Model Of Concept Learning Research Articles

Concept Learning

The dominant model of concept learning in experimental psychology, the hypothesis testing model, is described and its applicability to education is questioned. One issue related to the generality of this model was tested by an experiment in which subjects attempted to figure out a simple but difficult to discover concept presented by exemplars over a long series of trials. Based on their performance on this task, subjects were categorized into three levels of concept attainment. It is argued that hypothesis testing theory adequately describes only the behavior of the most successful (Level 1) subjects. Based on the differential levels of performance on old vs. new items and the subjects’ verbal reports of their selection strategies, Level 2 subjects appear to have relied heavily on memorization of previously seen exemplars. However, their above-chance performance on the new (postcriterion) items indicates some level of concept attainment. Since most of these adult subjects were totally unable to v...

A TASK-FREE CONCEPT LEARNING MODEL EMPLOYING GENERALIZATION AND ABSTRACTION TECHNIQUES

Abstract An interactive concept learning model is described. The model utilizes generalization and introspection techniques for building a set of concepts in the form of a context-sensitive associative network, and is able to describe its memory structure in simple English sentences. Four task domains are examined: contract bridge bidding, poker betting, learning of paired associates, and the diagnosis of liver disease.

Sequence Effects in Rule Learning and Conceptual Generalization

Subjects were given 12 training problems based on four primary bidimensional rules, the problems mixed so that the rule changed after every problem (to facilitate adoption of a truth-table strategy; condition M) or systematically blocked by rule (condition S). Their next tasks were to learn an unpracticed complementary rule, to discover a tridimensional rule, and to identify a rule based on novel and meaningful stimuli. The findings show that condition M facilitated the subjects' ability to abstract, generalize, and extend a higherorder strategy. They suggest the importance of the sequence as well as the variety of rules and imply a hierarchical model of concept learning.

Concept learning and the intradimensional spread of the relevant values of the instances I. Effects of intraclass variability

Abstract.— The effect on concept learning of the intra‐class variability of the relevant values of the instances was investigated. 20 subjects took part, and two ranges of the concerned intraclass variability, one about twice the other, were used. Depending upon the learning strategy used by the subjects the smaller intraclass variability was shown to have a limited facilitating effect on concept learning. Some facilitation thus occurred when strategies involved a rather unselective attention to the appearance characteristics of the instances but not when they involved selective focusing on these characteristics, or on elements of relevant knowledge established pre‐experimentally. The results were discussed in relation both to relevant S‐R‐conceptions and information processing models of concept learning. It was concluded that automatic generalization as described in S‐R‐theories might have to be integrated into the information processing models for some categories of learning.

Effects of attribute identification training on rule effects in an attribute identification transfer task

The hypothesis that attribute identification (AI) pretraining would reduce or eliminate rule effects in AI tasks was investigated. Ss were given 0, 4, or 8 AI problems based on the same rule prior to transfer to a final AI problem. Results indicated that familiarity with AI problems was not a key factor in eliminating rule effects in transfer. Implications for a two-component model of concept learning were discussed.

Continuous-discontinuous concept attainment as a function of individual differences in cognitive style.

Continuities and discontinuities in concept learning were studied in 30 white, male, middle socio-economic level, kindergarten children differentiated with respect to field independent-field dependent cognitive style. The children, homogeneous as to age and Verbal IQ, were presented with a two-choice simultaneous discrimination problem. Backward learning curves from the last error before criterion were constructed. As predicted, the learning curve for the field-independent group was characterized by discontinuity over precriterion trials, with no statistical dependence from one trial to the next. The curve for field-dependent Ss exhibited marked continuity, with significant departures from stationarity, and significant dependence in learning from one trial to the next. The results were interpreted in light of the incremental and mediational models of concept learning, and highlights the use of perceptual orientations in the formation of mediators as a developmental function.

A model of memory in concept learning

A concept learning model was developed and tested in two conjunctive attribute identification tasks. The model includes assumptions about the focus of attention, decision making, and memory for stimulus information and prior decisions. Predictions are made about how S changes his hypothesis following an error. Procedures in both tasks allowed inference of the subject's current hypothesis. The hypothesis selections and error statistics were in the majority of cases accurately predicted by the model. Deviations from predictions on hypothesis sampling occurred for naive Ss but not for trained Ss who were required to state a hypothesis on each trial.

An Extension of Four Single-Cue Process Models to Conjunctive Concept Learning

A family of 4 different process models of single-cue concept learning was extended to 2-category conjunctive concept learning. The fit of these 4 extended models to data from 120 Ss solving conjunctive problems over a wide range of stimulus complexity was assessed. None of the models did a thoroughly satisfactory job of fitting the data over all levels of stimulus complexity. Possible reasons for this lack of fit were discussed.

An on-line technique for system identification

In most practical applications, dynamic systems can be modeled by differential equations with prescribed forms but unknown parameters. An analog method is developed for identifying these unknown parameters. The method is based on the learning model concept. Therefore, it is capable of performing on-line identification in the presence of noise. It has shown that single-variable and multivariable linear systems can be easily identified. Also the same algorithm is applicable to zero memory nonlinear systems. Hence the proposed method is regarded as very useful Various examples simulated on a digital computer are presented and the results are good.

OBSERVABLE CHANGES OF HYPOTHESES UNDER POSITIVE REINFORCEMENT.

In mathematical models of concept learning it has consistently been assumed that positive reinforcement cannot lead to a change of the hypothesis determining the overt response. When hypotheses are experimentally identified and recorded along with positive and negative reinforcements of stimulus-response pairs, it can be shown that hypotheses may change after a positive reinforcement. Positive reinforcement has an information content for subjects that has not yet been adequately recognized in concept formation studies.