Consumer Inference

John Kim, Oakland University
Frank R. Kardes, University of Cincinnati
ABSTRACT - The importance of inference processing has been noted by many distinguished consumer researchers (e.g., Alba and Hutchinson 1987; Huber and McCann 1982; Johnson and Levin 1985; Kardes 1988; Meyer 1981; Sawyer 1988; Simmons and Lynch 1991; Yi 1990). The urgings of these researchers and the importance of this topic has led to an increase in research activities, and our understanding of the phenomenon. However, our understanding of inference processing is still limited. Thus, this session attempted to further stimulate inference research by focusing on recent empirical research on how consumers "go beyond the information given" to fill gaps in knowledge that inevitably follow from the reliance on incomplete information. Furthermore, the session endeavoured to facilitate the development of an integrative theory of consumer inference that can be useful in understanding consumer response to missing information.
[ to cite ]:
John Kim and Frank R. Kardes (1992) ,"Consumer Inference", in NA - Advances in Consumer Research Volume 19, eds. John F. Sherry, Jr. and Brian Sternthal, Provo, UT : Association for Consumer Research, Pages: 407-410.

Advances in Consumer Research Volume 19, 1992      Pages 407-410

CONSUMER INFERENCE

John Kim, Oakland University

Frank R. Kardes, University of Cincinnati

ABSTRACT -

The importance of inference processing has been noted by many distinguished consumer researchers (e.g., Alba and Hutchinson 1987; Huber and McCann 1982; Johnson and Levin 1985; Kardes 1988; Meyer 1981; Sawyer 1988; Simmons and Lynch 1991; Yi 1990). The urgings of these researchers and the importance of this topic has led to an increase in research activities, and our understanding of the phenomenon. However, our understanding of inference processing is still limited. Thus, this session attempted to further stimulate inference research by focusing on recent empirical research on how consumers "go beyond the information given" to fill gaps in knowledge that inevitably follow from the reliance on incomplete information. Furthermore, the session endeavoured to facilitate the development of an integrative theory of consumer inference that can be useful in understanding consumer response to missing information.

OVERVIEW

Consumers are often required to make judgments and purchase decisions about products on the basis of limited or incomplete information. Because complete information is rarely available to consumers, it is important to understand how consumers respond and adapt to incomplete information.

The four papers presented in the special topics session investigated consumer inference processes using several very different approaches: some have strategically deleted portions of the text of advertisements to examine the persuasive impact of inferred versus explicit claims; some have manipulated the contents of brand-by-attribute matrices presented on an information display board to examine the relationship between information search and decision making; and some have used Bayesian approaches to examine the diagnostic value of inferences. Although very different products, contexts, and methodologies have been applied, a unified theoretical framework is beginning to emerge that may be useful in addressing unresolved issues. Specifically:

1. All four papers emphasize that several different responses to missing information are possible. Consumers often fail to recognize that judgment-relevant information is missing. When omissions are overlooked, inferential activity is unlikely. Even when omissions are detected and when gap-filling inferences are likely to be generated, several different informational bases for inference making are often available and several different mechanisms for producing inferences can be activated.

2. All four papers adopt a contingency perspective. The results of Sawyer's experiment show that the likelihood of inference generation depends on the consumers' level of need for cognition and on the strength of the link between arguments and conclusions. Yi's study shows that attribute information is often open to multiple interpretations and that the accessibility of judgment-relevant information stored in memory influences which interpretation is likely to be inferred. Simmons and Lynch's experiment shows that inferences are likely to be formed only when several conditions are met: (1) the consumer must detect an omission, (2) an appropriate inference rule must be accessible from memory, and (3) the inference must be perceived to be useful for achieving the goals of the consumer. Kardes, Kim, and Lim show that prior knowledge produces multiple effects: knowledgeable consumers are more likely to generate inferences, and if inferences are generated, knowledgeable consumers are more likely to use their inferences to guide subsequent judgments and decisions.

3. The papers suggest consumer inference processes have important managerial implications across a wide variety of settings (advertising, product management, consumer decision making), attesting to the richness of the concept. Sawyer and Yi focus on the development of creative, new, and effective ad executions. Simmons, Lynch, Kardes, Kim, and Lim focus on the perceived diagnosticity of inferences and on the implications of perceived diagnosticity for consumer decision making. The perceived diagnosticity or importance of various attributes and benefits also has important implications for the design of products.

The session, thus, addressed some unresolved issues in consumer inference processing. Issues such as when inference are formed, what inferences are formed, and the effects of of these inferences on judgment and choice were discussed in the papers. The abstracts provided by the authors are presented next.

ABSTRACT - S

 

DETERMINANTS OF THE PERSUASIVENESS OF OPEN-ENDED VERSUS CLOSED-ENDED ADVERTISEMENTS

Alan G. Sawyer, University of Florida

Open-ended advertisements do not have an explicit conclusion that tells the audience what it should infer from the information in the ad. Alternatively, an ad can be closed-ended and specifically tie up loose ends and offer a message conclusion for an audience. There are several reasons why communication researchers predict that open-ended messages might be more persuasive than closed-ended ones (see Sawyer 1988 for a detailed review). Perhaps most importantly, the absence of any obvious conclusion may lead a motivated audience to try to infer one. In order to reach a conclusion, a motivated audience is more likely to process the central message arguments (Petty and Cacioppo 1986) of open-ended messages since they are unable to rely on the message conclusion available in a closed-ended message. An audience motivated to infer a conclusion is more likely to generate it its own thoughts and inferences when exposed to an open-ended ad than when exposed to a closed-ended ad (Kruglanski 1980). Attitudes resulting from effortful self-generated inferences should be more accessible (Kardes 1988) and more positive than attitudes resulting from less effortful processing of inferences explicitly provided in a message (Linder and Worchel 1970).

However, this potentially greater persuasiveness of a no conclusion message is accompanied by several risks. First, the audience may not process the central message arguments and, with no explicit conclusion to summarize the intended position of the message, be unable to infer a conclusion. Second, a message recipient might process the information but either draw no inference from that evidence or draw a different inference than intended. Therefore, it would seem important that the audience be sufficiently motivated to process the contents of an open-ended message in order to make a conclusion and that the message be structured so that the audience is able to make the conclusion intended by the message.

Sawyer and Howard (in press) confirmed the above speculation that the audience had to be involved in order for open-ended messages to be more persuasive than closed-ended ones. This result was replicated across several measures (including brand evaluation, purchase intention, and choice), two products, and two experiments and held after a one week delay as well as immediately after ad exposure. Although it is uncertain what exact process(es) account for these results, it seems that inferences play an important role. Kardes (1988) used a brand evaluation response latency procedure and found that involved subjects viewing an open-ended advertisement were more likely to have already formed an inference (consistent with faster responses) compared to involved subjects exposed to a closed-ended ad. However, unlike Sawyer and Howard's results, Kardes did not find that the open-ended ads were more persuasive; although the attitudes were more accessible, they were not significantly more positive.

This presentation will discuss the role of inferences in the persuasive process(es) evoked by open-ended advertisements and then present the results of a new experiment that tests for additional moderating variables. Two variables - one measured and one manipulated - are examined. The first is the subject variable of need for cognition. Subjects high in need for cognition might be more likely to infer a conclusion just as those who are more educated (Cooper and Dinerman 1951). Since the open-ended message requires more extensive and effortful processing, a closed-ended message may leave considerably less room for further elaboration. Need for cognition might interact with the type of message such that open-ended messages are especially effective for an involved, high need for cognition audience, whereas the effectiveness of closed-ended messages might not vary for involved audiences with varying levels of need for cognition.

A second variable is the obviousness of a conclusion. This variable might explain the absence of an effect on attitude valence in Kardes' study. The advertisement in Kardes' study frequently mentioned the brand name (nine times in the open-ended ad and six times in the closed-ended ad), contained a large picture of the product captioned with the brand name, and never mentioned any other brand. Thus, even a casual processing of either ad might have led to a realization of which brand was concluded by the ad to be the best. Although subjects in Kardes' two message conditions differed in whether they came to a conclusion about whether the advertised brand was superior on three attributes they may not have differed in whether they made a conclusion about brand preference. If both messages were equally likely to result in conclusion drawing about the brand, then no difference in persuasion should result. In Sawyer and Howard's experiments, an equal amount of information was given about each of four competing brands in a comparative ad format and the target brand was not highlighted in any way. Thus, a conclusion about which brand was best was not possible unless a subject processed at least some of the details of the ad information that would lead to that conclusion.

The present experiment manipulates the obviousness of the conclusion and crosses that variable with message type (open-ended vs. closed-ended) for involved subjects. The obviousness of the message conclusion is accomplished by adjusting the two messages in the Sawyer and Howard study to create brand name in the opening headline (e.g., "Edge wants you to discover the Difference" vs. "Discover the Difference"), label the pictured product with the brand name in the comparative ad information (vs. not highlighting) and include the brand name in the bottom copyright (vs. a non-linked parent name).

Paired comparisons and a test of the interaction between the obviousness of the sponsoring brand and whether a message is closed-ended or open-ended will be used to assess whether, as predicted, the non-obvious sponsor, open-ended ad is most persuasive and there is little or no persuasion difference among the other three ads - the obvious sponsor, open-ended ad and the two closed-ended ads. Other measures that attempt to assess subjects' perceptions of their effort in processing the ads' information and whether they made inferences will also be analyzed.

It is expected that this presentation will lead to a better understanding of the conditions conducive to effectiveness of an open-ended ad which depends on the audience making inferences on their own instead of merely processing the advertiser's conclusion in the ad. Together with the other papers in this session, this presentation should help the audience better understand the role of inferences in persuasion and information processing.

 

CONSUMER INFERENCE MAKING AND INDIRECT EFFECTS OF ADVERTISING

Youjae Yi, University of Michigan

Advertisements frequently emphasize salient attributes of products so that consumers' beliefs about these attributes will change. It is commonly accepted that changes in beliefs lead to changes in attitude. Much research has therefore focused upon changes in attacked beliefs and related these belief changes to attitude change. However, substantial evidence suggests that an ad can affect unattacked elements as well. Since a consumer may infer beliefs about aspects of a product that are not mentioned in the ad, ad messages about one attribute can indirectly influence beliefs about other product attributes. That is, an ad may have indirect effects on unmentioned beliefs as well as direct effects on target beliefs.

Furthermore, an ad often contains information that can be interpreted in several different ways. For example, when consumers hear that a car is large with luxury features, they might infer either that the car will provide good riding comfort or that the car will have low gas mileage. In such cases, the product evaluation will depend on which inference is drawn form the presented information. As a consequence, we should understand how and when these inferences are made in order to fully understand advertising effects.

This paper proposes that indirect effects of an ad on unmentioned beliefs are a function of attribute interrelationships and information accessibility. The proposed presentation will include the theoretical development extending research on information accessibility as well as the results of studies conducted to test the theoretical framework. This research should be of interest to researchers who study basic information processing, inference making, attitude change, and consumer responses to advertising. The research is also relevant to practitioners of advertising. By investigating the moderator variables of inference making, this research will provide insights into the situations when strong indirect effects of advertisements are expected.

 

CONTINGENT INFERENCE MAKING

Carolyn J. Simmons, University of Illinois

John G. Lynch, Jr., University of Florida

Recent evidence suggests that inferences about missing attributes are common only under certain conditions (Dick, Chakravarti, and Biehal 1990; Lim, Olshavsky, and Kim 1988; Simmons and Lynch 1991). Dick et al. have presented an integrative model that predicts that an inference about a missing attribute will be made under the following conditions: the missing attribute is noted, a relevant inference rule is accessible in memory, the inputs to that rule are also accessible, the inference is believed to be diagnostic, and alternative diagnostic inputs are not accessible or available externally. We manipulated factors that determined whether a missing attribute would be noted (competitive products did or did not provide information about the target missing attribute), the accessibility of an inference rule in memory (subjects were or were not reminded of a previous class exercise in which they were given the inference rule), and the relative diagnosticity of an inference for the task at hand (low diagnosticity = evaluation task or choice task in which neither alternative was described in terms of the target missing attribute, high diagnosticity = choice task in which one alternative was described in terms of the target missing attribute -- the assumption is that the desire for dimensional comparisons will make an inference highly diagnostic in this case). The inputs to the inference rule were always externally available. The results were generally consistent with the Dick et al. model. Inferences were unlikely unless the context drew attention to the target missing attribute. Increasing the accessibility of an inference rule increased the incident of inferences only when the inference was highly diagnostic.

 

CONSUMER EXPERTISE AND THE PERCEIVED DIAGNOSTICITY OF INFERENCES

Frank R. Kardes, University of Cincinnati

John Kim, Oakland University

Jeen-Su Lim, University of Toledo

Recent empirical evidence indicates that consumer expertise is an important determinant of the extent and the direction on inferential activity about products (Maheswarn and Sternthal 1990). Consumers who are knowledgeable about a particular product category possess the cognitive resources required to elaborate extensively on product-related information, and consequently, consumers are more likely to generate elaborative inferences spontaneously when level of prior knowledge is high (versus low). Given that inferences had been formed, consumers are more likely to trust their inferences and consider their inferences valid when prior knowledge is high. In contrast,when consumers consider themselves unknowledgeable about a particular product domain, they may distrust their inferences so much that they may consider explicitly-provided conclusions to be more valid than their own self-generated conclusions.

Subjects were exposed to four sets of arguments implying four separate conclusions about the benefits of a target product (a compact disc player). In spontaneous inference conditions, subjects were exposed to the arguments only (the conclusions were omitted). In prompted inference conditions, subjects were exposed to the arguments only and were asked to answer questions about the implied conclusions. In explicit conclusion conditions, the implied conclusions were presented explicitly. Subjects were blocked into high and low prior knowledge groups on the basis of their scores on a ten-item multiple-choice test about disc players.

For each implied benefit (conclusion), subjects estimated the percentage of high- versus low-quality CD players possessing the target benefit on scales from 0 (0 percent) to 10 (100 percent). Bayesian likelihood ratios were computed from these subjective probabilities following Herr, Kardes, and Kim (1991). The ratio provides a direct measure of perceived diagnosticity. The results revealed that low knowledge subjects perceived the target conclusions as nondiagnostic in spontaneous inference conditions, moderately diagnostic in prompted inference conditions, and highly diagnostic in explicit conclusion condition (although the latter effect was not statistically significant). For high knowledge conditions, the results were nonsignificant. However, the direction of the diagnosticity results were opposite of the low knowledge conditions. High knowledge subjects viewed the target conclusions as moderately diagnostic in explicit conclusion conditions, and highly diagnostic in spontaneous and prompted inference condition. This pattern suggests that low knowledge consumers are unlikely to generate inferences spontaneously, and if they are prompted to generate inferences, they regard their self-generated inference to be less diagnostic than explicit conclusions. High knowledge consumers, on the other hand, spontaneously generate inferences and tend to view these self-generated conclusions as more diagnostic than explicit conclusions.

REFERENCES

Alba, Joseph W. and J. Wesley Hutchinson (1987), "Dimensions of Consumer Expertise," Journal of Consumer Research, 13 (March), 411-454.

Cooper, Eunice and Helen Dinerman (1951), "Analysis of the Film 'Don't be a Sucker,' A Study of Communication," Public Opinion Quarterly, 15 (Summer), 243-264.

Dick, Alan, Dipankar Chakravarti, and Gabriel Biehal (1990), "Memory-Based Inferences During Consumer Choice," Journal of Consumer Research, 17 (June), 82-93.

Herr, Paul M., Frank R. Kardes, and John Kim (1991), "Effects of Word-of Mouth and Product-Attribute Information on Persuasion: An Accessibility-Diagnosticity Perspective," Journal of Consumer Research, 17 (March), 454-462.

Huber, Joel and John McCann (1982), "The Impact of Inferential Beliefs on Product Evaluations," Journal of Marketing Research, 14 (August), 324-333.

Johnson, Richard D. and Irwin P. Levin (1985), "More than Meets the Eye: The Effect of Missing Information on Purchase Evaluations," Journal of Consumer Research, 12 (September), 169-177.

Kardes, Frank R. (1988), "Spontaneous Inference Processes in Advertising: The Effects of Conclusion Omission and Involvement on Persuasion," Journal of Consumer Research, 15 (September), 225-233.

Kruglanski, Arie W. (1980), "Lay Epistemological Process and Contents: Another Look at Attribution Theory," Psychological Review, 87 (January), 70-87.

Lim, Jeen-Su, Richard W. Olshavsky, and John Kim (1988), "The Impact of Inferences on Product Evaluations: Replication and Extension," Journal of Marketing Research, 25 (August), 308-316.

Linder, Darwyn E. and Stephen Worchel (1970), "Opinion Changes as a Result of Effortfully Drawing a Counterattitudinal Conclusion," Journal of Experimental Social Psychology, 6 (October), 432-448.

Maheswaran, Durairaj and Brian Sternthal (1990), "The Effects of Knowledge, Motivation, and Type of Message on Ad Processing and Product Judgments," Journal of Consumer Research, 17 (June), 66-73.

Meyer, Robert J. (1981), "A Model of Multiattribute Judgements Under Attribute Uncertainty and Informational Constraints," Journal of Marketing Research, 18 (November), 428-441.

Petty, Richard E. and John T. Cacioppo (1986), "The Elaboration Likelihood Model of Persuasion," in Advances in Experimental Social Psychology, Vol. 19, ed. Leonard Berkowitz, New York: Academic Press, 123-205.

Sawyer, Alan G. (1988), "Can There Be Effective Advertising Without Explicit Conclusions? Decide for Yourself," in Nonverbal Communication in Advertising, eds. Sidney Hecker and Daviv W. Stewart, Lexington, MA: Lexington, 159-184.

Sawyer, Alan G. and Danial J. Howard (forthcoming), "The Effects of Omitting Conclusions in Advertisements to Involved and Uninvolved Audiences," Journal of Marketing Research, in press.

Simmons, Carolyn J. and John G. Lynch, Jr. (1991), "Inference Effects Without Inference Making? Effects of Missing Information on Discounting and Use of Presented Information," Journal of Consumer Research, 17 (March), 477-491.

Yi, Youjae (1990), "The Effects of Contextual Priming in Print Advertisements," Journal of Consumer Research, 17 (September), 215-222.

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