most closely related to what I care about here. It's the property of mental states/phenomena/elements (and possibly some non-mental phenomena/things) being about things (in general other than themselves), e.g. a belief in X is about X, valuing Y is about Y.

Reference

Reference is very similar to intentionality but focusing on names or externalized symbolic constructs rather than mental elements (the kind of names focused on most often are proper names).

The reference of a name can be context-dependent. If I know many people named John, me saying "John" can refer to a different John, depending on the context. There's also deixis where reference changes with the context in an especially regular way, e.g. "here", "now", "tomorrow", etc.

Moreover, there is a difference between intended [sic] (sender's) reference and inferred/interpreted (receiver's) reference. When I say "John"¹ I refer to some specific John but what John I intend to depends on the ground truth of my intention (to the extent that there is ground truth to my intention). When somebody receives my message including the phrase "John", they use the context (rest of the message, their background knowledge, etc.) to infer which John I'm talking about.²

An important distinction is sense versus reference as introduced by Gottlob Frege. The reference of a term is the thing it's referring to. Its sense is the way by which the term referring to the reference. E.g. "the morning star" and "the evening star" both refer to the same thing (Venus) but they do so via different senses.

Interpretation

Interpretation is at least a "two-place verb". Whenever one interprets, one always interprets some X as some Y. X is taken to be an instance of Y, or an evidence for Y, or something like that. Observing X or [taking X in] induces a reference to Y.

Given ground truth (what is the Y that X "really means" or something like that), an interpretation can be wrong, or rather invalid. An interpretation can also be wrong even if there is no ground truth strictly speaking but there are some criteria on what kind of interpretation is valid,³ so maybe here we want to talk about something weaker than truth, like "coherence"?

Re-interpretation happens when some X that has previously been (provisionally?) interpreted as Y is now taken interpreted as Y'. X is re-interpreted when it's "seen in a new light", or some additional facts came to awareness, altering the judgment of an appropriate interpretation (this sounds a bit too cognitive/cerebral than it actually is). It's a change of reference: $(X \to Y) \implies (X \to Y')$.⁴ So it seems like it is at least a "three-place verb", one interprets X as Y in the context Z.⁵

(Probably relevant: Heidegger's "taking-as", affordances in psychology.)

Interpretation is relevant for the processes of reinterpreting oneself (one's V-things?), being one possible way of mild self-modification (see e.g. the dual role of V). So re-interpreting one's value would come to expanding one's underestanding such that one now better knows what a given value is trying to point at or should be interpreted as.

This, however, raises the question. Why should a mind think that a value needs to be reinterpreted? Why should a mind think that its previous interpretation has been mistaken/wrong? This presumes a kind of (subjective?) judgment of value validity. The context of interpretation has changed, rendering the incumbent interpretation invalid. Reinterpretation can have ripple effects. If X is included in the context Z of interpretation of X', re-interpreting X alters Z, which may cause the need to reinterpret X'.⁶

Interpretation seems like a bigger issue, so I'm going to discuss it in a separate page.

Registration

Registration is a concept introduced by Brian Cantwell Smith (BCS) in On the Origin of Objects. I understood the book only partially on the first pass, so can't do it full justice but here's a sketch.

The "base" of the world is flux, with no point-of-view-of-the-universe God-given big-O Objective individuals. One "region" of the world can track another "region" being within its effective reach. Simple tracking roughly denotes a causally driven servoloop. Even though tracking is not necessarily symmetrical, so far there's no (proto-) subject/object(-like region) to speak of.

If, upon separation (the two regions going out of each other's effective reach), one of the regions remains coordinated/synchronized with the other region (takes on responsibility) we start talking about non-effective coupling. The region that remains synchronized despite no effective connection thus becomes the s-region. It becomes the s-region due to taking on responsibility; and consequently the other region (that the s-region is synchronized with) becomes the o-region.

This introduces the possibility of error, a kind of normativity, grounded in verifiability of the connection in case of the effective connection becoming re-established.

However, the mechanisms for maintaining said coupling need not be intrinsic/internal to the s-region. Rather, they can (and typically largely are) embodied in a lot of supportive infrastructure. The s-region's coupling-maintenance capability depends on "its own" cleverness as well as on the cleverness of the supportive infrastructure.

However, this is not registration yet. To be/[count as] registration/intentional, the s-region needs to exploit the separation to stabilize the object as an object. The s-region must play with the relation, even at the cost of a certain amount of destabilization, so as to stabilize the far end (the o-region); sediment the o-region as a discrete autonomous individual partially distinct from the s-region, locate it in the wider world.

Thus, gaps in effective connection are strictly necessary to bring an object into intentional focus.

Separation between the s-region/subject and the s-region/object introduces/is the freedom allowing the former to take the latter as an object. This makes room for acts of abstraction, detachment, stabilization.

Aboutness

"Aboutness" is a propertization of the adverb "about". It seems like whenever it's used in philosophy, it could be substituted with intentionality or reference.

Meaning

The word "meaning" is polysemous. One meaning of "meaning" is basically reference, primarily in the context of ~symbolic constructs, although things that are not symbolic constructs that can also be interpreted as meaningful (e.g. the meaning of a dream). The other meaning of "meaning" is exemplified by phrases like "meaning of life", "to find meaning", or "this means a lot to me".⁷ The latter meaning is more easily rephrasable in terms of V-terms, like "meaning of life" → "one's ultimate purpose", "meaning in life" → "something that drives you in a way you reflectively approve of".

Representation

Representations are things that stand on the "subject" side of some intentional/referential relationships. A painting of Mount Everest is a representation of Mount Everest. A mental image of a person is a representation of that person. But not all intending/referring things are representations. However, a proper name is not a representation of the name's referent. Is a thought about a person a representation of that person? Ambiguous. Representation is a specific kind of standing for something, probably a fuzzy notion where an intentional thing can be more or less representational and perhaps it's not a one-dimensional axis or one Thing but rather many axes or multiple Things. The concept of representation is provisional; it may dissolve when we better understand the different apparent facets of representation and intentionality, the concept of representation will dissolve away.

But adopting it provisionally for now, what might be characteristic features of representations that distinguish it from non-representational intenders?

• picturesque?
• composed of elements?

re presentation

#-/TODO mechinterp lol?

Representation according to BCS

According to Brian Cantwell Smith, if we project registration (a high-dimensional thing) onto the subject, we get representation; if we project it onto the object, we get ontology. I intend to read more of Cantwell Smith to understand this better.

The map-territory metaphor runs into the homunculus problem.

More generally perhaps, I'm not sure what is the utility and reason for introducing intermediaries between our thoughts (mental elements?) and what they are (supposed to be) about. These intermediaries come under the names like "proposition", "representation", and the like. It seems like the main reason for this is to translate them into a form that is closer to formal and thus easier to transmit losslessly, given that the sender and receiver share sufficient context.

What is intentionality?

It seems like a kind of relation between things but a peculiar one, in that the object being related to need not exist at all. E.g. one might love (feel like one loves?) a person who doesn't even exist, e.g. a literary character. (This seems to puzzle or have puzzled philosophers quite a bit.) The object may exist in the past but no longer (e.g. beliefs abuot one's dead grandma), in which case the element may have been acquired "directly" (grandma that died when one was 10yo) or "indirectly" (grandma died before one's birth). The object may exist in the future.

Some philosophical accounts claim that mental states whose referents seem to [be about]/[intend to] non-existing things are are [are not about]/[don't intend to] anything at all. This is too big a price to pay (in isolation at least⁸). We would like to have an account of intentionality that handles intending to existing things and intending to non-existing things in one swoop, "elegantly". In that case, the onus of intentionality is being put on the intending thing and there must be something like the term trying to point at something, even though the thing doesn't quite exist in the world. People can have coherent, very belief-like beliefs about things that don't exist, regardless of whether they know that these things don't exist (e.g. fictional characters) or believe that they do exist (e.g. ancestor spirits).

#-/HERE

The most natural-seeming-to-me way to do this (without falling for "pure" dispositionalism, intentional stance, etc) would be to distinguish between a representation of X (henceforth «X») and X.

Intuitively, there is a kind of "internal"/"disconnected" feeling/attitude that we seem to have toward X, independently of whether X exists (or, more weakly, whether our conception of X is adequate).

So suppose that a mind M is in an intentional relation R to X, the mind is in relation «R» to «X». «R» is the internalized "mirror" of R and «X» is the internalized representation of X. «R»(M,«X») is causally upstream from R(M,X). If X exists, then «R»(M,«X») causes R(M,X), i.e. a relation of e.g. M behaving lovingly towards X. «R» can manifest only contextually, be latent, elicited/elicitable by specific contexts.

«R»(M,_) can be (the internalized mirror of) a relation of "believing not to exist", which would then cause M to exhibit surprise upon encountering an actual X.

There is some kind of phenomenological/local/actual aspect of «R» that makes «R» have the same "structure", regardless of whether the object of the corresponding R is actual/real/existent or not. The local («R») part is real/actual, regardless of whether the distal (R) part is.

When we use the phrasing "M R about X", we conflate two meanings: «R»(M,«X») and R(M,X) (the latter plausibly via «X»).

Problems/caveats regarding this idea:

• «X» is a part of M. So we would need a coherent idea of how the whole of M «R»-relates to «X». Perhaps through abstraction?
• This is still very skeletal. There is a want for more meat on the bones.
• It feels like it's trying to reconcile representationalism and dispositionalism whereas I have an intuition that we should actually build a bottom-up account that deconfuses us about both in one swoop.
• Identificaiton/recognition of referentiality hinges on a mind (the same or different mind) doing registration. This feels like it's eliding the Mystery in a way similar to the intentional stance. On the other hand, it's compatible with Brian Cantwell Smith's view that (most) registration is unregistered by default.

Reference

I am interested in the mechanics of reference. Not what makes us think that one thing (e.g. a mental state or a name) refers to another thing (e.g. a state of the world), which is what academic philosophy seems to be primarily concerned with (see e.g. SEP), including teleosemantics.

Beliefs and V-things refer to something about the World.⁹ But the World is Big. Some beliefs refer to most abstract mathematical facts that don't always map neatly onto the world. Beliefs and V-things can be about other beliefs and V-things (possibly causing some weird dynamics).

Reference means normativity, possibility of error: the referans should match its referandum.¹⁰ This normativity goes in different directions for beliefs and V-things. A belief should change to match its referandum, whereas a V-thing should change its referandum to match itself.

What does it make for a referans and a referandum to match each other (sufficiently)? The match cannot be assessed by either looking at just the referans or just the referandum. Both are need to measured, at (sufficiently) the same time.¹¹ A connection needs to be established between them. Some kind of reaction is expected if they don't match and another (lack of? null?) reaction is expected if they do match.

What would I lose if I couldn't talk about reference and related stuff? I wouldn't be able to say that "I have a belief about X" or "I have no beliefs about X". In general, I wouldn't be able to communicate some kinds of normatively laden links between (aspects/states of) my mentality and the rest of the World. E.g. I have some belief about X but then I take this belief as another thing in the world and communicate by referencing this belief, rather than its referandum.

Overall, reference (and the rest) seems pretty mysterious. More specifically, its mystery has a specific kind of smell to it that may indicate that it's not a Thing, only seeming to be a Thing, an illusion of a Thing; so that once we understand how the illusion arises, there won't be anything left to explain.

Here's an attempted example of how this may happen.

Suppose that we have a hatch of ducklings and their duck mommy. The ducklings have some in-built-ish element that "hooks onto" a salient object within the first 13-16 hours and produces another element that makes the ducklings reliably follow the object (and broadly treat it as their mother). From then on, the latter "runtime developed" element can be said to refer to this object. It has possibility of error, i.e. there are "adversarial" inputs that look like the object-to-be-followed but are actually not the object. The former, "in-built-ish" element can be said to refer deicticly to the duckling's mother; it's an in-built-ish element type that is instantiated in every (normally/properly functioning) duckling. In ecologically typical circumstances, the tracker element hooks onto the duckling's mother. It can then be said to refer to the mother. In other circumstances, it may hook onto a corgi or whatever.

The in-built-ish element's reference feels less like a reference than the developed element's reference. Why is that? One possible reason is that it is more prone to failure/misreference. It's easier to fool the in-built-ish element's referentiality (just put whatever big object instead of the duckling mommy close to the ducklings) than the tracker element's referentiality (you would need a sufficiently similar object and as far as I understand, the ducks can distinguish between their mother and other similar ducks). Another reason is that its reference is deictic, with different referrents in ducklings from different hatches, albeit same sense/character.

Given that the duck mommy, the elements of the ducklings' minds, etc are Things and that they can use some pattern detection machinery to detect other Things, making it possible for them to robustly do what they should do (i.e. accomplish some specific purposes) … I don't feel the need to talk about reference anymore.

(The discussion so far has implicitly presupposed that that the ducklings, their mommy, the dog, and the elements in the duckling brains are all Things. Thingness is not given, however, it needs to be derived in some way. Thingness still feels like a mystery but given that it is possible to achieve, grounding reference in Thingness seems like a reasonable possibility.)

Is [Thingness (ontos?) + pattern detection + robust purposefulness] sufficient for reference? Maybe it's the wrong way to state this. Rather, maybe we should ask:

This is a simple example and it's probably missing many things. For one thing, I would assume ducks to have some non-trivial degree of object permanence and so they would understand that if the mommy disappears from their sight, she's still there. For a slightly more complicated example, I would expect that if the ducks were regularly immobilized, have their mom removed from the view and returned always at the same place, then they would learn to expect the mom to appear back at the same place, every time. This is a very rudimentary form of adjusting your inner state/element (the referans) so that it keeps tracking (referring to) the object (the referand). What does "keeps tracking" mean? Minimally, when it reappears (when it re-enters your effective reach, "meaningful causal lightcone"), the referans is in the same state as it would be, had the referrand stayed within the effective reach for the whole time. More ambitiously, the referrant remains in the (approximate) state in which it would be had the referrand stayed within the effective reach for the whole time.

Generally, a Thing has some purpose and in order to accomplish that purpose, it needs to make use of another Thing, and so it needs to be somewhat coupled to that other Thing's state, even if it's not within its effective reach.

This doesn't seem to satisfiably account for cases like humans deriving latent information from distant/noisy signals for which they have no practical use. Suppose the knowledge of WASP-17b is completely useless. I still want to talk about my beliefs about the planet as tracking the planet or missing it. (One possible patch is to use inter-subjective cross-verification but it doesn't sound convincing.)

Teleosemantics?

From Abram's post on teleosemantics:

Ok, but what does "accurately reflect" mean? Teleosemantics (or at least this way of stating teleosemantics) seems to define "reference" in terms of another mysterios/load-bearing concept "reflection" which seems to be locally synonymous with "representation".

I'm also not sure about what makes something a symbolic construct and whether we should count mental elements as symbolic constructs.

Further down in the post:

This still doesn't address what the correspondence in question actually is about.