Murdock, J. W. & Goel, A. K. (1998). A functional modeling architecture for reflective agents. In Proceedings of the AAAI98 Workshop on Functional Modeling and Teleological Reasoning Madison, Wisconsin. 50  Home/Search   Document Details and Download   Summary   Related Articles   Check

....nature of goal trees leads to a natural breakdown of required high level functionality in terms of low level components. Uses of goal trees in computer science research include user interface task modeling [Dix] goal based requirements representations [Dardenne] and reflective redesign [Murdock]. A behavior trace is a representation of an actual or desired program execution. As such it tends to illustrate the design of a system by providing selective examples rather than a comprehensive view. Uses of behavior traces in computer science research include scenario based requirements ....

J. W. Murdock and Ashok Goel. "A Functional Modeling Architecture for Reflective Agents." AAAI-98 Workshop on Functional Modeling and Teleological Reasoning, Madison, WI, July, 1998.

....while ManSART uses the term view manipulation. Our work is complementary to the work done in DALI and ManSART. In our work on legacy system evolution, we have architectural perspectives derived not only from source code, but from requirements engineering activities[26] program adaptation analysis[23], architectural analysis[14] and domain analysis[7] These sources may create perspectives at different levels of abstraction. While pure source code derived perspectives may at least have some naming consistency and comprehensive theme, we can expect no such help when resolving other more diverse ....

Murdock, J. and Goel, A. "A Functional Modeling Architecture for Reflective Agents." AAAI-98 workshop on Functional Modeling and Teleological Reasoning, Madison, WI, July, 1998.

....is to display a file, that external browsers perform display actions, etc. Thus processes of reflection, i.e. a system s reasoning about itself, may be particularly valuable in supporting agent evolution. Autognostic [Stroulia, 1994, Stroulia Goel, 1994, Stroulia Goel, 1996] and Sirrine [Murdock Goel, 1998] are two closely related systems which do reflective reasoning. These systems are based on a theory that elements of cognition can be viewed as abstract devices. Thus reasoning systems can be thought of as made up of distinct components which combine to produce the overall behavior of the system. ....

....is largely defined by the knowledge it uses: knowledge of a system s own reasoning. A core issue which I intend to address in the proposed research is that of the language which is used to encode knowledge of self. I have been developing a language known as TMKL, Task Method Knowledge Language [Murdock, 1998a] which I hypothesize will be able to support reflective reasoning processes such as agent evolution. Work on TMKL is an extension of the Task Method Knowledge (TMK) modeling framework [Stroulia, 1994, Goel et al. 1996, Griffith, 1997, Murdock Goel, 1998] Most of the existing work on TMK ....

[Article contains additional citation context not shown here]

Murdock, J. W. & Goel, A. K. (1998). A functional modeling architecture for reflective agents. In Proceedings of the AAAI98 Workshop on Functional Modeling and Teleological Reasoning Madison, Wisconsin. 50

.... ToRQUE2 system we have modeled four reasoning strategies for exploratory problem solving: model based search (MBS) model based analogy (MBA) structure based model transformation (SBMT) and limiting case analysis (LCA) These strategies are represented in the Task MethodKnowledge (TMK) language (Murdock and Goel, 1998). In the remainder of this paper we describe a control process which enables the selection of strategies through reflection. We claim that this process is sufficient for modeling exploration in science. Reflection A theme which runs through a vast range of research paradigms in cognition is the ....

....clearly erroneous, they are motivated to reason about how they got to that state, even if the process which got them their was largely nonreflective at the time. Much of the past work on TMK models has focused on the first three topics above: adaptation (Stroulia, 1994; Stroulia and Goel, 1995; Murdock and Goel, 1998), explanation (Goel et al. 1996; Goel and Murdock, 1997) and multi strategy reasoning (Punch et al. 1996) In this paper we consider yet another possible application of reflection: that of exploratory reasoning. This idea extends the notion of multi strategy reasoning; instead of simply ....

Murdock, J. W. and Goel, A. K. (1998). A functional modeling architecture for reflective agents. In Proceedings of the AAAI98 Workshop on Functional Modeling and Teleological Reasoning, Madison, Wisconsin.

....with extensive experimentation in this field, there is still only a limited degree of understanding as to what constitutes a useful description of a software system. Task Method Knowledge (TMK) models were originally created as a formalism for allowing intelligent systems to reason about software [2, 4, 5, 6]. These intelligent systems use the information encoded in TMK to manipulate and alter software modeled within the TMK framework. The fact that these systems are able to do so, strongly implies that the content of TMK models is useful for reasoning about software. This suggests that the content of ....

J. William Murdock and Ashok K. Goel. A functional modeling architecture for reflective agents. In Proceedings of the AAAI98 Workshop on Functional Modeling and Teleological Reasoning, Madison, Wisconsin, July 1998.

....this body of work makes to the argument I am making here is: An accurate causal, compositional, functional model of a complex system inherently constitutes a deep understanding of such a system. The Task Method Knowledge (TMK) modeling language (Stroulia, 1994; Goel et al. 1996; Griffith, 1997; Murdock and Goel, 1998) provides a causal, compositional, functional framework for describing cognitive capabilities. TMK models have been used to support processes such as explanation and adaptation in a variety of AI systems. For example, Murdock and Goel, 1998) describes a recently developed agent architecture for ....

....(Stroulia, 1994; Goel et al. 1996; Griffith, 1997; Murdock and Goel, 1998) provides a causal, compositional, functional framework for describing cognitive capabilities. TMK models have been used to support processes such as explanation and adaptation in a variety of AI systems. For example, (Murdock and Goel, 1998) describes a recently developed agent architecture for implementing, executing, and adapting TMK models. TMK models are very much an extension of the physical device modeling framework from which they were derived. The division of reasoning into tasks (i.e. functional elements) and methods (i.e. ....

[Article contains additional citation context not shown here]

Murdock, J. W. and Goel, A. K. (1998). A functional modeling architecture for reflective agents. In Proc. AAAI98 Workshop on Functional Modeling and Teleological Reasoning, Madison, Wisconsin.

....the version in Autognostic but is considerably simpler since the language is used simply to control processing and not to enable self redesign. ffl SIRRINE 1 is essentially an extension of Autognostic; the language used in this project is essentially the language of Autognostic. ffl SIRRINE 2 [Goel Murdock, 1998] is a relatively simple system which is being developed concurrently with the writing of this report. It is intended to be act both as a vessel for exploring very short term research goals and as a prototype for a new system to be based on the TMKL language being developed here. SIRRINE 2 uses a ....

Goel, A. K. & Murdock, J. W. (1998). A functional modeling architecture for reflective agents. In Proc. AAAI98 Workshop on Functional Modeling and Teleological Reasoning (in press) Madison, Wisconsin.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC