![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
guslI feel a strong desire to develop systems to help me reason and communicate, and I am frequently reminded of the need for this kind of software. For example, whenever I have difficulty communicating an abstract idea with someone.
Suppose you want to communicate a simple argument (e.g. a scientific or philosophical deduction) to someone.
The medium of Natural Language is the standard way of communicating arguments, and most of the time it succeeds with only a few words (with the intended audience, at least).
Yet, there are many advantages of communicating in logical form (one of which is that computers can understand you then). But because the context / background knowledge may be quite large, it can be very hard to translate a NL expression into a logical form: first we would need to have a formal theory of the domain!
In the domains of scientific theories, this problem is solvable by doing logical formalizations of the domain, preferably multiple parallel formalizations, to reflect the multiple ways people can think about things, and the multiple ways in which they argue (or "prove") things.
There exist people doing work on argumentation systems, common sense reasoning, multimodal reasoning and on formal theories of science, but I don't know of anyone integrating these ideas towards creating a medium for theorizing. I dream of a medium where people could communicate their theories naturally and yet formally.
I want to speak a language that everyone understands (or can come to understand by following a well-defined sequence of steps); once this exists, there will be little room left for rational disagreement (and when there is any, it will be clear exactly *where* they disagree). Philosophy books will be much shorter.
I sometimes find this inexpressibility of natural language very frustrating and this is part of my motivation to be interested in such systems. I am also quite excited about the possibility of interacting with computers. Once someone formalizes a scientific theory meaningfully, it can be privately tutored to millions of people. The computer would help students play with the theories, using things such as example generation and diagram generation to illustrate principles (theorems); it could create homework problems and correct them, etc. and at a more advanced stage it could help the theorist concretize his intuitions about relations with other theories (e.g. analogy) or the meta-theory.
Suppose you want to communicate a simple argument (e.g. a scientific or philosophical deduction) to someone.
The medium of Natural Language is the standard way of communicating arguments, and most of the time it succeeds with only a few words (with the intended audience, at least).
Yet, there are many advantages of communicating in logical form (one of which is that computers can understand you then). But because the context / background knowledge may be quite large, it can be very hard to translate a NL expression into a logical form: first we would need to have a formal theory of the domain!
In the domains of scientific theories, this problem is solvable by doing logical formalizations of the domain, preferably multiple parallel formalizations, to reflect the multiple ways people can think about things, and the multiple ways in which they argue (or "prove") things.
There exist people doing work on argumentation systems, common sense reasoning, multimodal reasoning and on formal theories of science, but I don't know of anyone integrating these ideas towards creating a medium for theorizing. I dream of a medium where people could communicate their theories naturally and yet formally.
I want to speak a language that everyone understands (or can come to understand by following a well-defined sequence of steps); once this exists, there will be little room left for rational disagreement (and when there is any, it will be clear exactly *where* they disagree). Philosophy books will be much shorter.
I sometimes find this inexpressibility of natural language very frustrating and this is part of my motivation to be interested in such systems. I am also quite excited about the possibility of interacting with computers. Once someone formalizes a scientific theory meaningfully, it can be privately tutored to millions of people. The computer would help students play with the theories, using things such as example generation and diagram generation to illustrate principles (theorems); it could create homework problems and correct them, etc. and at a more advanced stage it could help the theorist concretize his intuitions about relations with other theories (e.g. analogy) or the meta-theory.
![[identity profile]](https://www.dreamwidth.org/img/silk/identity/openid.png){kind=small}
(no subject)
Date: 2005-01-08 08:55 am (UTC)I want to do this, too. I am especially interested in methods of communicating things that I see intuitively, or exploring ideas I have, through some sort of collaborative system.
Once someone formalizes a scientific theory meaningfully, it can be privately tutored to millions of people. The computer would help students play with the theories, using things such as example generation and diagram generation to illustrate principles (theorems); it could create homework problems and correct them, etc. and at a more advanced stage it could help the theorist concretize his intuitions about relations with other theories (e.g. analogy) or the meta-theory.
I am also interested in developing such a system. I think a lot of learning can be done far more effectively (especially by some students) using a system like that. In fact, this is one of my greatest interests in terms of things I would like to bring about as soon as possible. I've been sort of overwhelmed about where to begin.
My own experience with school was one of a substantial amount of boredom and disinterest, and I think that learning can be made far more interesting when there is active interactivity, and the ability for users to explore content at their own pace. And, this could be integrated with mechanisms to assess what they have come to understand in a motivating way.
One way to visualize ideas I found recently is http://cmap.ihmc.us/. It's quite a leap from that sort of thing to a more interactive system, but I'd like to see what can be done.
(no subject)
Date: 2005-01-08 07:17 pm (UTC)I think a good way to begin is:
* implement a system that plays argumentation games in FOL
* formalize a simple, easy scientific theory in FOL
That would already be a system where the student would play argumentation games against the computer. Once this works, you would probably want to work on:
* automatic problem generation
* example generation
* the user interface
* adding more theories
* more general logics
None of these parts seem hard. Which of course means I should be doing it.
A more generic approach is to:
* look at how students learn things in the classroom or with a book
* try to improve the process by replacing inefficient parts
(no subject)
Date: 2005-01-08 09:29 pm (UTC)Btw, how did you find my journal?
(no subject)
Date: 2005-01-08 09:49 pm (UTC)I'm very interested in discussing how something like this could be developed - and probably working on developing it. I think that some of the motivational principles relevant to video games could be applied to other forms of learning, using such an interactive system.
(no subject)
Date: 2005-01-08 09:53 pm (UTC)A few questions:
* What is FOL?
* Are you thinking of something web-based, or a standalone software app?
* What programming languages are you thinking of?
A DOOD named Flora
Date: 2005-01-11 08:03 am (UTC)(no subject)
Date: 2005-01-12 08:20 am (UTC)This is a nice goal, but I think impossible, at least based on Quine's arguments for the impossibility of radical translation (I think that's in "Word and Object"? Wherever he uses the word "gavagai".) and Putnam's model-theoretic argument against the possibility of precise reference. (Basically it comes down to the Lowenheim-Skolem theorem, pointing out that two non-isomorphic domains can satisfy all the same first-order formulas in any given language.)
So teaching people the actual meanings of this one uniform language will be hampered by the same difficulties speakers of supposedly the same natural language already face, when they can never be sure they don't use the same words in subtly (but perhaps importantly) different ways.
The idea of multiple parallel formalizations sounds really interesting though.
(no subject)
Date: 2005-02-02 09:59 pm (UTC)FOL is First-Order Logic: propositional logic + quantifiers (i.e. for all, there exists).
The programming language shouldn't be important, but since you've mentioned it, I think Lisp is TheRightThing.
The form of the app is also not important at this stage. We can worry about such details once the core of the system is working. The point is that it has potential. The medium is secondary.