The Role of Conversational Norms and Sensitivity to Omissions in Judgment Based on Limited Evidence

ABSTRACT - When judging an object described by limited evidence, people often make judgments based on the evaluative implications of what is known and fail to adjust for what is unknown. Consequently, people tend to form extreme and confident judgments of an object even when little information is provided. The present study investigates the extent to which the tendency to form strong judgments on the basis of weak evidence stems from conversational inferences and assumptions about the intent of the communicator or from insensitivity to the limitations of the presented evidence. Participants received a brief description of a target object provided by a credible or a noncredible communicator. Attributions for omissions (missing information) were assessed either before or after assessing overall evaluations of the target object. The least favorable attributions and the least favorable evaluations were formed toward a target object described by a non-credible communicator, but only when evaluations were elicited after attributions. In the remaining conditions, participants exhibited more sensitivity to the credibility of the communicator and to the limitations of the evidence. Implications of the results for understanding judgment based on limited evidence are discussed.

Citation:

David C. Houghton, Frank R. Kardes, and David M. Sanbonmatsu (1998) ,"The Role of Conversational Norms and Sensitivity to Omissions in Judgment Based on Limited Evidence", in NA - Advances in Consumer Research Volume 25, eds. Joseph W. Alba & J. Wesley Hutchinson, Provo, UT : Association for Consumer Research, Pages: 146-150.

Advances in Consumer Research Volume 25, 1998      Pages 146-150

THE ROLE OF CONVERSATIONAL NORMS AND SENSITIVITY TO OMISSIONS IN JUDGMENT BASED ON LIMITED EVIDENCE

David C. Houghton, Northwest Nazarene College

Frank R. Kardes, University of Cincinnati

David M. Sanbonmatsu, University of Utah

Edward A. Ho, University of Utah

Steven S. Posavac, University of Utah

ABSTRACT -

When judging an object described by limited evidence, people often make judgments based on the evaluative implications of what is known and fail to adjust for what is unknown. Consequently, people tend to form extreme and confident judgments of an object even when little information is provided. The present study investigates the extent to which the tendency to form strong judgments on the basis of weak evidence stems from conversational inferences and assumptions about the intent of the communicator or from insensitivity to the limitations of the presented evidence. Participants received a brief description of a target object provided by a credible or a noncredible communicator. Attributions for omissions (missing information) were assessed either before or after assessing overall evaluations of the target object. The least favorable attributions and the least favorable evaluations were formed toward a target object described by a non-credible communicator, but only when evaluations were elicited after attributions. In the remaining conditions, participants exhibited more sensitivity to the credibility of the communicator and to the limitations of the evidence. Implications of the results for understanding judgment based on limited evidence are discussed.

INTRODUCTION

People must frequently rely on incomplete, fragmentary, or limited evidence when forming judgments of objects (e.g., persons, places, things) or issues (e.g., social, political, economic topics) encountered in everyday life. In many instances, social communications, advertisements, and mass media messages tend to be remarkably brief and uninformative. Nevertheless, people seem to try to make sense of whatever evidence is presented to them, and, in some cases, people seem to be willing to draw judgments and conclusions on the basis of minimal information that is of limited or questionable probative value. The reasons for this apparent willingness to judge objects using limited evidence is unclear.

Recent research suggests that, in some instances, people may be insensitive to the limitations of the given evidence. In one relevant study, participants differing in prior knowledge about cameras received either a large or a small amount of attribute information about a target camera (Sanbonmatsu, Kardes and Herr 1992). Most participants formed extreme and confident evaluations of the target regardless of how much or how little information was given. However, participants who were highly knowledgeable about cameras formed moderate (less extreme) judgments of the target when little information was provided. Highly knowledgeable individuals appear to be sensitive to omissions (missing or unknown information) and to limitations of evidence.

Inferential judgments that "go beyond the information given" (Bruner 1957) also appear to be contingent on sensitivity (or insensitivity) to evidential limitations. In one study, participants differing in prior knowledge about bicycles were given a brief description of a target bicycle containing no information about its durability (Sanbonmatsu, Kardes and Sansone 1991). Participants were asked to judge the durability of the bicycle either immediately after the description was presented or one week later. Participants formed moderate inferences about durability immediately after stimulus presentation because it was readily apparent that no direct evidence bearing on durability was available. In delay conditions, however, sensitivity to limitations of evidence should decrease (due to forgetting) and inference extremity should increase. This result was obtained for unknowledgeable and moderately knowledgeable participants. Highly knowledgeable individuals, however-who were presumably sensitive to limitations of evidence across conditions-formed moderate inferences in delay as well as in no-delay conditions.

Other variables, in addition to prior knowledge levels and the timing of judgment (i.e., delay or no delay following stimulus exposure), may also influence sensitivity to the limitations of evidence. Contextual variables may sensitize people to the limited strength or weight of the given information. Consistent with this hypothesis, recent research shows that sensitivity to limitations of evidence and judgment adjustment toward a moderate position increases when a target object is judged in the context of a similar object described on attribute dimensions different from those used to describe the target, or when a target object is judged in the context of a completely different type of object described by a relatively large amount of information (Sanbonmatsu, Kardes, Posavac and Houghton 1997). These results suggest that the detection of specific omissions or the awareness of a general lack of information without the detection of specific omissions, respectively, may increase sensitivity to limitations of evidence.

Another important contextual variable that may influence judgment based on limited evidence is the social context in which information is communicated. Recent research on Grice’s (1975) cooperation principle and maxims of conversation suggests that interpersonal communication is guided by conversational norms and by inferences and assumptions based on these norms (Hilton 1990, 1995; Schwarz 1994; Schwarz, Strack, Hilton nd Naderer 1991; Wyer and Gruenfeld 1995). These norms proscribe communicators to cooperate with recipients by providing information that is relevant, truthful, informative, and clear (the maxims of relation, quality, quantity, and manner, respectively). Correspondingly, recipients should assume that communicators are trying to be relevant, truthful, informative, and clear, and these assumptions should have a strong impact on judgment. Hence, recipients may weigh presented information highly, not because the information itself seems important and relevant, but because recipients infer that communicators try to provide information that is important and relevant.

Recipients should also make assumptions about omissions or about information that communicators fail to mention. Cooperative communicators should provide no more information than is needed to satisfy interactional goals. Cooperative communicators should also deliberately omit or fail to mention irrelevant information because irrelevant information may potentially distract or confuse recipients. Hence, the tendency of recipients to perceive presented information as important and to perceive omitted information as unimportant may stem not from insensitivity to the limitations of the given evidence, but from conversational inferences and assumptions about the intentions and actions of communicators.

Grice’s (1975) cooperation principle presupposes that communicators construct messages intentionally, and that communicators wish to and have the ability to cooperate with recipients. However, communicators differ with respect to their communicative intentions, wishes, and abilities. Consequently, recipients need to be sensitive to these differences. Politicians and advertisers do not always wish to cooperate by providing complete and accurate information. Unknowledgeable or ineffective communicators may wish to cooperate, but may lack the issue- relevant knowledge or communicative skills needed to provide clear and accurate information. Consequently, recipients should be sensitive to communicator credibility cues and should adjust their judgments accordingly.

When investigating attributional processes in conversational inference, it is important to recognize that attributions are not generated spontaneously in all instances (Hastie 1984; Weiner 1985). Sometimes attributions are formed spontaneously or on-line as information is encountered (Hastie and Park 1986), and sometimes attributions are measurement induced (Feldman and Lynch 1988). Exposure to attributional measures may prompt respondents to engage in attributional reasoning processes that might not have occurred otherwise (i.e., in the absence of measurement). One useful procedure for determining whether attributions were measurement induced or not is to vary the order in which attributions (the presumed mediator) and overall evaluations (the dependent variable) are measured (Feldman and Lynch 1988). Relative to spontaneously-formed attributions, measurement-induced attributions are far more sensitive to measurement order. Specifically, measurement-induced attributions should influence evaluations only when attributions are measured prior to assessing evaluations. Otherwise, measurement-induced attributions have not yet been formed and therefore cannot influence evaluations. Spontaneously-formed attributions, on the other hand, are formed independently of measurement and should influence evaluations regardless of measurement order.

To summarize, the apparent willingness of people to form judgments based on limited information may occur for one of two very different reasons: (a) people may be insensitive to omissions and to limitations of evidence, or (b) people may discount omissions due to conversational inferences that imply that the communicator should omit irrelevant details to facilitate cooperative communication. The present experiment investigates these possibilities by manipulating the credibility of the communicator of a relatively brief message, and by manipulating the order in which attributions for omissions and overall evaluations of the described target object are measured. The insensitivity to limitations of evidenc hypothesis suggests that participants should form extreme evaluations of a briefly described object except when the presence of attributional measures prompts participants to consider the credibility of a noncredible communicator very carefully. Moreover, the joint influence of communicator credibility and measurement order on evaluations should be mediated by the perceived limitations of the evidence. By contrast, the conversational inference hypothesis suggests that a main effect for communicator credibility should be observed: more extreme evaluations should be formed toward the object described by the credible communicator than by the noncredible communicator. Moreover, attributions for omissions should mediate the effect of communicator credibility on evaluations.

METHOD

Participants

One hundred and twenty-six University of Cincinnati undergraduates participated to obtain extra course credit in an introductory marketing course. Participants were randomly assigned to experimental conditions.

Procedure

The study was described as an investigation of product perceptions. Participants read a description of a new automobile model referred to as "Model A." Favorable information about the product’s repair record, acceleration, ride, anti-lock brakes, warranty, price, and standard features (A.M./F.M. tape player, anti-lock brakes, air conditioning) was presented. However, information about several attribute dimensions-including exterior styling, steering and handling, and gas mileage-were omitted. The description was held constant across all experimental conditions.

Communicator Credibility. Before participants received the product description they were told that this information was transcribed exactly from an article published in Consumer Reports (high communicator credibility) or from an advertisement placed in a local newspaper, the Cincinnati Post (low communicator credibility). To bolster the communicator credibility manipulation, in high credibility conditions, participants were told that Consumer Reports accepts no advertising and is published by the Consumers’ Union, a nonprofit organization serving only consumers. In low credibility conditions, participants were told that the advertisement was sponsored by a local car dealership. In addition, the message source was printed in large, bold, underlined letters placed above the description. The volume, date, and page number of the article/ advertisement was printed below the description.

Order of Measurement. Half of the participants received the attribution scales (and the inference and attribute rating scales) before receiving the overall evaluation scales. The remaining participants received the overall evaluation scales first.

Overall Evaluation. Participants evaluated the target product on a 9-point scale anchored by 1=one of the worst and 9=one of the best relative to other compact cars in the same price range.

Attributions for Omissions. Participants indicated their attributions for omissions on four 9-point scales anchored by 1=totally disagree and 9=totally agree. Participants were told that the article/advertisement provided information about the styling, power, air conditioning, radio, and price of the product, but provided no information about the gas mileage, exterior styling, and other attributes. Participants were asked to indicate the extent to which they agree or disagree with each of four possible explanations for the omissions: 1) the communicator wanted to mention only the most important attributes of the "Model A" (Important), 2) there was simply not enough space to mention other attributes of the "Model A"(Space), 3) the communicator wanted to catch the attention of the readers by mentioning a few positive attributes of the "Model A" (Attention), and 4) the communicator wanted to discuss the most positive attributes of the "Model A"; the article/advertisement did not mention the exterior styling and other important attributes because the "Model A" is below average on these dimensions (Favorable). Attribution scales having negative evaluative implications (3 and 4 above) were reverse scored.

Ratings of Presented Attributes. Participants were told that the article/advertisement provided information about the exterior styling, steering and handling, and gas mileage of the "Model A," and participants were asked to rate the "Model A’s" performance on each of these dimensions relative to other compact models in the same price range on 7-point scales anchored by 1=far below average and 7=far above average.

TABLE 1

EVALUATIONS, ATTRIBUTIONS, INFERENCES, ATTRIBUTE RATINGS, AND PERCEIVED SUFFICIENCY JUDGMENTS AS A FUNCTION OF COMMUNICATOR CREDIBILITY AND ORDER OF MEASUREMENT

Inferred Values for Missing Attributes. Participants were told that the article/advertisement provided no information about the power and the warranty of the "Model A," and participants were asked to rate the "Model A’s" performance on these dimensions relative to other compact models in the same price range on 7-point scales anchored by 1=far below average and 7=far above average.

Perceived Sufficiency Judgments. Participants were asked "to what extent was the information that was provided sufficient to evaluate the 'Model A’ automobile? That is, to what extent were you given enough information to evaluate the 'Model A’ automobile?" Participants were asked to indicate their responses on a 9-point scale anchored by 1=not at all enough and 9=highly sufficient.

RESULTS

Overall Evaluations

Mean overall evaluations as a function of communicator credibility and order of measurement are presented in row 1 of Table 1. Lower scores on this index indicate less favorable evaluations of the target object. As expected, contrast analysis (Rosenthal and Rosnow, 1985) showed that the least favorable evaluations were formed in the noncredible communicator/attributions measured after evaluations cell than in the remaining three conditions, F(1, 122)=51.36, p<.0001. This pattern suggests that measuring attributions for omissions prior to measuring evaluations increases sensitivity to conversational norms. Uncooperative noncredible communicators may try to hide negative aspects of a product they are trying to sell by omitting important information concerning these dimensions. Conversely, recipients assume that cooperative credible communicators present important information and omit unimportant information. Consequently, conversational inference processes, when they occur, encourage recipients to form less favorable attributions for omissions and less favorable overall evaluations when the communicator is low versus high in credibility.

Attributions, Inferences, Attribute Ratings, and Perceived Sufficiency Judgments

Mean scores on the attribution measures as a function of communicator credibility and order of measurement are presented in rows 2 through 5 of Table 1. Lower scores indicate less favorable attributions for omissions. Contrast analyses showed, as predicted, that the least favorable attributions for omissions were formed in the noncredible communicator/evaluations measured last cells than in the remaining three conditions, for attribution 1 (important, t=2.69, p<.005), 3 (attention, t=1.62, p<.10), and 4 (favorable, t=2.84, p<.005). The test for attribution 2 (space) was nonsignificant (t<1). The observed order of measurement effect suggests that attribuions for omissions were not formed spontaneously while participants processed the target information. Instead, exposure to attribution scales prompted participants to consider attributional implications that would not have been considered otherwise.

Mean inferred values for missing attributes are presented in rows 6 and 7 of Table 1. Contrast analyses showed no significant effects for communicator credibility or order of judgment elicitations.

Mean ratings of presented attributes are presented in rows 7 through 9 of Table 1. Contrast analyses showed that the least favorable ratings were formed in the noncredible communicator/evaluations measured last cell than in the remaining three conditions for style (t=3.46, p<.001), power (t=4.12, p<.001) and handling (t=3.58, p<.001)

Mean judgments of the perceived sufficiency of the presented information for evaluating the target object are presented in row 11 of Table 1. Planned contrasts revealed a significant main effect for order and a significant order by communicator credibility interaction. The presented information was perceived as less adequate when attributions were measured after evaluations than when evaluations were measured after attributions (t=2.03, p<.025). This pattern suggests that recipients are generally insensitive to the sufficiency or adequacy of the presented information. However, exposure to attribution scales prompts recipients to consider the limitations of the presented evidence more carefully. Also as predicted, the information was judged to be less sufficient in the noncredible communicator/evaluations measured last cell than in the remaining three conditions (t=3.36, p<.001).

TABLE 2

MEDIATION ANALYSES

Mediation Analyses

A series of regression models were estimated using the Baron and Kenny (1986) approach for testing mediation. In step 1, the mediators were regressed on the independent variable. In step 2, the independent variable was regressed on the dependent variable. In step 3, the independent variable and the mediators were regressed on the dependent variable. The results of these analyses are presented in Table 2. The independent variable was coded as a dummy variable: 1=noncredible communicator/attributions measured before evaluations; 2=the remaining cells. As Table 2 indicates, the contrast dummy variable influenced inferences, attribute ratings, and perceived sufficiency. Moreover, the contrast influenced evaluations (b=-1.389, p<.0001), and this effect was reduced (to b=-.727 , p<.0001) when the mediators were included in the regression equation. Together, these results suggest that inferences, attributions, and perceived sufficiency may mediate the effect of communicator credibility and order of measurement on evaluations.

DISCUSSION

The present research sought to determine whether people’s willingness to form judgments based on limited evidence stems from conversational inference processes that encourage recipients to discount omissions when the communicator is credible or from insensitivity to omissions and to the limitations of the presented information. The results provide little evidence of omission discounting and strong support for the insensitivity to omissions hypothesis. Discounting requires recipients to be sensitive to the credibility of the communicator regardless of the order in which evaluations and attributions are measured. By contrast, the results revealed weak effects of communicator credibility except when attributions were elicited prior to evaluations. This order of measurement effect suggests that attributions for omissions were not formed spontaneously prior to measurement. Instead, exposure to attributional measures was necessary to induce recipients to consider why the communicator failed to provide infrmation bearing on several specific aspects of the target object. Moreover, attributions for omissions failed to mediate the joint effect of communicator credibility and order of measurement on overall evaluations.

The effects of order of measurement observed in the present experiment also suggest that recipients tend to neglect omissions routinely and spontaneously. Only when recipients were exposed to attributional scales that required consideration of omissions and reasons for omissions did they demonstrate awareness of the limitations of the presented information. This cognizance was manifested in terms of less favorable attributional judgments, less favorable inferences about missing attributes, lower ratings of the perceived sufficiency of the presented information, and less favorable overall evaluations when participants were sensitized to omissions due to exposure to attributional measures. Moreover, the joint effect of communicator credibility and order of measurement on overall evaluations was mediated by inferential judgments and by perceived sufficiency judgments. Consistent with the results of prior research (Kardes and Gurumurthy 1992; Kardes and Sanbonmatsu 1993; Sanbonmatsu et al., 1991, 1992, 1994, 1996), sensitivity (versus insensitivity) to the limitations of the presented evidence encouraged recipients to form moderate (less extreme) judgments of the target object, even when the content of the communication was held constant. Moderate evaluations are frequently appropriate under conditions of limited information because moderate positions are easily justified to oneself and to others (Tetlock 1992) and are readily updated in light of new evidence (Cialdini, Levy, Herman and Evenbeck 1973; Jaccard and Wood 1988).

It is important to emphasize that the results of the present experiment do not undermine the importance of conversational norms or of recipients’ inferences about the intentions and actions of communicators. Ironically, the high degree of insensitivity to limitations of evidence observed in the present study may have blocked or prevented the occurrence of Gricean attributional reasoning processes. Conversational inferences may well have occurred if recipients had been more sensitive to omissions. Indeed, when exposure to attributional measures forced recipients to focus on omissions, less favorable attributions about the behavior of the communicator and less favorable evaluations of the target object were found.

REFERENCES

Baron, Reuben M. and David A. Kenny (1986), "The Moderator-Mediator Variable Distinction in Social Psychological Research: Conceptual, Strategic, and Statistical Considerations," Journal of Personality and Social Psychology, 51, 1173-1182.

Cialdini, Robert B., Alan Levy, C. Peter Herman and Scott Evenbeck (1973), "Attitudinal Politics: The Strategy of Moderation," Journal of Personality and Social Psychology, 25, 100-108.

Feldman, Jack M. and John G. Lynch (1988), "Self-Generated Validity and Other Effects of Measurement on Belief, Attitude, Intention, and Behavior," Journal of Applied Psychology, 73, 421-435.

Grice, H. Paul (1975), "Logic and Conversation," In Syntax and Semantics 3: Speech Acts (pp. 95-113), (Eds.) P. Cole and J. L. Morgan, New York: Academic Press.

Hastie, Reid (1984), "Causes and Effects of Causal Attribution," Journal of Personality and Social Psychology, 46, 44-56.

Hastie, Reid, and Bernadette Park (1986), "The Relationship Between Memory and Judgment Depends on Whether the Judgment Task is Memory-Based or On-line," Psychological Review, 93, 258-268.

h Denis J. (1990), "Conversational Processes and Causal Explanation," Psychological Bulletin, 107, 65-81.

Hilton, Denis J. (1995), "The Social Context of Reasoning: Conversational Inference and Rational Judgment," Psychological Bulletin, 118, 248-271.

Jaccard, James and Gregory Wood (1988), "The Effects of Incomplete Information on the Formation of Attitudes Toward Behavioral Alternatives," Journal of Personality and Social Psychology, 54, 580-591.

Kardes, Frank. R. and Kalyanaram Gurumurthy (1992), "Order-of-Entry Effects on Consumer Memory and Judgment: An Information Integration Perspective," Journal of Marketing Research, 29, 343-357.

Kardes, Frank. R. and David M. Sanbonmatsu (1993), "Direction of Comparison, Expected Feature Correlation, and the Set-Size Effect in Preference Judgment," Journal of Consumer Psychology, 2, 39-54.

Rosenthal, Robert and Ralph L. Rosnow (1985), Contrast Analysis: Focused Comparisons in the Analysis of Variance. Cambridge, England: Cambridge University Press.

Sanbonmatsu, David. M., Frank R. Kardes and Paul M. Herr (1992), "The Role of Prior Knowledge and Missing Information in Multivariate Evaluation," Organizational Behavior and Human Decision Processes, 51, 76-91.

Sanbonmatsu, David. M., Frank. R. Kardes and Carol Sansone (1991), "Remembering Less and Inferring More: The Effects of the Timing of Judgment on Inferences about Unknown Attributes," Journal of Personality and Social Psychology, 61, 546-554.

Sanbonmatsu, David. M., Frank R. Kardes, Steven S. Posavac, and David C. Houghton (1997), "Contextual Influences on Judgment Based on Limited Information," Organizational Behavior and Human Decision Processes, 69, 251-264.

Sanbonmatsu, David. M., Sharon Shavitt and Bryan D. Gibson (1994), "Salience, Set Size, and Illusory Correlation: Making Moderate Assumptions about Extreme Targets," Journal of Personality and Social Psychology, 66, 1020-1033.

Schwarz, Norbert (1994), "Judgment in a Social Context: Biases, Shortcomings, and the Logic of Conversation," In Advances in Experimental Social Psychology (Vol. 26, pp. 123-162), ed. M. P. Zanna, San Diego, CA: Academic Press.

Schwarz, Norbert, Fritz Strack, Denis J. Hilton and Gabi Naderer (1991), "Base Rates, Representativeness, and the Logic of Conversation: The Contextual Relevance of 'Irrelevant’ Information," Social Cognition, 9, 67-84.

Weiner, Bernard (1985), "'Spontaneous’ Causal Thinking," Psychological Bulletin, 97, 74-84.

Wyer, Robert. S., Jr., and Deborah H. Gruenfeld (1995), "Information Processing in Social Contexts: Implications for Social Memory and Judgment," In Advances in Experimental Social Psychology (Vol. 27, pp. 49-91), ed. M. P. Zanna, San Diego, CA: Academic Press.

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