Nodes in a network

Metaphysics & Epistemology
1

bongo fury
When I made the point (badly) I nearly said “nodes in a network”. Dang!

Jamal
I feel like I’ve been overusing it lately.

Once more can’t hurt?

Compare: Quine’s web of belief. The nodes are scientific sentences, the web is scientific theory as a whole. Not just any. The web isn’t just idle speculation, but active assertion about the world. So hypotheses effectively debunked are nodes erased.

Neither is it a big formal theory, whose assumptions and dependencies are clearly underlined. Were it so, verification could be (at least on occasion) conclusive piece by piece, but Quine argues that holism is inescapable. So we may acknowledge that some nodes are, either singly or in concert, in conflict with others. Just that we shouldn’t expect a clear outcome (with or without erasures), even in principle.

I’ll stand corrected on this crude caricature, of course.

Now consider, the web imagined by some of us when using a google search. What would be so wrong with allowing ai to survey the documents provided by the search algorithm and summarise them? Wouldn’t that be a reasonable stab at visualising the relevant part of the web of belief?

(And surely the recommended holism accommodates treating empirical research as part of the web of belief even when not strictly scientific; and I’ve no objection to that.)

What I have objected to is respecting the ai summary on a par with a humanly authored one. But isn’t this concern with authorship irrelevant in the Quine analogy? The sentences in that network speak for themselves, as it were.

Yes, and this is (for some of us) an advantage of Quine’s metaphor. It cuts out the psychological middle-men (ideas, concepts, attitudes). But as such, it exposes all the better the question how the web of belief-language is able to relate to its subject-matter. The answer to which is of course the testing of it against reality (the “tribunal of experience”). Holistically, yes, but in labs, or in the field. Whereas, use of a google search is in order to navigate to nodes that are secure and supported in their positions in the network, owing to previous efforts of verification and validation.

In Quine’s more rarefied conception, assuming a narrower sense of science, the links between nodes are implications and suggestions between sentences. In a looser scenario that accommodates google searches, we can imagine the links will often be signals of justification and endorsement (or otherwise) between larger texts. Just like with any library search: follow this citation to that source, from there to another, etc.

But any selection involved in this process (so all of the process) must presume to adjust the relation between belief-web and world. Hence the traditional importance of authorship at every node. Citing a source provides only provisional validation, conditional on the author’s accountability. And a library search is just as much an act of authorship for which you make yourself accountable, if you publish it. It’s a new node, with an author.

Well, a google search isn’t a new node, despite all the bias in the selection of items, because you know you asked for any texts containing a certain string, with no further criteria. So not all, nor a representative sample, of such texts. You will do the due diligence.

Whereas an “AI summary” is apparently embraced by most as a genuine act of intelligence and erudition, but with no acknowledgement of the dilemma of a missing and hence unaccountable author.

Which is why I draw the line at that particular technology, and can’t understand why capitalism doesn’t want to help me avoid it. :joy:

So, there’s my attempt to flesh out “nodes in a network”.

I wondered if it connects with the thesis lately mooted regarding the status of “ideas floating free of their expression”? (Citation pending. Was Jamal and Foolosophy.)

2

Consider four types of ideas that might be said to “float free of their expression”:

  1. The “idea” that a poem communicates.

  2. The “idea” that a complex mathematical calculation communicates.

  3. The “idea” that a philosophical argument communicates.

  4. The “idea” that an AI summary communicates.

I’m guessing that most of us would have very decided responses about which of these ideas can be said to be independent of their expression. So the question needs a good bit of analysis. But I like your speculation about an AI summary as, possibly, a snapshot of a particular web of belief. Are we sure the author is missing, though? Is “invisible” closer to the mark?

3

No! I meant, unidentified.

Yes! … Or, “unidentified”? I suppose identification might not be impossible. But a kind of identification that facilitated accountability? :thinking:

4

Yes. But couldn’t we (following Nelson Goodman, I would think) characterise instead a class (narrower or wider in the different cases) of acceptable paraphrases of the text?

5

“Unidentified” is best. Maybe unidentifiable too, in the way that an author of a manuscript from hundreds of years ago is unidentifiable – but still, there was such a person.

It’s been a long time since I read Goodman. Is the suggestion that, if a paraphrase is acceptable, then we’ve identified a synonymy of “ideas”?

6

I guess not! Synonymy being too problematic a notion for Goodman, as for Quine.

Ok. I agree, any of those 4 kinds of text might seem to communicate an “idea” also captured, easily or awkwardly, by some paraphrase. Where easily, perhaps we would say the idea is independent of expression?

But we could at least be ready to redeem (more or less similar) “ideas” as (more or less similar) “paraphrases”?

We don’t have to consider the texts as literally conveying any abstract items, even though the metaphor seems natural enough.

7

Yes, possibly, or we might decide that there are cases where no paraphrase is possible within the discourse. For example, I can perhaps paraphrase what a poem means, but in doing so I move out of poetry and into something else – lit-crit prose, maybe. Arguably there’s no poetic paraphrase of a poem, because it would also have to paraphrase the “really poetic”, non-discursive qualities of the original poem. So, if this is a case where the poem’s idea (taken as including the poetic) cannot “float free of expression,” it’s so not because paraphrase is difficult, but because it’s impossible.

I think those who feel this way about philosophical prose have something similar in mind: The language we choose to express our ideas must always fall short of something paraphrasable within the discourse of philosophy. Is this true? I’d say: Sometimes, but not always, depending upon philosophical commitments and style.

Right. But what about a mathematical “text”? If it isn’t conveying an abstract item (such as the solution of an equation), what in the world is it doing? This is the other end of the paraphrase spectrum: Math doesn’t need paraphrasing because anything that “says the same thing” within the discourse of mathematics simply is the same thing.

(Though hmmm, what about different ways to solve an equation? I invite mathematicians to weigh in. Does it make sense to call one a “paraphrase” of the other?)

8

Speaking for myself, no, it’s not about degree of approximation to an idea, which is impossible to measure, because chimerical anyway.

For me, it’s about comparing bits of language that are identifiable, e.g. texts, and using historical facts (or claims) of authorship to guide comparison. As well as using, of course, experiments in paraphrase. Assuming the reality and even primacy of “ideas” is liable to obscure the epistemological utility of being able to actually identify bits of language?

9

Well, I’d say most of my teaching is paraphrase. Analogy.