RULES Patty Nordstrom Hien Nguyen
Dec 31, 2015
Cognitive Skills
Cognitive skills are any mental skills that are used in the process of acquiring knowledge; these skills include reasoning, perception, and intuition.
Cognitive skills refer to those skills that make it possible for us to know.
Production Rules
• Production rules constitute a framework for understanding human cognition
• Production rules are if-then statements or condition-action pairs
Ex. If it snows, then I'll go skiing
Ex. If status='OK' and type=3 then count+1
Representational Power
• Represent general information about the world
• Represent information about how to do things in the world
• Represent linguist regularities
• Inferences such as modus ponens
Computational Power
• Problem solving• Searching, space, heuristics
• Planning• Sequence of rule
• Decision making• Learning
• Acquisition, modification, application
• Language
Psychological Plausibility
• Rule-based systems can account for different types of learning– power law of practice
– conditioning
Practical Applicability
Learning consists of rules so how can this be applied to helping students better acquire rules– Computer tutors– Rule based cognitive systems
• ACT & ACT-R (Adaptive Control of Thought—Rational)
• SOAR (Soar is used by AI researchers to construct integrated intelligent agents and by cognitive scientists for cognitive modeling)
Frameworks
• Frameworks – set of constructs that define important aspects of cognition.
• Frameworks – cannot make predictions, but you can add assumptions to make theories
Theories
• Theories still cannot make precise predictions
• Add assumptions about a specific situation and it is a model of that situation
Models• Models - theories with assumptions
about its application to a specific situation
• Many models possible within a theory
• Production system are theories of human cognition
Cognitive Architectures
• Cognitive architectures are proposals about the structure of human cognition
• Cognitive architecture tries to provide a complete, if abstract, specification of a system
• Production system are theories of human cognition because they are architectures
Features of Production Systems
• Each production rule is a modular piece of knowledge (a well-defined step of cognition)
• Complex cognitive processes:– String a sequence of rules– Writing to working memory (goal setting, etc)– Reading from working memory
• Rules are condition-action asymmetrical• Rules are abstract & apply in many situations
How do production systems operate?
• Pattern matching– Production’s condition vs. contents of working
memory
• Conflict resolution
• Firing a production
-> CYCLE
How to write a production system model?
• Write a set of production rules to perform the task
• For AI, production systems are used as programming formalisms– Precise, complete theories of tasks– Without cognitive modeling
Examples
• A production system for addition
• Various production system architectures:– PSG: first production system implemented as
a computer program– OPS systems
• Efficient pattern matching and conflict resolution
– ACT systems: ACTE, ACT*, ACT-R• Include a separate declarative representation
– SOAR system
ACT-Rhttp://act-r.psy.cmu.edu/about/
• A cognitive architecture: a theory about how human cognition works.
• A framework
• A cognitive skill is composed of production rules.
Are rules psychologically real?
• Appropriateness of rules in describing skilled behavior
• Ability to predict the details of that behavior
Problems
• Is ACT-R the right production system theory?
• Assumption: production system framework is the right way to think about cognitive skill.
Implementation Level Problems
• Algorithm level vs. Implementation level– High-level programming language vs.
machine level implementation
• It is difficult to identify what is going on at the implementation level. – Uniqueness: which implementation is the
underlying internal structure?– Discovery: which implementation matches the
behavior?
Implementation Level Problems
• Uniqueness Problem– Neural approach: use neural-like
computations
• Discovery Problem– Rational approach -> ACT-R– Cognition is adapted to environment structure:
• Memory• Categorization• Causal inference• Problem solving
Intelligent Tutoring Systems
• Previously– CAI vs. ICAI– Impractical
• Costly• Time• No established paradigm for enabling students to acquire
knowledge.
• Now– Cost reduced, advances in AI and cognitive
psychology -> shorter time, advances in cognitive science -> instructional design implications
ITS Model
• knowledge of the domain
• knowledge of the learner
• knowledge of teacher strategies
http://coe.sdsu.edu/eet/Articles/tutoringsystem/start.htm
What an ITS must do
• accurately diagnose students' knowledge structures, skills, and styles
• diagnose using principles, rather than preprogrammed responses
• decide what to do next
• adapt instruction accordingly
• provide feedbackhttp://coe.sdsu.edu/eet/Articles/tutoringsystem/start.htm
ACT-based approach to intelligent tutoring
• Goal structure
• Instruction in Context
• Immediacy of Feedback
• Examples: the Geometry Tutor, the LISP Tutor
Design Scenario
• In your group, discuss the design of an intelligent tutoring system that teaches HTML to highschool students. Please use the ACT-R cognitive architecture and discuss the use of production rules in your design.
• FOCUS:– The degree of learner control– Individual vs. collaborative learning– Situated learning– Intelligent Tutor System vs. regular Computer-Aided
Instruction
References
• http://act-r.psy.cmu.edu/about/• http://en.wikipedia.org/wiki/ACT-R • http://coe.sdsu.edu/eet/Articles/tutoringsystem/start.htm• http://www.cs.cmu.edu/~listen/videos/
1998_video_10_min/lis06.mpg• http://act-r.psy.cmu.edu/papers/Lessons_Learned.html • http://www.ncrel.org/sdrs/areas/issues/content/cntareas/
reading/li1lk23.htm• http://www.audiblox2000.com/cognitiveskills.htm• http://www.britannica.com/eb/article-9053169/modus-
ponens-and-modus-tollens