Reading to make a decision or to reduce cognitive dissonance? The effect of selecting and reading online reviews from a post-decision context

This research challenges the presumption that reading online reviews solely affect reader attitudes toward a prospective decision. Instead, readers may strategically select and read reviews after a decision. This research advances and tests hypotheses proposing that: (1) post-decision readers select decisionsupportive reviews earlier and more frequently; and (2) the reviews they read affect the subsequent cognitive dissonance they experienced. Two studies employed an original post-decision experiment and demonstrated that readers expressed biased review selection. Interestingly, the reviews they read affected and increased their cognitive dissonance. The findings contributed to understanding the complex dynamic of online reviews in a post-decision context. © 2016 Elsevier Ltd. All rights reserved. Online reviews (e.g., customer reviews on Amazon.com) continue to be an important and popular source of information (Dellarocas, 2003; Dellarocas, Zhang,& Awad, 2007). These reviews originate from other users; as a result, products retailers undoubtedly utilize their influences to facilitate purchase behavior (Purnawirawan, Eisend, Pelsmacker, & Dens, 2015). Readers often perceive online reviews as more credible and more trustworthy than traditional advertising (e.g., Huang, Chou, & Lan, 2007). Based on this advantage, prior research related to online reviews has largely attended to the effect of reading reviews on reader attitudes (e.g., Doh & Hwang, 2009; Kim & Hollingshead, 2015; Lee, Rodgers, & Kim, 2009). These attitudes serve as conceptual antecedents toward some decisional and behavioral outcomes of interest (e.g., purchasing a laptop computer; choosing to use e-cigarettes). The decisionor outcome-oriented perspective is especially evident in recent review articles and summaries (e.g., Kim & Hollingshead, 2015; Walther & Jang, 2012). These authors outlined various sources of persuasive influence on participatory, or Web 2.0, websites and their associated effects on reader attitudes. Under this participatory view, the extent to which the review valence, or the extent to which reviews are positive or negative, serve as a generalized source of influence that directionally and aggregately correspond to influence reader attitudes (e.g., Purnawirawan et al., 2015; Walther, Liang, Ganster, Wohn, & Emington, 2012). Importantly, even after making a decision (e.g., product purchase), readers can still select and read online reviews. This postdecision selection and reading behavior theoretically serves a different motivation than the motivation for making an accurate decision (Festinger, 1957; Fischer & Greitemeyer, 2010). More important, this motivation activates specific selection biases and associated effects. The current research advances and tests such a post-decision dynamic to examine and explore the alternative effect of online reviews on readers. The current research focuses this post-decision dynamic at a theoretical and practical level by reporting on a series of systematic research. This research starts by providing empirical evidence establishing that readers do frequently utilize and read reviews after a decision. Next, this research examines how post-decision readers select reviews and the associated effect in a controlled laboratory design. Finally, this research provides a replication of the findings in an online environment. The following sections provide rationale and justify the relevance for investigating online reviews from a post-decision context. 1. Online reviews: The importance of a post-decision perspective On participatory websites, readers often first confront the task Y. Liang / Computers in Human Behavior 64 (2016) 463e471 464 of selecting which reviews to read from the plethora of reviews available online. Popular retail websites, such as Amazon.com, have amassed a large number of product reviews. Otterbacher (2009) reported an average of 340 reviews posted for products randomly sampled on Amazon.com. Ghose and Ipeirotis (2007) also reported a high number of reviews per product (M 1⁄4 138), 1339 reviews being at the high end for a single product. It is not surprising that a PEW Internet Survey found that 30% of internet users reported feeling overwhelmed by the amount of information they found while shopping or researching online (Horrigan, 2008). Given the abundance of available reviews, readers' motivation likely guide how their selection and subsequent reading behaviors. Readermotivation differs based onwhether a decision has taken place (i.e., pre-versus post-decision). Before a decision (e.g., product purchase), readers may seek information to make an accurate decision (Fischer & Greitemeyer, 2010). After a decision, readers may seek information to justify the decision they made (Cummings & Venkatesan, 1976; Fischer & Greitemeyer, 2010). This motivational difference leads to a discrepancy in how readers both select and regard the information they receive (Fischer & Greitemeyer, 2010; Hart et al., 2009). Examining reviews from this post-decision perspective offers vital and unique contributions to understanding how participatory websites affect readers. Participatory websites (Walther & Jang, 2012) are distinct communication environments where multiple sources of influences and cues (e.g., star ratings, helpfulness ratings, number of comments, likes) exist simultaneously. These cues may facilitate the selection process by serving as sorting cues that best address the readers’ motivation. For example, readers may select reviews they perceive as higher in quality prior to a decision; after making the decision, they may select reviews that support their decision to reduce dissonance. Specifically, a decision activates a motivational mechanism to support the original decision. This activation suggests that there are characteristics of selected reviews that systematically attract attention and selection. This postdecision motivation also affects outcomes other than attitude (e.g., increasing or reducing cognitive dissonance). In a postdecision context, decision accuracy is not the primary motivation for information-seeking. Addressing this argument may yield corresponding data that provide boundary conditions for the existing research on the effect of online reviews. In addition, it empirically demonstrates how reviews serve a differential purpose after a decision. 2. Participatory websites and aggregated userrepresentations The motivational effect may also alter how readers regard different sources of persuasive influence on participatory websites. There are multiple sources, and each source specifically references the type of cues and signals left by users (Walther & Jang, 2012). One such cue, star ratings, conveys information regard the contents of a particular review and allows readers to select reviews without fully reading the textual content. Prior to making a decision, the cue conveys specific valence characteristic regarding the review. For example, star ratings conveys the extent to which a review is positive or negative toward a product. However, the product decision alters the manner to which this cue may be interpreted as decision-supportive (i.e., congenial) or decision-unsupportive (i.e., uncongenial) (Hart et al., 2009). This suggests that one cue provides different signals on the basis of whether a decision has been made (Fischer & Greitemeyer, 2010; Hart et al., 2009). The current research focuses on this review valence (i.e., star ratings), one of the most prominent cues present on platforms such as Amazon.com. This type of cue categorically fits Walther and Jang's (2012) description of aggregated user-representations (AUR). Their framework describes different sources of persuasive influence and how they juxtapose in influencing on reader. AUR, in the case of star ratings, is deliberate in that users intentionally leave the rating to help readers in making the decision. However, the post-decision perspective takes a different theoretical contention that AURs can support alternative functions for dissonance reduction as opposed to directly influencing reader attitude toward a decision. This contrast is elaborated in that post-decision motivation alters the potential meaning conveyed by AUR, explained below. Pre-decisionally, the information conveyed by star ratings is clear. Positive reviews (e.g., 5 out of 5 stars or 4 out of 5 stars) recommend the product to other readers. This type of review often contains corresponding written information in the text that is favorable toward the product. Alternatively, negative reviews (e.g., 1 out of 5 stars or 2 out of 5 stars) dissuade other consumers from a decision by providing an overall unfavorable rating and corresponding written information that discourages the purchase. As opposed to valence, post-decisional star ratings can convey the congeniality of a review. Congeniality depends on the results of a prior decision (Hart et al., 2009). Congenial reviews, regardless of valence, confirm a decision. Uncongenial reviews disconfirm the decision. For example, if a reader purchased an Apple computer, congenial reviews include positive reviews of the Apple computer. In addition, congenial reviews include negative reviews of alternative products that the reader did not purchase (e.g., Sony computers). Uncongenial reviews include negative reviews for the Apple computer and positive reviews for alternative products (e.g., Sony computers). The need to alleviate the dissonance may lead readers to find congenial reviews and avoid uncongenial reviews (Chatterjee, 2001; Cummings & Venkatesan, 1976; Soutar & Sweeney, 2003). The current research applies and extends cognitive dissonance theory (Festinger, 1957) to examine the postdecision dynamic on participatory websites. 3. Cognitive dissonance theory Cognitive dissonance theory (Festinger, 1957) specifies the presence of cognitive elements. One or more elements that are inconsistent with one another lead to cognitive dissonance. Festinger (1957) notes that “[a decision between two alternatives, each with positive and negative aspects] is probably the most usual type of decision situation ... dissonance will result when action is taken” (p. 36). Specifically, the importance of the decision, the attractiveness of the unchosen alternative, and the degree of cognitive overlap affect the magnitude of dissonance. “The importance of the decision will affect the magnitude of dissonance that exists after the decision has been made” (Festinger, 1957, p. 37). For example, deciding between products a person may actually purchase or own is more important than a hypothetical decision. An important decision invokes more cognitive elements. For instance, learning about a product that one may potentially purchase corresponds with important cognitive elements regarding future consequences of receiving the product (e.g., reliability, quality) and the prospect of using it. The attractiveness of the unchosen alternative also affects cognitive dissonance. Selecting one attractive alternative (e.g., a product) means rejecting the other attractive alternative (e.g., another product) with its own associated desirable features. Put plainly, choosing between two similarly attractive alternatives results in “giving up” something, and thus produces cognitive dissonance. For example, if a reader views a Samsung and a Sony phone as nearly equal in howmuch he/she would like to own the phone, this person will experience a high degree of dissonance from the decision. However, when two alternatives are not similar in Y. Liang / Computers in Human Behavior 64 (2016) 463e471 465 attractiveness and the chosen alternative is clearly superior, the decision produces little dissonance. According to Festinger (1957), if two alternatives share similar attributes (different from attractiveness discussed previously), the cognitive elements related to these two alternatives overlap. For products, attributes involve product features, specifications, capabilities, among others. When a high degree of overlap occurs, little cognitive dissonance is aroused. Festinger demonstrates this context with an example involving a choice between two identical products where the only difference is that one costs $5.00 while the other one costs $4.99. The two products share a high degree of overall features in terms of cost, or cognitive overlap. The choice of either product would result in little or no cognitive dissonance, despite the similarity in the attractiveness of price. Conversely, cognitive dissonance occurs when individuals chooses between two alternatives with features that qualitatively differ from each other. In this situation, decrease in overlap invokes different cognitive elements associated with each of the qualities. The first research question addresses the effect of the three factors affecting the magnitude of cognitive dissonance in the context of online reviews on participatory websites: RQ1: After making a decision, what is the relationship between the importance of the decision, the attractiveness of the unchosen option, and the degree of cognitive overlap on cognitive dissonance in a participatory website environment? 4. Effect of decision on review selection and cognitive dissonance Since its inception in 1957 and a later addendum in 1964, cognitive dissonance theory had undergone tremendous test and development over the span of 50 years (For a review, see Cooper, 2007). Recent meta-analytic reviews (Hart et al., 2009) and revised models of the theory (Fischer & Greitemeyer, 2010) concluded with a clear support the effect of dissonance on communication behaviors. These behaviors manifest in terms of how individuals select messages (e.g., online reviews) after making a decision. The current study adopts the original and robust tenets of cognitive dissonance theory (Festinger, 1957) to experimentally test the dissonance effects on participatory websites. The goal is to examine how individuals may utilize this more recent and newer communication technology, given its broad availability, ease for message selection, and presence of cues, to serve the alternative goal of dissonance reduction explained by the classic theory. In the context of consumer decisions, Oliver (1997) explained that the dissonance is uniquely maximized during the “gamma stage” when consumers made a decision but have not yet received the product or service to have direct experiences. This is an important stage as it connects the psychological state of the consumers to the communication messages that they may select. The dissonance affects message exposure in terms of online reviews, which in turn affects dissonance. The characteristics of these reviews may even influence post-purchase attitudes, which is an important antecedent to product regret and possible product return intentions (Das & Kerr, 2010). As a psychological discomfort, cognitive dissonance motivates individuals towards its reduction. To alleviate cognitive dissonance, individuals have both cognitive and behavioral strategies. Cognitive strategies involve psychological changes associated with the cognitive elements (e.g., changing existing cognitive elements, changing the importance of elements, adding elements, and/or removing elements). The current research focuses on the behavioral strategies that affect how dissonant individuals select and read online reviews using cues available on participatory websites. One behavioral strategy is to avoid information that produces additional cognitive dissonance. Specifically, individuals avoid uncongenial information that challenges or opposes the decision. When readers prefer congenial informationmore than uncongenial information, a congeniality bias occurs (Hart et al., 2009). Hart et al.’s (2009) meta-analysis demonstrated an ample and robust effect of the congeniality bias effect (average d 1⁄4 0.36) over the span of 91 studies across five decades of research. To facilitate the selection process, the current study employed star ratings (i.e., number of stars), a cue that provides information of whether a review was congenial or uncongenial to a prior decision. Indicators of congenial exposure involves the number of reviews read and reading time. Specifically, more exposure of reading time supports attitude consistency (Knobloch-Westerwick & Meng, 2009): H1a. Cognitive dissonance leads individuals to read more number of congenial reviews than uncongenial reviews. H1b. Cognitive dissonance leads individuals to spend more time reading congenial reviews than uncongenial reviews. In addition, cognitive dissonance also guides the order in which readers examine reviews. Although readers can read a series of reviews, congenial reviews that reduce cognitive dissonance should systematically attract more attention than uncongenial reviews. The selection behavior manifests in terms of the order to which readers select reviews (congenial or uncongenial). Specifically, readers strategically select information, or congenial reviews, earlier than uncongenial reviews (e.g., Winter & Kr€ amer, 2012). H2. Aftermaking a decision, individuals generally select congenial reviews to read earlier than uncongenial reviews. In turn, exposure to online review, either congenial or uncongenial, affects cognitive dissonance. Specifically, reading congenial information bolsters cognitive elements that support the original decision. The resulting cognitive dissonance should diminish. Applying this concept: H3a. Reading more number of congenial reviews than uncongenial reviews affects subsequent cognitive dissonance. H3b. Spending more time reading congenial reviews than uncongenial reviews affects subsequent cognitive dissonance. 5. Initial evidence demonstrating post-decision dissonance An initial survey collected data to ascertain if individuals read online reviews after a product decision. Two questions were included as a part of a larger survey study related to consumer behavior (n 1⁄4 425). The first question prompted participants, “How often or rarely do you read online customer reviews for a product after you purchased it?” Approximately 40% of participants reported between sometimes to very often. The second question asked participants to describe the type of information they found after they purchased a product, ranging from negative to positive. An overwhelming majority (86%) of the participants reported the information they found as somewhat positive to very positive. Results from this initial survey offered empirical justification to demonstrate that individuals do select and read reviews after a purchase decision, and the information appear to bias towards supporting the product decision. 6. Study 1 (lab) e the effect of decision on review selection and cognitive dissonance stimulus pretesting A pilot study pretested the stimuli for the main studies (Study 1 and Study 2). The main studies focused on examining the effects of Y. Liang / Computers in Human Behavior 64 (2016) 463e471 466 cognitive dissonance on review selection, and how reviews affected cognitive dissonance. The experiential design involved participants who ranked products within a single product category (e.g., coffee makers), selected between two ranked products (e.g., the second and third product on a ranked list of 10 products), and read additional online reviews about the product that they selected. The pilot study examined different product categories. 6.1. Pretesting procedures An offset group of 94 participants viewed three pre-selected product comparison matrices. Each matrix consisted of information about 10 products from a single product category (i.e., 10 coffee makers, 10 blenders, or 10 vacuum cleaners). Participants viewed one comparison matrix at a time. For each matrix, participants received pictures of 10 products within the product category. Participants also received descriptions regarding relevant product attributes for that category. For instance, relevant attributes for blenders included motor power, number of speeds, jar capacity, jar material, and dimensions. Participants ranked the products within the category from best toworst quality, rated each product's quality from good to bad on a single 15-point semantic differential item, and reported the overall difficulty in ranking all 10 products within the category using an original scale with four 7-point semantic differential items: difficult/easy, demanding/undemanding, takes a long time/take a short time, and challenging/unchallenging Participants also reported their overall involvement with each product category (Zaichkowsky, 1994; a 1⁄4 0.91). 6.2. Pretesting results Empirical determination for the product category reflected three criteria: low or moderate product category involvement, similar quality ratings among products within the category, and higher difficulty during the ranking task. Coffee makers was selected given its low product category involvement (M 1⁄4 2.69), low variance for product quality ratings (mean variance 1⁄4 11.59) and product ranking difficulty (M 1⁄4 3.99).