Overview

[sthippo]

from @rick333:

My thoughts on "awok Overview" are as follows:

1. I understand what it is to delineate the logical structure of a document. This is what happens when a seminar on Plato attempts to map the argument by laying out its premises and inferences after the teacher asks, "What is Plato's argument on page 317?" This reminds me of diagramming sentences, except that instead of various sentence parts such as subject, adjective, compound direct object, etc., we really only have premises, inferences and conclusions, although the inferences can be logical or probabilistic. A typical probabilistic structure could be: "The truth of P makes the truth of Q more likely, unless R is true, in which case the truth of P makes the truth of Q less likely." This seems equivalent to "The truth of P and not-R make the truth of Q more likely, but the truth of P and R make the truth of Q less likely." These claims could be phrased in terms of justification: "If I believe that Q is 30% likely to be true, and then I learn that P and not-R are both true, then I am justified in bumping up my confidence in the truth of Q to 70%," although admittedly in that philosophy seminar on Plato there was no talk of probabilistic inference, and I am not sure how we would assign these probabilities. Perhaps the probabilities arise from the worrying holistic point associated with Quine that justification depends to some degree on millions of other claims, so that in 30% of the possible worlds represented by my beliefs before I learn P and not-R, Q is true, but in 70% of the possible worlds represented by my current beliefs plus P and not-R, Q is true.
2. Perhaps the next step for us is to create an inventory or bestiary of logical structures. That way we would have a library of structures to draw from. This is probably what you mean by "define abstract models of logical structures."
3. As I am reading your list of eight objectives, I am reminded of an article I read recently about a piece of software named Aladdin, which is maintained by 2300 tech employees at BlackRock. The firm rents out the use of Aladdin to 40 clients while also using Aladdin to guide their own investing on behalf of clients. I got the sense that Aladdin contains a model of the universe of financial instruments so that it can predict, for example, what is likely to happen to a complex portfolio if the Fed were to lower interest rates. It also presumably knows all the correlations among the prices of various financial instruments to that it can detect covert erosions in a portfolio's true degree of diversification. What seems remarkably similar in my mind are (1) the logical claim that "If interest rates go down, then this portfolio will increase in value," which has the logical structure P implies Q, and (2) the causal claim, composed of the same words, which asserts that P causes Q, and (3) the justificatory claim that belief in P justifies a belief in Q. One of your objectives is to "explore what happens when assumptions are changed," and Aladdin is certainly doing a lot of that while functioning as a local web of knowledge covering only the financial realm, consisting of myriad logical connections, causal connections, correlational connections and justification connections among propositions.
4. I am a little worried that my complete lack of knowledge of programming is going to hinder my ability to contribute to this project. I am also suspecting that you and I need some dedicated time together in a room with a big whiteboard, so that we can start to map out the components of the overall project and decide which components need our attention at the outset. An additional worry is that if BlackRock has 2300 people doing nothing but helping its network of 5000 computers to model the financial world, then it is going to be difficult for the two of us to model the world as a whole -- which is what a web of all knowledge would do.
5. There is a lot to recommend forging ahead on our own, because it is better to think things through for oneselves than to get lost in the weeds of one or another published literature. On the other hand, I wonder whether, if we had a map of this project, it would turn out that 13 of the 37 component challenges have been solved by other people in some literature or other.

[sthippo]

Excellent comments. Responses:

1. I think your description is exactly right.
2. A Bestiary! And here I was thinking of a mere classification. By all means, a bestiary. That deserves a thread of its own. Here is a start (just the question, no answers yet): https://github.com/Rethinkers/awok/issues/2
3. I think your example illustrates very well the difference between a semantic model and a logical model. Think of a spreadsheet. Typically every row in a spreadsheet except for a header and perhaps a total holds the same exact formulas, but the data is different. The logical model of a web of knowledge is more like the formulas in a spreadsheet, whereas the semantic content is like the data. It can take 2300 people to fill column A in all the rows and at the same time just a handful of people to come up with the formulas and fix them when they break.
4. For the reasons you lay out, I think a modest goal makes sense -- a web of a small amount of knowledge, rather than all knowledge. I will be very happy to achieve that.
5. I am all in favor of finding ideas in the published literature, and also of forging ahead. Each informs the other.