Comment: Consumers' Memory For Product Knowledge

Reid Hastie, Psychology Department, Northwestern University
[ to cite ]:
Reid Hastie (1982) ,"Comment: Consumers' Memory For Product Knowledge", in NA - Advances in Consumer Research Volume 09, eds. Andrew Mitchell, Ann Abor, MI : Association for Consumer Research, Pages: 72-73.

Advances in Consumer Research Volume 9, 1982      Pages 72-73


Reid Hastie, Psychology Department, Northwestern University

To begin with, I would like to propose a general five-part model to describe the cognitive representations and processes that yield a decision such as a judgment of product desirability or a preference among products. This general model is summarized in Figure 1. First, there is generic knowledge which includes general information about classes of products, instances exemplifying the products, the existence of different types of products (implying knowledge of correlations between product attributes such as luxuriousness and gas consumption of an automobile), and information about the attributes or dimensions that are relevant and important in making decisions concerning the products (e.g., the distribution of existing product types along the price dimension).

Second, there is individual knowledge about specific products that are involved in a judgment or choice. This would include information such as the prices, color, taste durability, features, etc., of each product. Furthermore, this knowledge structure would include information about relationships among the products. For example, that product X is more expensive than product Y, or that product A and product B are manufactured by the same company.

Third, there is information that serves as ingredients to the judgment or choice strategy. This information might include specific facts from either individual or generic knowledge stores concerning the products, as well as information that is inferred (not given) about the individual products that may be relevant to a specific judgment strategy. For example, if product durability is deemed to be important in making a decision and durability information is not available, the subject will be likely to infer product durability for each product under consideration. This durability information would be an ingredient in judgment and decision stages of the overall process.

Fourth, there is a judgment procedure that combines :he ingredient information to render a characterization of each product that is reflected in :he subject's individual product ratings. The most general judgment procedure appearing in virtually every theory of judgment and choice is the linear weighted combination rule (e.g., Information Integration Theory's weighted averaging rule, Social Judgment Theory's additive composition rule, Decision Theory's expected utility, and so forth). However, alternative representations of the product, for example, as an image composed of perceptual and conceptual features, would also be admitted.

Fifth, decision strategies are concerned with commitment to an action such as purchasing or not purchasing a product. Here there is a greater variety of processing models. Some strategies compare unitary representations of products to choose an optimum (highest expected utility) or a minimally acceptable (satisficing) item; some search product-by-product for a conjunctive intersection of jointly sufficient features or a disjunction of separately sufficient features; some involve an attribute-by-attribute comparison of products that eliminates products until a winner is chosen (elimination by aspects); some compare products to a prototype instance to select the product most similar to the ideal instance: and so forth.



The arrows in Figure 1 indicate the flow of information from component to component in the general model. It is important to note that information from generic knowledge and individual product knowledge are inputs into the judgment procedure as well as inputs for inferred ingredients for the procedure. For example, information in generic form about the importance of product attributes in choices will influence some of the judgment and decision procedures (e.g., weights on a dimension, order in which attributes are considered, and so forth). Similarly, generic knowledge will determine what some of the ingredients look like, particularly when inferences concerning unknown attributes of a product must be made. Individual product knowledge will have its primary effect on the decision strategy through the form in which individual products are displayed. For example, one display format of several products will induce a subject to use an attribute-by-attribute elimination strategy while another display will lead to a product-by-product strategy.

The most important message from Figure 1 is that theoretical issues concerning the representation of generic product knowledge and individual product knowledge must be considered in the light of the functions that this information is expected to serve in natural decision tasks. Thus, it is important to have information in generic knowledge structures concerning the distribution of product types and the importance of various product attributes to guide the formation of a judgment procedure or, when inferences are made, to create ingredients for the judgment procedure. It is interesting that many theoretical issues in cognitive research on the structure of very long-term semantic memory have been resolved by a consideration of the deductive processes (analogous to judgment strategies) that use the information. Thus, a hierarchical representation of information concerning noun categories is a particularly economical structure for semantic memory if one of the main functions of semantic memory is to allow category membership inferences. The simple task analysis, summarized in Figure 1, may serve as a useful beginning to an analysis of the function-knowledge relationship in consumer choice and judgment tasks.

In short, what I am saying is that a typology of judgment and choice strategies (part of a larger task analysis) will be a useful precondition to an analysis of knowledge content and structures in various memory stores. I am optimistic that there are currently only a few relatively simple judgment procedures and decision strategies that are relevant to most consumer decisions.

The following comments on papers presented at the Association for Consumer Research meeting panel on memory for product knowledge were stimulated by the preceding analysis of memory function. Jerry Olson has attempted to measure the structure and content of generic knowledge of products. His correlational analysis suggests that he prefers a dimensional or spatial model as a metaphor for the structure of the generic knowledge store. If all of the choice and judgment strategies that we wish to postulate in the more general model require only similarity judgments among products, or preference orderings among products, the spatial metaphor may be adequate. However, if more complex relations among products are needed to infer ingredients to the judgment strategy or in forming a decision strategy then the spatial metaphor will fail and must be replaced by hierarchical or more complex structures. Furthermore. there are types of information that I suggest should be included in generic knowledge (e.g., information about the correlations among product attributes or information about the importance of attributes in decision procedures) that are not included in Olson's model.

The paper by Gabriel Biehal and Dipankar Chakravarti reported work that touches on the judgment procedure, decision strategy, individual product knowledge, and generic product knowledge, but without committing itself to a specific process model or structural representation for any of the components. In particular, I think a first goal of this work should be to attempt to identify the decision strategy employed by subjects attempting to perform the choice task. A second goal would be to characterize subject's generic knowledge concerning the relevant products, or the ingredient information input to the judgment procedure. Individual product knowledge was measured using a standard recall task. It would also be important to explore some of the relations between generic knowledge SUCh as subjective dimension weights for the decision and memory for individual product information.

The paper by Raymond Burke and Thomas Srull uncovered an interesting phenomenon that involves an interaction between decision strategy and individual product knowledge (knowledge for a product when a related product is presented or is not presented). Here, again, generic product knowledge was not measured nor was an effort made to capture the subject's judgment process. The phenomenon appears to be of substantial magnitude and a number of plausible explanations can be advanced for future test. First, research on interference in A-B, A-D paired associate learning has shown that, under some conditions, presentation of the A-D term facilitates recall of the A-B term. Presumably facilitation occurred because of increased rehearsal of the A-8 term of presentation of the A-D term. Second. an explanation based on Watkin's cue-overload principle postulates that the effectiveness of a retrieval cue for a particular item of information is inversely related to the number of items of information associated with that cue. Thus. a cue associated with many pieces of information is "overloaded" and is a relatively ineffective cure for any single piece of information. Third. a schema assimilation interpretation based on the work of Thorndyke and Hayes-Roth suggests that as more related events occur, a schema representation is induced, but at the same time memory for the details of events is lost.

Eric Johnson's paper reported an application of cluster analysis fitting a hierarchical structure to the representation of individual product information. My intuitions agree with Johnson's analysis that a simple spatial model will not be sufficient. However, it would be interesting to see some sharp tests of "badness-of-fit" comparing spatial and hierarchical tree structure fits or a summary of specific characteristics of the data that defy representation in the simple spatial mode. Johnson's earlier work has constrained possible models for the subjects' decision strategy in choice and judgment tasks with his product materials. However, again, t would be important to specify a clearly defined process model sufficient to perform Johnson's choice and Judgment tasks. (I believe that his model DECIDER is a useful step in this direction.)

In summary, I am advocating attention to three theoretical questions. One. what functions will generic product knowledge and individual product knowledge be used to perform in natural consumer decision tasks? Two. what are the major judgment procedures and decision strategies (expressed as process models) that subjects use when evaluating consumer products? Three. what are the inferred information ingredients that are required as input in each of the judgment procedures or decision strategies?