Desktop version

Home arrow Education arrow Knowledge and Action

Source

Associative Learning Models

In this class of theories, human choice is conceptualized as a learning process (Busemeyer & Myung, 1992; Bush & Mosteller, 1955). Learning consists in changing the propensity to select a gamble according to the experienced outcomes. Good experiences boost the propensity of choosing the gamble associated with them, and bad experiences diminish it (e.g., Barron & Erev, 2003 ; Denrell, 2007; Erev & Barron, 2005; March, 1996). Two associative-learning models that have been proposed to capture decisions from experience are the value-updating model (Hertwig, Barron, Weber, & Erev,2006) and the instance-based learning (IBL) model (Gonzalez & Dutt, 2011).

The value-updating model stipulates that learners update their estimates of the value of the gamble after each new draw from it. Specifically, the model computes the weighted average of the previously estimated value and the value of the most recently experienced outcome. The model includes two parameters, namely, the number of draws and a recency parameter. The former parameter is determined empirically; the second is adjustable (i.e., fitted to the data). Importantly, the model does not necessitate representation of probabilities. Furthermore, the best fitting parameter in a model competition indeed suggested a substantial recency effect (Hau et al., 2008).

The IBL model also stipulates a learning process but goes beyond the relatively simple assumptions of the value-updating model: It is assumed that a choice (given that it is not automatically reproduced) represents the selection of the option with the higher utility (blended value). An option’s blended value is a function of its associated outcomes and the probability of retrieving corresponding instances from memory (contingent sampling). Memory retrieval depends on memory activation, which, in turn, is a function of the recency and frequency of the experience. Activation is specified by the mechanism originally proposed in Adaptive Control of Thought—Rational (ACT-R; Anderson & Lebiere, 1998), a cognitive architecture used by cognitive psychologists to model problem-solving, learning, and memory. The IBL model is particularly attractive because it “predicts not only the final consequential choice but also the sequence of sampling selection” (Gonzalez & Dutt, 2011; p. 529; but see Hills & Hertwig, 2012) and because it offers a single learning mechanism (leading up to an instance’s activation) across all experiential designs (Fig. 8.1b-d). Indeed, in a quantitative comparison of models, Gonzalez and Dutt (2011) were able to show that the IBL model predicts final experience- based decisions as well as or better than any other proposed model (including, for instance, the value-updating model and cumulative prospect theory).

 
Source
Found a mistake? Please highlight the word and press Shift + Enter  
< Prev   CONTENTS   Next >

Related topics