CHREST Tutorial

CHREST is a cognitive architecture that models human perception, learning, memory and problem solving. Its distinctive features include an emphasis on the importance of perception and attention, and the incorporation of human constraints, such as limitations on short-term memory and processing speed. CHREST is particulary suited to modelling human learning in domains involving high-level (symbolic) information.

For the latest information about CHREST, please visit our website: http://chrest.info, sign up to our mailing list, and follow us on twitter @chrest_news.

Tutorial Schedule

This tutorial should take approximately 2 hours. The lecture slides may be viewed by clicking on the link in the table below.

Papers

We have provided some sample papers to introduce the CHREST literature. Note that some of these are preprints. The latest publications are announced on the mailing list, and posted at http://chrest.info.

Theory and introduction

• Chunking theories in human learning: a good first article to read, and the main citable resource.
• Discrimination nets, production systems and semantic networks: explains how discrimination networks may be used to implement a variety of knowledge structures.
• Attention mechanisms in the CHREST cognitive architecture: describes the attention mechanisms within CHREST in some detail.
• Expert Memory: A Comparison of Four Theories: compares some competing theories on expertise with the chunking/template theories.
• Expertise, models of learning and computer-based tutoring: relates the principles of EPAM/CHREST to topics in education, such as the role of multiple representations, and the zone of proximal development.

Some results

Chess

Chess has been influential in the psychological study of expertise, ever since the early work of de Groot in 1946 established the importance of the recall experiment.

• Expertise and intuition: relates the template theory to an understanding of intuition.
• Recall of random and distorted chess positions: study of the recall experiment.
• Five seconds or sixty?: results on chess perception, and the effect of presentation time.
• In search of templates: experimental and simulation results on the presence of templates.
• Multiple boards: a study of how the ability of master chess players to remember multiple chess positions supports the template theory.
• Interpretation (in AGI-12 proceedings): a study training CHREST to reproduce board evaluations provided by a human player.

Implicit Learning

• CHREST models of implicit learning and board game interpretation (in AGI-12 proceedings): A replication of a recent experiment in implicit learning.

Problem Solving

Recent work with CHREST has looked at adding problem-solving behaviour. This has mostly been limited to chess.

• Checking chess checks with chunks: a model of how humans detect whether the king is in check.

Language

• Modelling syntactic development in a cross-linguistic context: a good introduction to MOSAIC and the language experiments.

Ageing

• An investigation into the effect of ageing on expert memory with CHREST: a look at the parameters within CHREST, and an exploration of which ones predict age-related loss of ability.

Neuroscience

The use of fMRI studies in cognitive science has not had a long history, but some useful experiments have been performed.

• Brain localisation of chunks: some fMRI results to compare the locations of brain activity with the predictions of the chunking theory.

Software

The CHREST implementation is in the 'Software' folder of this CDROM. The latest version is available from http://chrest.info/software.html