CMSC 477/677 - Spring 2005
Discussion Questions for Class #5, February 13
Reading: Anderson, "ACT: A Simple Theory of Complex Cognition";
Anderson et al.,
"An Integrated Theory of the Mind."
You may also want to look at the tutorial slides (from the ACT-R
tutorial website,
click on the Introduction
to ACT-R 5.0 slides (they don't seem to have updated the slides
for ACT-R 6.0).
It's a little
hard to follow, since it's just a presentation without context, but will
give you a flavor of what an ACT-R model looks like.
Note that your paper summary need only cover the assigned Anderson papers,
but it's fine if you include commentary/summary of the tutorial slides, or
any other related material that you choose to read. The 1996 paper is
a nice, relatively short overview. The 2004 paper is more up to
date, and contains some interesting results, but you will probably
want to skim through some of the details of the experiments.
Questions:
- What's a cognitive architecture?
- What does ACT-R stand for?
- What is Anderson's main claim in the 1996 article? (Be
prepared to
give the direct quote, and also to summarize in your own words.)
- Production rules are the fundamental component of cognitive reasoning
in ACT-R.
- What's a production rule?
- How are production rules triggered?
- What is the mechanism for resolving conflicts between competing competitions?
- How are the conflict-resolution weights learned?
- What is a chunk in ACT-R?
- How are chunks activated?
- What are the components of the activation level?
- What is meant by spreading activation?
- What does Anderson mean by declarative vs. procedural
knowledge?
- Walk through the example in Figure
3 [Anderson 1996]. Does this seem to you
to be a reasonable model of how people solve simple equations? Why
or why not? Do you think that this approach would scale well to complex
cognitive behaviors such as planning a trip or selecting courses for the
spring semester? Why or why not?
- Describe some domains that you think ACT-R would be especially well
suited for.
- Describe some domains that you think ACT-R would not be well
suited for.
- Do you think that ACT-R is a good model of human cognition?
Why or why not?
- Do you think that ACT-R is a good model to use for implementing artificial
cognition? Why or why not?
Soar vs. ACT-R
- Soar and ACT-R both have production rules. What are the similarities
and differences in the nature or application of these production rules?
- Soar and ACT-R both have chunks. What are the similarities
and differences in the nature or application of these chunks?
- Try to identify domains that would be (1) good domains to apply either
Soar or ACT-R; (2) good domains to apply Soar but not ACT-R; (3) good domains
to apply ACT-R but not Soar; and (4) good domains for neither ACT-R nor
Soar applications.