gobet-lane-tutorial

Description

CHREST Tutorial: Simulations of Human LearningFernand Gobet ([e-mail])Department of Human Sciences, Brunel University,UXBRIDGE, Middlesex, UB8 3BH, U.K.Peter C. R. Lane ([e-mail])School of Computer Science, University of Hertfordshire,College Lane, HATFIELD, Hertfordshire, AL10 9AB, U.K.Abstract theory (Chase & Simon, 1973), CHREST implementsthe essential aspects of the template theory (Gobet &CHREST (Chunk Hierarchy and REtrieval STructures)Simon, 2000). In spite of its historical and contempo-is a comprehensive, computational model of humanrary importance, and the diversity of domains in whichlearning and perception. It has been used to success-modelling has been successfully carried out, the num-fully simulate data in a variety of domains, including:the acquisition of syntactic categories, expert behaviour, ber of people who use or understand the principles andconcept formation, implicit learning, and the acquisition operation of an EPAM/CHREST model remains small.of multiple representations in physics for problem solv-The tutorial is structured so that participants will:ing. The aim of this tutorial is to provide participantswith an introduction to CHREST, how it can be used to 1. Acquire a comprehensive understanding of themodel various phenomena, and the knowledge to carry CHREST computational model and its relation to theout their own modelling experiments. chunking and template theories of cognition;2. Explore ...

Informations

• Publié par: Tapav
• Nombre de lectures: 19
• Langue: English

Extrait

CHREST Tutorial:Simulations of Human Learning

Fernand Gobet ([e-mail]) Department of Human Sciences, Brunel University, UXBRIDGE, Middlesex, UB8 3BH, U.K.

Peter C. R. Lane ([e-mail]) School of Computer Science, University of Hertfordshire, College Lane, HATFIELD, Hertfordshire, AL10 9AB, U.K.

Abstracttheory (Chase & Simon, 1973), CHREST implements the essential aspects of the template theory (Gobet & CHREST (Chunk Hierarchy and REtrieval STructures) Simon, 2000).In spite of its historical and contempo- is a comprehensive, computational model of human rary importance, and the diversity of domains in which learning and perception.It has been used to success- fully simulate data in a variety of domains, including: modelling has been successfully carried out, the num- the acquisition of syntactic categories, expert behaviour, ber of people who use or understand the principles and concept formation, implicit learning, and the acquisition operation of an EPAM/CHREST model remains small. of multiple representations in physics for problem solv- The tutorial is structured so that participants will: ing. Theaim of this tutorial is to provide participants with an introduction to CHREST, how it can be used to 1. Acquirea comprehensive understanding of the model various phenomena, and the knowledge to carry CHREST computational model and its relation to the out their own modelling experiments. chunking and template theories of cognition; 2. Exploresome key learning phenomena supporting the Developing detailed process models of cognitive phenom- chunking theory by taking part in a verbal-learning ena is important to the development of cognitive sci- experiment; ence, as only then can cognitive theories be used to gen- 3. Attempt to match their own data with the perfor- erate quantitative predictions for complex phenomena. mance of a CHREST model of verbal learning; and The history of computational modelling includes many 4. Beintroduced to the implementation of CHREST in diverse approaches, from models of single phenomena sucientdetailtobeginmodellingtheirowndata. (such as Young and O’Shea’s model of subtraction), to integratedmodelscoveringawiderangeofdierentphe- Wehavechosenaverbal-learningexperiment(serial- nomena (such as Soar and ACT-R), to over-arching prin- anticipation method) for introducing participants to ciples, which guide the development of models in dis- CHREST for the following reasons:the experiment is parate domains (e.g. connectionist approaches, or em- historically important; it was one of the motivations be- bodied cognition). hind the development of EPAM; it can be carried out in The EPAM/CHREST tradition, which forms the a short period of time; striking learning phenomena are heartofthistutorial,hasbeenprovidingsignicant readilyobservable,inspiteofthebrevityoftheexper- models of human behaviour since 1959.Early models iment; the motivation and requirements for the experi- ofEPAMprovidedtheimpetustodevelopthechunk- mentaregenerallyclear;and,nally,itillustratessome ing theory, which has been an important component key features of the EPAM/CHREST architecture. in theories of human cognition ever since.Focusing References on learning phenomena, EPAM and CHREST place Chase, W. G., & Simon, H. A. (1973).Perception in a great emphasis on how the model’s information is chess.Cognitive Psychology, 4, 55–81. learnt through interactions with an external environ- ment. Thus,EPAM/CHREST models are typically de- de Groot, A. D., & Gobet, F. (1996).Perception and veloped from large quantities of naturalistic input.For Heuristics of the professional eyememory in chess.. example, in modelling expert perception of chess players, Assen: VanGorcum. actual chess games are used. Feigenbaum, E. A., & Simon, H. A. (1984).EPAM-like Historically, CHREST is derived from the EPAM (Ele- models of recognition and learning.Cognitive Science, mentary Perceiver and Memorizer) model of Feigenbaum 8, 305–336. and Simon (1984). In both models, learning happens as Gobet, F., Lane, P. C. R., Croker, S., Cheng, P. C-H., the creation and elaboration of a discrimination network. Jones, G., Oliver, I. & Pine, J. M. (2001).Chunking In addition, CHREST has mechanisms for the auto- mechanisms in human learning.TRENDS in Cogni- matic creation of schemata and for the creation of ‘lateral tive Sciences, 5, 236–243. links’, which can be used for creating elementary produc- tions or elementary semantic links.CHREST can thus Gobet, F., & Simon, H. A. (2000).Five seconds or sixty? be situated between production systems such as Soar and Presentation time in expert memory.Cognitive Sci- connectionist systems.Just as EPAM was the compu- ence, 24, 651–682. tational embodiment of the key aspects of the chunking