Information Utilization: a Validation Study

Peter C. Wilton, University of California
ABSTRACT - Historically, most consumer information processing research has focussed upon the use of persuasive information in brand choice tasks. Of potentially increasing importance, however, is a second class of information which might be labelled objective, or non-persuasive, in nature. Much less is known about consumer use of this information.
[ to cite ]:
Peter C. Wilton (1986) ,"Information Utilization: a Validation Study", in NA - Advances in Consumer Research Volume 13, eds. Richard J. Lutz, Provo, UT : Association for Consumer Research, Pages: 354-359.

Advances in Consumer Research Volume 13, 1986      Pages 354-359

INFORMATION UTILIZATION: A VALIDATION STUDY

Peter C. Wilton, University of California

[Work on this project has been sponsored by a grant from the National Science Foundation, #PRA 81-10052, with the author and John G. Myers as principal investigators. The views expressed in this paper are those of the author and do not necessarily reflect the position of the Foundation.]

ABSTRACT -

Historically, most consumer information processing research has focussed upon the use of persuasive information in brand choice tasks. Of potentially increasing importance, however, is a second class of information which might be labelled objective, or non-persuasive, in nature. Much less is known about consumer use of this information.

In an earlier paper, we presented a reduced model of the process of utilization of objective information in consumer judgment tasks. The model has been developed and tested extensively in a series of controlled laboratory experiments. In the current paper, we present partial validation results from a field study of information use among a sample of potential adopters of a new technology. The results are strongly supportive of the proposed model and laboratory findings.

INTRODUCTION

In a series of recent articles, the process of utilization of objective (non-persuasive) information in judgment tasks has attracted the attention of consumer researchers (Deshpande and Zaltman 1982, 1984; John and Martin 1984; Wilton and Myers 1984, 1985, Sternthal and Craig 1982). As used in this paper, objective information refers to stimuli whose primary purpose is to inform, rather than persuade, consumers. Models of the utilization process have been offered for both managerial and (traditional) consumer contexts. While the unit of analysis (individuals in institutions vs. individuals in households) may introduce certain context-specific model constructs, there also exist many theoretic commonalities across the two contexts. In this paper, we focus upon some of these commonalities, and report partial results of a field study designed to validate selected information utilization constructs developed and tested in a series of earlier laboratory experiments.

A Model of Information Utilization

Since this paper focusses largely on validation of a model of the utilization of objective information in judgment tasks developed in a laboratory setting (Wilton & Myers 1984, 1985), discussion of model development will here be kept relatively brief. In the section which follows, therefore, we provide only a summary description of the model, and its associated hypotheses. More elaborate discussion of utilization theory can be found in the articles referenced above, as well as Bettman (1979), Berg et alia (1978), Weiss (1979), Sabatier (1978), Gerstenfeld and Berger (1980) and, for persuasive information, Cacioppo and Petty (1983) and Mitchell (1983).

Previous studies of the utilization of objective information have generally suffered from one or more of the following limitations. First, there appears to be a paucity of empirical data amenable to controlled hypothesis testing, particularly among (but not limited to) studies reported outside the marketing discipline. Second, among causal studies supported empirically, there remains considerable ambiguity over both the definition and measurement of key information utilization constructs. Many of these frustrations have been carefully reported by Deshpande and Zaltman (1985, 1984). In response to both these problems, a parsimonious model of the utilization of objective information has been proposed and tested in a laboratory setting by Wilton and Myers (1984, 1985).

This model, shown in Figure 1, attempts to refine the information utility construct, as well as enhance understanding of the process by which utility judgments are formed and modified. According to this model, when asked to choose from among a set of information items, consumers will first develop a set of expectations of the utility of each item. These utility expectations are hypothesized to be a function of the level of uncertainty associated with the judgment task, and the expected "performance" (part-worth) of the information along a set of attributes. In this paper, these attributes are parsimoniously described as the relevance, novelty, credibility and comprehensive ease of the information. The performance expectations are also based upon the individual's store of existing knowledge about the judgment task, and the set of cues (words, phrases, headings) available to summarize some block of more detailed information on an object. The individual is presumed to be a utility maximizer, and will if allowed, exhibit information choice and cognitive elaboration patterns consistent with his/her utility expectations.

The model also proposes a process of (dis)confirmation of utility expectations to describe the modification of utility judgments as a result of exposure to information. In the consumer satisfaction literature (Oliver 1977, 1979; Churchill and Suprenant 1982; Wilton and Tse 1983), expectancy disconfirmation has been offered as an important paradigm for explaining the consumer's response to a product following usage. According to this paradigm, a consumer's expectations of product performance may be confirmed (results consistent with expectations) or disconfirmed (results discrepant with expectations) during the consumption experience. If expectations are disconfirmed, the results may be either positive or negative. Negative disconfirmation (and hence dissatisfaction) will occur whenever the product is seen to perform worse than expected, while positive disconfirmation (and hence satisfaction) will occur whenever the product performs better than expected. In this paper, we extend the expectancy disconfirmation paradigm to the consumption of information products and services by explicitly modelling the formation of utility expectations and by manipulating the disconfirmation experience.

Although necessarily incomplete, the proposed utilization model does include factors found to be important in other contexts, some of which (e.g., information characteristics) are associated with significant measurement problems, and also introduces other factors (e.g., task situation and expectancy disconfirmation) which have not yet been examined empirically. The weakening of external validity implied by this simplified model, however, is considered a necessary trade-off against improved construct and process measurement, given the significant problems associated with prior utilization studies.

The testable hypotheses which can be derived from this model of information utilization can be summarized as follows: H1: Individuals facing higher task uncertainty will perceive greater utility for, and show greater use of, information than individuals facing relatively lower task uncertainty.

FIGURE 1

SIMPLIFIED MODEL OF INFORMATION UTILIZATION

H2: Utility for information can be represented as a multiattribute function of its characteristics. In particular, utility for information is positively and uniquely related to the:

H21: relevance

H22: credibility

H23: novelty

H24: comprehensive ease

of arguments expected/perceived to be contained in the item.

H3: Modification of utility judgments can be represented as a process of expectancy disconfirmation during, or following, information exposure. In particular:

H31: the higher the expected utility for information, the more likely expectations will be negatively disconfirmed during utilization, and

H32: the lover the expected utility for information, the more likely expectations will be positively disconfirmed during utilization.

Extended discussion of the theoretical support for these hypotheses is contained in Wilton and Myers (1985). It should be recognized here, however, that hypotheses H1, H22 and H 23 are by no means obvious. Our hypotheses concerning the relationship between uncertainty and information utility, and the effects of novelty on utilization of information are in direct contrast to those proposed by Deshpande and Zaltman, who argue that (in organizational contexts) new information and/or high uncertainty will lead to lower, rather than higher, utilization. Similarly, evidence concerning the effects of credibility of arguments on utilization is also equivocal. Sternthal et alia (1978) and Kehret-Ward (1984), for example, in studies of source credibility, have found that moderately credible sources may actually induce higher cognitive elaboration among subjects than sources which are highly credible. Thus, both constructs warrant further study in information utilization research.

Model Operationalization

The model has been tested in a controlled laboratory experiment reported by Wilton and Myers (1985), and subsequently validated in a field setting, the results of which are the focus of the current paper. In these experiments, subjects were asked to study a data-base of objective information about a new technology (prepared by a non-partisan organization) with a view towards evaluating the technology. Manipulated variables in these experiments were task uncertainty (2 levels) and expectancy disconfirmation (3 levels).

(i) Task Uncertainty Under the high uncertainty task condition, subjects were asked simply to examine as much or as little of the database as they felt they needed in order to identify all the important issues associated with the technology, to be used in preparing a background briefing to their superior. Since no explicit decision was required and the task parameters are essentially unbounded, this condition is referred to (after Berg et alia 1978) as a conceptual task orientation. Under the low uncertainty condition, subjects were asked to complete a hierarchical budget- or manpower-allocation problem, allocating a fixed set of resources across a pre-identified set of policy/investment options. Since this task is decision-related, it is referred to as an instrumental task orientation. For both conditions, arguments contained in the data-base bore a direct relation to the task assignment.

(ii) Expected Utility Disconfirmation Expectancy disconfirmation can be manipulated by controlling the utility level of the information to which the subject is first exposed. In general, negative disconfirmation of expectations is more likely to occur among individuals who already hold high expectations for the information, while positive disconfirmation is more likely to occur among individuals holding low expectations for the information. If expectations are low, then either the information will in fact be seen (during processing) as low utility, in which case expectations will be confirmed; or the information will be seen as higher in utility than anticipated, in which case expectations will be disconfirmed positively. Analogously, if expectations are high, then high utility information during use will be confirming, while low utility information will be negatively disconfirming.

In the current experiments, subjects first provided measures of the expected utility of information contained in various chapters/sections of the available data-base (both overall and by attribute). For any given subject, these scores were then used to identify which information was relatively higher, lower or indifferent in expected utility. Using a quota procedure, subjects were then assigned to one of three forced-exposure conditions. In the high utility condition, a subject was given information from the set of items which he had rated as having high expected utility. In the low utility condition, the subject was given information from the set of items scored low in expected utility. Subjects who rated the expected utility of all information items equally were assigned to the indifferent utility condition and given an information item chosen from this set at random. This manipulation thus served to operationalize the expectancy disconfirmation process described above.

Following this forced-exposure treatment, subjects were allowed free access to all remaining information contained in the data-base, subject to a time constraint of approximately one-and-a-half hours. Data-collection and information-exposure sequences of the research employed a computer-interactive system called STARSIS, or Strategic Technology Assessment ReSearch and Information System. This environment also precluded any possibility for group interaction among subjects. Total time for completion of all experimental tasks averaged two-ant-one-half hours, for which subjects (in the laboratory experiment) received twenty dollars.

LABORATORY RESULTS

Subjects for the laboratory experiment were 57 female and 94 male volunteer graduate students at a large Western University, evenly divided among Masters and Doctoral degree programs in a wide variety of disciplines, including seventy-two percent from outside the School of Business Administration. The data-base for the laboratory experiment comprised a 65-page National Science Foundation report discussing the nature and consequences of life-extending technologies in the U.S.. Results from this experiment, discussed in detail in our earlier paper, can be summarized as strongly supporting the overall model and each individual hypothesis.

Information was shown to have higher utility in tasks which were not tied to a choice dilemma subjects in the instrumental task condition not only expected lower utility, but also reported lower utility after information was made available. Utility scores estimated as a linear combination of ratings of the information along its attributes provided high correlations with alternative, self-reported, overall measures of expected/perceived utility. Consistent with hypotheses H 21 to H;!4, each individual attribute was shown to have a positive association with overall utility. A positive relationship for the novelty attribute is important since it suggests that not only can new arguments be tolerated during utilization, but that anticipation of new information is a strong inducement towards initial utilization.

Utility judgments were also significantly affected by the expectancy disconfirmation manipulation. For subjects in a high expected utility condition, expectations of both the relevance and novelty of arguments were strongly negatively disconfirmed during exposure, with resulting declines in overall utility ratings. Conversely, low initial expectations on these attributes tended to produce positive disconfirmation with corresponding increases in overall utility ratings.

Thus, in a controlled laboratory environment, our results indicate that the influence of task uncertainty and information characteristics on the process of formation and modification of utility judgments can be validly measured. In the remainder of this paper, we focus our attention on validating the proposed model in more complex (externally valid) task environments.

FIELD VALIDATION

Subjects for the field validation experiment were sixty landscape specialists and maintenance engineers in a large, Western, state-funded transportation organization. This sample represented the entire population of advisors and potential adopters for a new technology (integrated pest management) for controlling scale infestations along the state highway system. Information about this technology was contained in a 40-page document prepared by a team of researchers at a large, prestigious Western university. Data from the report were mate available via the STARSIS computer-interactive data-collection system described earlier, in individual sessions lasting from one-and-a-half to three hours, administered in the subject's usual place of work activity.

Pre-Exposure Manipulation Validation

Table 1 gives the mean scores, and associated F-statistics on a variety of individual cognitive measures for each treatment condition, prior to exposure to objective information. The scores verify the effectiveness of the experimental manipulations. Subjects in the conceptual task condition, for example, show significantly higher expected utility for available information in completing the assigned task than do subjects in the low uncertainty, instrumental condition (means 4.4 and 2.9 respectively).

TABLE 1

GROUP DIFFERENCES, PRE-EXPOSURE

Similarly, consistent with the treatment design, expected utility scores for information assigned during the forced-exposure treatment are significantly lower for subjects in the low-expected-utility condition than for either subjects in an indifferent- or high-expected-utility condition (mean PECS scores, 7.2, 20.5 and 34.6 respectively). Thus, subjects appeared to enter the usage experience with variations in expectations consistent with the experimental manipulations or group assignments. On all other pre-exposure measures of the value of information or the innovation, the differences between experimental groups are not significant.

INFORMATION ATTRIBUTES AND INITIAL UTILITY EXPECTATIONS

According to the utilization model proposed, the expected utility score for information can be expressed as a function of ratings of that information along its attributes. This hypothesis is tested in Table 2 which shows the mean expected performance rating of information assigned during forced-exposure, on each information attribute, for each of the three expected utility groups.

If overall utility scores are related to the individual attribute scores, then these three groups should differ in their ratings of information attributes, and group membership should be predictable by some linear combination of the attribute scores.

TABLE 2

EXPECTED UTILITY FOR INFORMATION AS A FUNCTION OF INFORMATION ATTRIBUTES - DISCRIMINANT ANALYSIS - MEAN SCORES, F-RATIOS

The results are strongly supportive of the hypothesized relationship between overall utility expectations and individual attributes. On all but one attribute (expected comprehensive ease), subjects in the low-expected-utility condition expect assigned information to perform significantly worse than do subjects in either the indifferent- or high-expected-utility conditions. Consistent with hypotheses H21 to H23, the effect of each attribute on overall utility is positive; individual attribute ratings increase with overall expected utility. As can be seen from the expected novelty ratings, an absence of new arguments leads to low overall utility expectations.

The attribute ratings also provide a strong linear predictor of initial expectation condition. The Wilk's lambda obtained from a multiple discriminant analysis using expectations on all four attributes as predictors is .55, significant at p < .000; while overall percent correct group classification is 68.3. Thus, the results from our field study appear highly consistent with results obtained in our laboratory studies and provide further encouraging support for the proposed multiattribute information utility construct.

Expectancy Disconfirmation and Information Utility

Our utilization model also suggests that, in addition to information characteristics, initial expectations will influence the direction of the effect of exposure to information on subsequent perceived utility ratings. According to hypotheses H 31 and H32, subjects in 2 lower-expected-utility condition will experience positive disconfirmation of expectations and hence revise utility ratings upwards following information use, while subjects in a high-expected-utility condition will experience negative disconfirmation and hence revise utility ratings downwards.

Examination of the mean shifts in overall utility ratings for forced-exposure information, pre- and post-exposure, support the disconfirmation hypothesis. For subjects in the low-expected-utility condition, the difference between expected- and post-exposure, perceived utility ratings (using a 100-point constant-sum scale in which an "average" information item would be allocated 20 points) averaged 6.01 , compared to -0. 32 for subjects in an indifferent-expected-utility condition, and -5.2 for subjects in a high-expected-utility condition (F4,56=11.03, p <.01).

An explanation of these shifts in overall utility is provided by the pattern of disconfirmation on individual attributes. On the relevance attribute, for example, the mean shift for subjects in the low-expected-utility condition was 89 (using a self-report, 6-point bi-polar scale), compared to -.36 for subjects in the high-expected-utility condition (F4,56 =5-4, p <.01). Thus, on average, subjects holding low initial expectations of the relevance of the information find those expectations are positively disconfirmed during the usage experience. Since group shifts on the remaining information attributes are not statistically different, this would imply that, for the current stimulus, the relevance attribute plays the determinant role in expectancy disconfirmation.

Utilization and Technology Evaluation

An important issue in utilization research is the impact of information on evaluation of the target product or technology. While it is often easy to demonstrate shifts in evaluation of the technology following information exposure, it is usually more difficult to explain why such shifts are occurring. In this section, we attempt to relate individual cognitive processes to characteristics of the information to which the individual has been exposed. Table 3 gives the standardized beta coefficients for each information attribute obtained by regressing the average post-exposure (perceived) attribute ratings against each of three cognitive measures, attitude towards the technology, confidence in performing the assigned task, and overall evaluation of the report. The attribute ratings are the within-subject averages across all information items examined only (i.e., items not examined have been excluded).

The coefficients are intuitively meaningful and add insight to the cognitive processes associated with information utilization. First, evaluations of the technology are positively influenced by high perceived relevance and credibility, but low novelty, ratings. This result is interesting since it implies that new arguments are not a necessary condition to favorable evaluation of the technology, and may, if encountered, have a negative effect on those evaluations.

Second, post-utilization task confidence judgments are not affected by judgments of the information's relevance or credibility. Comprehensive ease and familiarity with the arguments encountered, however, both contribute significantly to higher confidence levels. Thus, information would appear to be being used as a confirmatory test against existent expertise.

Third, evaluations of the overall usefulness of the report in the judgment task can be attributed to the more "affectively-weighted" attributes of information relevance and credibility. These attributes are "affectively-weighted" in the sense that they denote higher-level processing of an argument once it has been read and understood. Technical ease and the number of new arguments, on the other hand, can be considered more "cognitively-weighted" (i.e., ask the respondent to report only whether the argument was known and could be understood, and not whether the premise should be accepted or rejected in the judgment task). While understanding of the argument would seem to be a pre-condition to higher level processing, it does not (at least for the current experiment), appear to significantly influence judgments of the utility of the overall data-base.

Of course other factors will also affect these judgments. Among constructs included in our motel, the nature of the task assignment will directly affect the confidence level associated with task completion. Our laboratory results suggest that conceptual tasks (i.e., those associated with broader judgment parameters), for example, will show greater improvement in confidence levels as a result of exposure to information, than will structured choice dilemmas such as that associated with the instrumental task in the current study. Our field results generally support this finding. Subjects in the conceptual task condition report a mean improvement in confidence level of 1.27 (on a 6-point, bi-polar scale), compared to .51 for subjects in an instrumental condition. This difference is significant at p < .05.

TABLE 3

COGNITIVE RESPONSE AND INFORMATION UIILIZATION POST-UTILIZATION

DISCUSSION

In this paper, we have reported partial results from a field study of the utilization of objective information in judgment tasks. The purpose of the study has been to validate among a sample of potential adopters or a new technology, results obtained in a controlled laboratory experiment of information utilization.

The study lends strong support to the laboratory findings, and hence the proposed model. Both task and information characteristics significantly affect the formation of expected utility judgments. Consistent with our laboratory findings, objective information is expected to have lower utility in judgment tasks tied to a choice dilemma than in tasks requiring initial understanding of the choice parameters. In our earlier paper, we discuss the types of consumer judgments which might represent these two task conditions. Without repeating this discussion, our results suggest that the focus of consumer researchers on the role of information in the brand choice problem may overlook an important class of judgments in which consumers will be more receptive to information.

The field results also support the multi-attribute information utility construct proposed and tested in the laboratory environment. This result is important since it indicates that problems in measuring the effect of information characteristics on utility judgments, reported in earlier utilization studies in field settings (Deshpande and Zaltman 1982, 1984), can indeed be overcome.

A unique contribution of the current model is the testing of the expectancy disconfirmation paradigm as a model of the process of modification of utility judgments, following use of information. Again, our hypotheses concerning the effect of initial expectation on utility judgments are supported by the field results. This also is an important finding, since it implies that rejection of information following exposure is more likely to occur among audiences initially highly receptive to the stimulus. Since subjects holding low expectations are generally positively surprised by the stimulus, our data suggest that utilization of information may actually be enhanced by attempting to lower, rather than raise, initial expectations.

The study also demonstrates the separability of information characteristics to explain inherently different cognitive and affective judgments. For the current sample, "knowledge-related" attributes of technical ease and familiarity with arguments are being used to assess the confidence which should be assigned to the cask judgments, while "affectively-related" attributes of relevance and credibility provide input to evaluations of the overall utility of the arguments. For evaluations of the target technology, both sets of factors combine.

As objective (non-persuasive) information data-bases become increasingly available to consumers through innovations in communications technology, the importance of this class of information in consumer judgments is likely to increase. Such charges are likely to present marketing managers with new, and often particularly complex, communications challenges. In this paper, we take the position that increased understanding of the use of objective information products and services will provide marketing managers with insight not only into how to more effectively respond to these challenges, but also into the role of information generally in consumer judgments. We advocate increased efforts by consumer researchers in this area and present a reduced paradigm and set of findings to suggest how this effort might begin to be implemented.

REFERENCES

Berg, Mark.R., et alia (197&) actors Affecting the Utilization of Technology Assessment Studies in Policy-making, National Science Foundation report No. NSF/PRA-7616805/1.

Churchill, Gilbert A. and C Suprenant (1982) "An Investigation into the Determinants of Consumer Satisfaction", Journal of Marketing Research, 19, (November), 491-504.

Deshpande, Rohit and Gerald Zaltman (1982) "Factors Affecting the Use of Market Research Information", Journal of Marketing Res., 19, (February), 14-31.

Deshpande, Rohit, (1984) "A Comparison of Factors Affecting Researcher and Manager Perceptions of Market Research Use", Journal of Marketing Research, 21, (February), 32-38.

Gerstenfelder, Arthur & Paul Berger (1980) "An Analysis of Utilization Differences for Scientific and Technical Information", Management Science, 26, (February) 165-179.

John, George & John Martin (1984) "Effects of Organizational Structure and Marketing Planning on Credibility and Utilization of Plan Output", Journal of Marketing Research, 21, (May), 170-183.

Kehret-Ward, Trudy (1984) "How to Make a Claim Central:Using Syntax to Direct Cognitive Elaboration", Working Paper, University of California, Berkeley.

Oliver, Richard (1977) "A Theoretical Reinterpretation of Expectation and Disconfirmation Effects on Posterior Product Evaluations", in Consumer Satisfaction, Dissatisfaction and Complaining Behavior, R. Day and R. Hunt, eds, Indiana Univ., Bloomington, 2-9.

Oliver, Richard (1979) "Product Satisfaction as a Function of Prior Expectation and Subsequent Disconfirmation: new Evidence", in New Dimensions of Consumer Satisfaction and Complaining Behavior, R. Day and K. Hunt, eds, Indiana University, Bloomington, 66-71.

Sabatier, Paul (1978) "The Acquisition and Utilization of Technical Information by Administrative Agencies", Admin. Science Quarterly, 23, (September) 396-417.

Sternthal, Brian and Samuel Craig (1982) Consumer Behavior: An Information Processing PersPective, Prentice-Hall, Englewood Cliffs, N.J.

Sternthal, Brian, Ruby Dholakia and Clark Leavitt (1978) "The Persuasive Effects of Source Credibility: Tests of Cognitive Response", Journal of Consumer Res., 4, 252-260.

Wilton, Peter C. and John G. Myers (1984) "Response to Innovation: The Mediating Effects of Technology Assessment", Working Paper # M-2, Center for Research in Management, Univ. of Calif., Berkeley.

Wilton, Peter C. (1985) "Task, Expectancy and Information Assessment Effects in Information Utilization Processes", Journal of Consumer Research, forthcoming.

Wilton, Peter C. and David K. Tse (1983) "A Model of Consumer Response to Communication and Product in Advertising and Consumer Psychology, A. Woodside and L Percy, eds, Lexington Books, 315-332.

----------------------------------------