Metacognition in Computation

Welcome to the metacognition in computation homepage. Here you will find information on what it is, why it matters, ongoing research, dissertation topics, challenge problems, upcoming events, etc.

Please feel free to write Mike Anderson: anderson -AT- cs -dot- umd -dot- edu with suggestions about what should be included here.

TOPICS:
  1. What is metacognition in computation?
  2. Who is interested in metacognition in computation?
  3. Are there any metacognition in computation related events?
  4. What are some open questions/ research topics?
  5. Are there other related sites and groups?

 

 

What is metacognition in computation?

Imagine two components, X and Y (where X and Y could be the same), related in such a way that state information flows from Y to X, and control information flows from X to Y.

Component X is in a monitoring and control relationship with component Y, and when Y is a cognitive component, we call this relationship metacognitive monitoring and control. Put formally, then, the research question for the subject of metacognition in computation is: what are the sets {X, Y, S, E}—where Y is a cognitive component of a computational system S, and E is its environment—such that having some X in such a relationship with Y provides benefits to the system (and what are these benefits)?

The central research question also suggests a whole host of sub-questions. For instance: how much, and what sort of state information is required for the effective monitoring of cognitive components? What are the options for the control of cognitive components—is it primarily a matter of stopping, starting and otherwise scheduling their operation, or are there effective ways to induce internal changes (e.g. learning)? How much needs to be known about the inner workings of the cognitive component to effectively use/evaluate state information, and give appropriate control commands? What are the kinds of benefits we expect to see from metacognitive components, and, more importantly, how should we measure them? When and why can metacognition cause harm? And, do the answers to these questions depend on the details of the systems in question, or is metacognition largely domain independent?

For a history and research review of the topic, see: Cox, M. T. (in press). Metacognition in computation: A selected research review. Artificial Intelligence.

 

 

Who is interested in metacognition in computation?

 

 

Are there any metacognition in computation related events?

  1. 2005 AAAI Spring Symposium on Metacognition in Computation
  2. 2004 DARPA Workshop on Self-Aware Computer Systems
  3. 1995 AAAI Spring Symposium Representing Mental States and Mechanisms

 

 

What are some open questions / research topics?

  1. Metacogntion in the LIDA Model: A Dissertation Problem

    The LIDA (Learning IDA) model describes a cognitive agent-control architecture based on a broad swath of human cognition, a cognitive theory of everything (Newell 1990). It incorporates sensation, perception, emotion, several forms of memory, consciousness, several forms of learning, and action selection (Franklin 2005, D'Mello and Franklin submitted). An earlier version of the same architecture, IDA, affords an AI technology with practical applications (Franklin 2001).

    The lack of metacognitive abilities constitutes a major gap in the LIDA model, an open problem. Previous attempts at adding metacogntion to similar software agents (Zhang Dasgupta, and Franklin 1998, Zhang and Franklin unpublished) have taken the wrong approach, that of adding a B-brain (Minsky 1985) based on completely different computational mechanisms.

    The LIDA model should allow coherent, integrated, metacognitive capabilities built entirely within its existing architecture. This would entail, among other issues, designing and/or learning attention codelets capable of recognizing the need for metacognitive intervention, as well as behavior streams to effect the needed interventions. Adding such a metacognition module to the LIDA model, and experimenting with it, should constitute a viable dissertation topic for a doctoral student in AI with an interest in metacognition.

    • D'Mello, S. K., and S. Franklin. submitted. A cognitive architecture capable of human like learning.
    • Franklin, S. 2001. Automating Human Information Agents. In Practical Applications of Intelligent Agents, ed. Z. Chen, and L. C. Jain. Berlin: Springer-Verlag.
    • Franklin, S. 2005. A "Consciousness" Based Architecture for a Functioning Mind. In Visions of Mind, ed. D. Davis. Hershey, PA: Information Science Publishing. 149-175.
    • Minsky, M. 1985. The Society of Mind. New York: Simon and Schuster.
    • Newell, A. 1990. Unified Theories of Cognition. Cambridge MA: Harvard University Press.
    • Zhang, Z., D. Dasgupta, and S. Franklin. 1998. Metacognition in Software Agents using Classifier Systems. In Proceedings of the Fifteenth National Conference on Artificial Intelligence. Madison, Wisconsin: MIT Press.
    • Zhang, Z., and S. Franklin. unpublished. Metacognition in Software Agents Using Fuzzy Systems.

 

 

Related Sites