Modeling The Fan Effect Using Dynamically Structured Holographic Memory - eScholarship

Modeling The Fan Effect Using Dynamically Structured Holographic Memory Matthew F. Rutledge-Taylor (mrtaylo2@connect.carleton.ca) Institute of Cognitive Science, Carleton University, 1125 Colonel By Drive Ottawa, Ontario, K1S 5B6, Canada Robert L. West (robert_west@carleton.ca) Institute of Cognitive Science, Department of Psychology, Carleton University, 1125 Colonel By Drive Ottawa, Ontario, K1S 5B6, Canada between reaction time and the fan of the sentence (the sum of the fans of each of the content words in a test sentence). For example, the time to judge a sentence appearing in the study set with fans of one for both the person and place was 1178 milliseconds, while for a sentence with fans of three for both words was 1514 milliseconds. As a basic phenomenon of human memory, the fan effect has served as a useful paradigm for testing theories of memory retrieval, including those pertaining to recall time and accuracy. DSHM is primarily a theory of how information is represented and structured in memory. However, as such, it also offers mechanisms for retrieving information from memory. Therefore, it ought to be consistent with the prevailing theories of memory retrieval. Mechanisms such as spreading activation (Anderson & Reder, 1999), processes such as retrieval effects (Anderson & Reder, 1999), representation effects (Goetz & Walters, 2000; Radvansky, et al., 1993), and memory capacity (Bunting et al., 2004) have all been proposed as responsible for, or playing a role in, producing the fan effect. The basic mechanisms in DSHM are cue based retrieval and interference. However, informational structures extracted from DSHM can be interpreted as having a particular activation value and are subject to spreading activation Abstract Dynamically Structured Holographic Memory (DSHM), a system based on holographic reduced representations, designed for modeling memory is described. A DSHM model of the fan effect (Anderson, 1974) is presented. Comparisons between the DSHM model and an existing ACT-R model (Anderson & Reder, 1999) are made. It is argued that DSHM can be interpreted as providing a lower-level account of several features of ACT-R such as activation, noise, and the ability to make errors. Keywords: cognitive modeling; the fan effect; holographic reduced representation; spreading activation. Introduction The aim of this paper is to demonstrate that DSHM embodies features of human memory, and does so in a way that is emergent from the manner in which information is structured within it. As such, DSHM provides an account of the implementation of several higher-level features of memory, such as activation and noise, which architectures such as ACT-R take to be basic. The value of this account is that it may help to bridge symbolic account of cognition (i.e., ACT-R), with neurally inspired accounts, especially those that make use of vector representations, e.g., the Neural Engineering Framework (Eliasmith & Anderson, 2003), and holographic reduced representations (Stewart & Eliasmith, DSHM DSHM is a model of memory that makes use of holographic reduced representations (HRR). It is based on Jones and Mewhort’s BEAGLE model of the lexicon (Jones & Mewhort, 2007). The details of the DSHM architecture and the similarities between BEAGLE and DSHM can be found elsewhere (Rutledge-Taylor & West, 2007). Here, only the details necessary for understanding the application of DSHM to the fan effect will be included. Objects in the DSHM system, referred to as items, are represented as pairs of vectors. Every vector in the system is the same length, which typically consists of hundreds or thousand of elements. Each vector consists of real valued elements and has a Euclidean length of 1.0. The environmental vector is a unique static internal representation of the item. The memory vector is dynamic and stores all of the associations existing between the given item and every other item in the system, from the given item's perspective. An item may also reference a set of sub-items, from which it is composed. This set of sub-items is flagged as bearing either an ordered or unordered relationship to one another, (e.g., the words in a The Fan Effect The fan effect is a well known memory phenomenon relating to the time required to affirm (or, reject) a given proposition as true. The effect was first described by J. R. Anderson (1974). Anderson generated a study set of 26 sentences each of which described a fact about one of 16 persons residing in one of 16 places; the phrase ‘the hippie is in the park’ is a well known example. The fan of a word is the number of sentences from the study set in which it appears. The fan of each person and place in the study set is one, two or three. Thus, the 26 sentences can be divided into cells of a three by three table according to the fans of the two content words. Anderson had each experimental participant memorize the study set, and then measured the time that elapsed between the subsequent presentation of a sentence and the participant's signal that the sentence was either a member, or not a member of the study set. The results showed that there was a positive correlation

Récupéré en direct depuis OpenAlex et désinversé. Les résumés ne sont pas conservés dans cette base de données : les index inversés représentent 8,6 Go des 9,3 Go de texte de la base, et le serveur dispose de 13 Go libres.