Talk:Frame problem

From Wikipedia, the free encyclopedia

Socrates
Philosophy Portal
This article is within the scope of the WikiProject Philosophy, which collaborates on articles related to philosophy. To participate, you can edit this article or visit the project page for more details.
??? This article has not yet received a rating on the quality scale.
??? This article has not yet received an importance rating on the importance scale.

Contents

[edit] Deleted external links

[edit] Solved?

We need to mention that the technical problem is regarded as solved.---- Charles Stewart 10:43, 16 Sep 2004 (UTC)

[comment posted on 15:13, 9 April 2006 moved at bottom]
In what sense do you mean it's solved? Are you refering to the use of default logics and so on? I wouldnt say that the frame problem is 'solved' per se, but there have been some good stabs at it. It's really a representational issue more than a problem. Deepak 13:39, 16 Sep 2004 (UTC)
The problem arose as a formalism for AI agents in the situation calculus. Over the last 10 years several formalisms were proposed that handled the frame problems in ways that most AI researchers regard as adequate (I know the fluent calculus, but there are several such problems). It's still an active source of research papers, but the emphasis is on what is the best way to model default reasoning, not whether it can be done. There's a brief duscission in the Stanford Encyc. article, I will see if I can track down a better reference. ---- Charles Stewart 07:43, 20 Sep 2004 (UTC)
There's Reiter's Solving the Frame Problem as well. My point is, as such, it is better not to talk about the Frame Problem as being solved or unsolved, but as an issue to be dealt with. Non-monotic/Default logics and the fluent calculus are designed partly so that the frame problem becomes manageable. That doesnt make it 'solved' in a general sense. Deepak 13:17, 20 Sep 2004 (UTC)
The SEP article talks of distinguishing the technical from the philosophical problem. I'm happy for there to be discussion of this as an ongoing spur to thought, provided there is some recognition that various solutions have been proposed and thought to have addressed the original McCarthy/Hayes problem.
I know of two other approaches that have been proposed: Frank Brown's modelling of situational change in a second-order quantified modal logic and an approahc based on linbear logic (forget the name). A brief survey of types of solution proposed would be in order. ---- Charles Stewart 11:22, 21 Sep 2004 (UTC)

[edit] Change

I have introduced a technical explanation of the problem as was initially specified in artificial intelligence. I have removed the following sentence (I would not object to reintroduction, if references are provided).

Suggested solutions to the frame problem include satisficing, heuristics, and rational ignorance.

Paolo Liberatore (Talk) 10:51, 1 October 2005 (UTC)

[edit] Really solved?

[moved from above to keep the above thread whole]

Is the frame problem really "solved"? In my opinion, in most cases this problem is only "elluded" by using some naive assumptions that would make AI systems less useful in real world applications. For example, in many proposals (such as those provided by situation calculus), a bold assumption is an "inertia" world that common sense law of inertia can apply. This means all the relationships between logic formulae used in precondition and effects are enumerated by the action axioms. In other words, there are NO axioms among the condition formulae themselves. This is kind of unrealistic in real world. Just imagine now I have to define some actions using formulae defined by a domain model. For instance, I want to design the actions of an agent reasoning about an OWL knowledgebase. Then the "inertia" world approach would require us to list not just the direct effects, but ALL of their logic consequences! —Preceding unsigned comment added by 68.68.104.203 (talk • contribs) 15:13, 9 April 2006

As it was defined, the frame problem is about the non-effects of actions. The problem of formalizing the indirect effects of actions is the ramification problem. The name "frame problem" has been at some point used to indicate more generally the problem of "formalizing domains with actions in logic", but I think this use is obsolete now. Regarding inertia, this was the initial assmption in the original problem; there are however some logics where fluents can be specified not to be inertial. - Liberatore(T) 10:06, 10 April 2006 (UTC)

[edit] Modular reasoning

The frame problem is also present in the world of modular reasoning in formal verification. Solution strategies here include modify-clauses, separation logic, ownership (dynamic or static) and dynamic frames. Could anybody add a section about it? If nobody does and nobody minds, I will write a paragraph about it within the next month. --Schoelle 13:55, 13 December 2006 (UTC)

[edit] General comment

For me it looks like a very specific scientific publication - too much details for an encyclopedia article. On the other hand, is it a part of computational philosophy or rather AI, or a research in mathematical logics ?

- For instance, is "the simplified example of the door and the light" a simple philosophical problem?

What about a critical viewpoint? - a visitor --87.20.192.228 23:16, 7 November 2007 (UTC)

[edit] Quote?

It might be interesting to quote the following section from the original McCarthy and Hayes paper that (I think) given a good informal description of the frame problem:

[...] in proving that one person could get into conversation with another, we were obliged to add the hypothesis that if a person has a telephone he still has it after looking up a number in the telephone book. If we had a number of actions to be performed in sequence we would have quite a number of conditions to write down that certain actions do not change the values of certain fluents. In fact with n actions and m fluents we might have to write down m n such conditions.

What to you think? --Schoelle 18:17, 10 November 2007 (UTC)