ایجاد واقعیت مشترک در مورد آزار و اذیت جنسی مبهم: نقش ابهام محرک در اثر تنظیم مخاطبان بر روی حافظه

Abstract By tuning messages about ambiguous information to their audience's attitude, communicators can reduce uncertainty and form audience-congruent memories. This effect has been conceptualized as the creation of shared reality with the audience. We applied this approach to representations of ambiguous antecedents of sexual harassment and examined whether the effect depends on the event's perceived ambiguity. Participants read a testimony about a supervisor's ambiguous behaviors toward a female employee and described the behaviors to an audience who had previously evaluated him positively or negatively. We manipulated perceived ambiguity of the testimony by including or omitting information about eventual, clear-cut harassment (known vs. unknown outcome). As predicted, participants aligned their messages and memory with their audience's evaluation only in the unknown-outcome condition, where epistemic uncertainty was higher. The findings highlight the role of epistemic needs in the communicative creation of a shared reality about a ubiquitous social situation with potentially harmful outcomes.

Introduction Social events, particularly people's behaviors, often allow different interpretations and are hence ambiguous. There are various ways in which perceivers can reduce such ambiguity and the concomitant uncertainty. A channel that has received little, but increasing attention is interpersonal communication (e.g., Berger and Calabrese, 1975, Echterhoff et al., 2009a and Higgins, 1992). One way in which interpersonal communication allows communicators to reduce uncertainty is audience tuning ( Higgins, 1992). Audience tuning occurs when communicators adapt their message to the audience's perspective or attitude regarding an event (e.g., Clark and Murphy, 1982 and Higgins, 1992). Audience tuning not only affects message formulation, but can also have consequences for communicators’ subsequent cognition, including memory for the originally encoded events or behaviors ( McCann & Higgins, 1992; for related accounts, see Chiu et al., 1998 and Marsh, 2007). For example, after communicators have tuned a message about a target person's ambiguous behaviors to an audience's (positive or negative) attitude toward the target, they often end up with memories of the target that are consistent with their audience-tuned message (Echterhoff et al., 2005 and Higgins and Rholes, 1978). To illustrate, an audience-tuning effect on memory occurs when a team member describes a newcomer's ambiguous behavior more positively to a team colleague who likes the newcomer, and later remembers the newcomer's initial behaviors more positively, consistent with the audience-tuned message. This audience-tuning effect on memory has been investigated within the saying-is-believing paradigm, where participants are given ambivalent behavioral information about a target person (Higgins & Rholes, 1978). They are asked to describe the target person to an audience who has already formed an impression about the target. Those who communicate with an audience who likes (vs. dislikes) the target person describe the target more positively. As a result, the valence of communicators’ memory of the original target material, assessed with a surprise free-recall task, matches the valence of their audience-tuned messages (for a review, see Echterhoff, Higgins, et al., 2009). Hence, both the communicators’ message and their memory are aligned with the audience's attitude. The effect can be regarded as communication-driven because it depends on the production of an audience-congruent message: the effect is not found when communicators, who know the audience's attitude, do not actually produce a message (Higgins and Rholes, 1978 and Higgins et al., 2007). Also, the effect of the audience's attitude on recall valence is often statistically mediated by message valence (McCann & Higgins, 1992). Thus, the effect is a case of conversational influence on communicators’ own memory (Hirst & Echterhoff, 2012). This audience-tuning effect on communicators’ memory has been conceptualized as the creation of shared reality (Echterhoff et al., 2009a and Hardin and Higgins, 1996). Shared reality is defined as the product of the motivated process of experiencing a commonality of inner states about the world (Echterhoff, Higgins, et al., 2009). According to shared-reality theory, the audience-tuning effect occurs to the extent that communicators are motivated to create a shared reality with their audience and succeed at creating that shared reality. The motivation for creating this commonality can be relational (Echterhoff et al., 2009b and Pierucci et al., 2013) and, more critical to our current analysis, epistemic. Epistemic motivation reflects the need to achieve a valid and reliable understanding of the world (Hardin & Higgins, 1996), to establish what is real (Higgins, 2012) and to reduce uncertainty (e.g., Kruglanski, 2004). The greater epistemic needs from ambiguous experiences, the more people seek a shared reality with others (Echterhoff et al., 2009a and Festinger, 1950). Saying-is-believing studies support the notion that the communicator's shared-reality motivation is critical. For instance, when audience tuning of the message is motivated by alternative, non-shared reality motives (e.g., obtaining a reward, or complying with blatant demands), communicators’ memory is not biased toward the audience's attitude (Echterhoff, Higgins, Kopietz, & Groll, 2008). Thus, audience tuning of the message will lead to audience-congruent memory only when audience tuning is sufficiently motivated by shared-reality concerns.2 In contrast, other research on communication effects on memory has centered on the cognitive processes during message production, such as selective rehearsal (Pasupathi, Stallworth, & Murdoch, 1998) or the formation, or activation, of a schema that continues to guide subsequent recall (Marsh, 2007). According to a shared-reality account, the epistemic needs driving shared-reality creation are elicited by the ambiguity of the stimulus material. However, this central assumption has never been tested in experimental research. The main purpose of the present research was to fill this gap. Specifically, shared-reality creation through audience tuning should be more likely under high (vs. low) epistemic uncertainty. So far, the epistemic-needs account has been examined in only one study (Kopietz, Hellmann, Higgins, & Echterhoff, 2010). In Experiment 2 by Kopietz et al. (2010), negative (vs. positive) ability feedback was used to manipulate high (vs. low) epistemic uncertainty in participants who were previously asked to form social judgments about characters depicted in ambiguous social interactions. It was found that only high-epistemic-uncertainty participants tuned their messages and memory to the audience's attitude. However, in this study, like in all extant saying-is-believing studies, the target description was carefully designed to be ambiguous, and the same version was used in all conditions. In our Experiment we manipulated perceived stimulus ambiguity for the first time to test its role in shared-reality creation through audience tuning.

. Results We report analyses of variance (ANOVAs) with partial eta squared (View the MathML sourceηp2) as an effect size measure for main and interaction effects and exact p values from two-tailed tests. 5.1. Manipulation checks The audience's attitude was perceived as more positive in the positive audience-attitude condition (M = 6.50, SD = 2.16) than in the negative audience-attitude condition (M = 4.03, SD = 2.17), F(1, 58) = 19.43, p < .001, View the MathML sourceηp2=.25. In the known-outcome condition, most participants correctly remembered receiving the sexual harassment outcome (93.3%), whereas the majority of participants in the unknown-outcome condition correctly remembered that they did not receive any outcome (76.7%). Moreover, as predicted, the measure of certainty (i.e., lower uncertainty) about the outcome of the scenario was higher in the known-outcome condition (M = 24.98, SD = 19.61) than in the unknown-outcome condition (M = 14.70, SD = 10.10), F(1, 58) = 6.52, p = .013, View the MathML sourceηp2=.10. Thus, participants in this latter condition appeared to be more uncertain about the supervisor's final behavior. 5.2. Message and recall valence Table 1 depicts the means for message valence (top panel) and recall valence (bottom panel). First, a 2 × 2 ANOVA yielded a main effect of audience attitude, F(1, 56) = 12.47, p = .001, View the MathML sourceηp2=.18. This effect was qualified by a significant interaction between audience attitude and outcome knowledge, F(1, 56) = 5.37, p = .024, View the MathML sourceηp2=.09. Follow-up analyses revealed a simple main effect of audience attitude in the unknown-outcome condition (for test statistics, see Table 1). Thus, participants in this condition adapted their message to their audience's attitude. However, in the known-outcome condition the effect was not statistically significant. The effect of outcome knowledge was non-significant, F < 1. Second, the evaluative tone of free recall was also biased in the direction of the audience's attitude, as revealed by a main effect of audience attitude, F(1, 55) = 4.39, p = .041, View the MathML sourceηp2=.07. The simple main effect of audience attitude was significant for participants without outcome knowledge, but not for participants in the known-outcome condition. No other effects reached significance, Fs < 1.14. 5.3. Accuracy of message and recall protocols We found no evidence that message and recall protocols contained more accurate reproductions in the known-outcome condition (M = 3.90, SD = 2.31 and M = 2.68, SD = 2.45) than in the unknown-outcome condition (M = 3.40, SD = 2.25 and M = 2.76, SD = 2.32), both Fs < 1, for message and recall, respectively. Thus, the reduced audience-attitude effects in the known-outcome condition cannot be explained by more accurate rehearsal or retrieval of the story elements. Note that our memory measure (recall valence) is designed to capture evaluative biases in free recall. Our measure thus differs from memory measures capturing the quantity of correctly remembered items ( Puff, 1982). The present divergence between the two measures is consistent with several audience-tuning studies ( Kopietz et al., 2009 and Kopietz et al., 2010), in which our valence measure and traditional, accuracy-oriented measures were dissociated. 5.4. Moderation of the audience attitude-message-recall path To establish the role of message production in the observed memory bias, we tested whether the observed effect of audience attitude on recall valence was mediated by message valence. Thus, we examined whether the present finding reflected a saying-is-believing effect, rather than merely a knowing the audience's attitude-is-believing effect. Importantly, according to our rationale the mediation should be weaker, or even absent, when the outcome is known. That is, the path from audience attitude via message valence to recall valence should be moderated by outcome knowledge ( Hayes, 2013). We tested such a moderated mediation (see Fig. 1) using the PROCESS macro for SPSS ( Hayes, 2013). Bootstrapping is considered to be the most valid and reliable method for assessing indirect effects ( Hayes, 2009). Because it does not make assumptions about parameter distributions ( Chernick, 2008), it produces robust estimates even with relatively small sample sizes. Moderated mediation analyses with audience attitude as IV, message valence as ... Fig. 1. Moderated mediation analyses with audience attitude as IV, message valence as mediator, recall valence as DV and outcome knowledge as the moderating variable. Note: The path coefficients are unstandardized regression coefficients. Coding: outcome knowledge (−1 = unknown, 1 = known); Audience Attitude (1 = positive, −1 = negative). *p < .05. Figure options We implemented Process model 7 by Hayes (2013) using 5000 bootstrap samples (see Fig. 1). Consistent with the earlier analysis of variance, the effect of audience on message valence (coded as −1 for negative attitude and 1 for positive attitude) was stronger when the outcome was unknown (b = .28, confidence interval from .14 to .41) than when it was known (b = .06, ns). Also, as predicted, the audience's attitude predicted message valence, which, in turn, predicted recall valence. However, the effect of the audience's attitude on message valence was moderated by outcome knowledge (dummy-coded as −1 = unknown, 1 = known). The indirect effect of audience’ attitude on recall valence through message valence was reliable only in the unknown-outcome condition, b (for the indirect effect) = .07, confidence interval from .01 to .14). In the known-outcome condition, this indirect effect was absent, b = .01, ns. Thus, this suggests that only in the unknown-outcome condition the production of an audience-consistent message influenced memory. While the obtained effects are not large, they are still consistent with the predicted role of outcome knowledge. 5.5. Audience trust Participants’ trust in their audience's attitude was positively correlated with both message bias and recall bias only in the unknown-outcome condition, r(28) = .41, p = .025, and r(28) = .36, p = .051, respectively. In the known-outcome condition, these correlations were not positive and nonsignificant, r(28) = −.20, and r(27) = −.27, ns. 8 The trust-message bias and trust-recall bias correlations (Fisher z-transformed) differed significantly between the outcome knowledge conditions, Z = 2.26, p = .023, and Z = 2.29, p = .022, respectively.