The differential effect of three types of conflicting online reviews on perceived informativeness

Online shopping, thanks to its efficient characteristics, has witnessed a dramatic evolution and expansion in the last few years (Xu, Benbasat, & Cenfetelli, 2020). Given the limitations of online shopping, where customers cannot physically feel and observe a product (Boardman & McCormick, 2018), customers instead read the data and information provided by the merchant about the color, size, and material of a product and attempt to guess whether the product meets the customer's expectations based on past experience or existing knowledge. In addition to the more official product descriptions provided by the merchant, another source of information that customers will draw on is the reviews and descriptions from previous buyers. Previous buyers may describe the product based on their use of the item, and in this way, provide other potential customers with helpful information to assist them in further confirming that the product may match their needs. In other words, potential customers' purchasing decisions are somehow influenced by what previous customers have commented on, so studying online reviews may help to understand customers' purchasing behavior (Ashraf et al. 2020; Vali, Xu, & Yildirim, 2021).

We consider online reviews by past consumers to be one of the main sources of product information that potential customers gather to help them decide whether to place an order. Therefore, perceived informativeness of online reviews is an important outcome as it can influence the subsequent purchase decision (Sun, Han, & Feng, 2019). The informativeness of a review is the overall information that can be apprehended through reading a review text. Factors that affect the perceived informativeness of an individual consumer review include the length of a review, aspect ratings, and the number of product attributes mentioned (Chen & Xie, 2008; Park, Lee, & Han, 2007; Li, Zeng, Xu, Liu, & Yao, 2020). Moreover, in reality, it's important to recognize that not all reviews will inevitably align with a singular attitude. This holds true even for the same product; hence, existing online reviews sometimes present inconsistencies. While one customer might leave a positive evaluation for a certain item, another could provide negative feedback. This variance in reviews can be attributed to the differing objectives customers have for the product, as well as distinct subjective assessments and perceptions they apply. Regardless, these conflicting appraisals can significantly impact a prospective customer's decision-making process. When an item garners both favorable and unfavorable descriptions, it becomes challenging for the potential buyer to gauge whether the product will meet their expectations, potentially leading to confusion (Ruiz-Mafe, Chatzipanagiotou, & Curras-Perez, 2018). Consequently, the uncertainty stemming from such information discrepancies leads us to consider conflicting reviews as an additional factor influencing the perceived informativeness.

Among the existing conflict reviews, there are three primary types of conflict. The first is conflicting opinions on attributes of an item, where at least one former buyer expresses a different, or even opposing, opinion about a particular attribute of the same item (Siddiqi, Sun, & Akhtar, 2020). For example, when confronted with a handbag of the same size and material, one customer comments that it is heavy, while another customer remarks that it is actually very light. The second is conflicting ratings of an item, and this conflict is more intuitively expressed by rating disagreement as opposed to the first conflict's textual description (Ruiz-Mafe et al., 2018; Bigne, Chatzipanagiotou, & Ruiz, 2020). For instance, one customer gave the item a rating of five out of five, while another customer gave it only the lowest rating of one. Lastly, the third type of conflict is the intensity of conflicting reviews of an item, which is more complicated than the first two, and represents the strength of opposition across the entire comment section (Yao, Fang, Dineen, & Yao, 2009; Ruiz-Mafe et al., 2018). For example, in ten reviews of an item, half of the customers report a positive attitude towards the item, while the other half report a negative attitude; whilst in ten reviews of another item, only one person shows a positive attitude, whereas the remaining nine customers all leave negative feedback. This means that the intensity of conflict is stronger for the latter than for the former, and we expect that the different intensities of conflict observed by potential customers will make a difference in their ultimate purchasing decision. Although some existing studies have categorized conflict reviews, a more thorough understanding is absent. We would like to take a more comprehensive perspective on the impact of conflict reviews in the e-commerce environment.

Existing research has investigated the effects of online review attributes, including volume, review valence, length, and submission date on sellers' sales (Ren & Nickerson, 2019), buyers' purchase intentions (Park et al., 2007; Luo, Duan, Shang, & Pan, 2021) and buyers' purchase decisions (Forman, Ghose, & Wiesenfeld, 2008).

Nonetheless, the impact of various forms of conflicting reviews on perceived informativeness remains unexamined (Vali, 2016). We consider that when potential customers are influenced by online reviews, such as reviews that lead to false perceptions of a product, and purchase an item that does not meet expectations and request a refund, the merchant has to pay for the return shipping costs as well as some opportunity costs, resulting in increased operational costs and a corresponding reduction in profits to the merchant. Through improved perceived accuracy of purchase decisions and thus minimizing the cost of incorrect purchases, there can be significant improvements in cost control throughout the supply chain from manufacturers to distributors (Christozov, Chukova, & Mateev, 2009). As such, the significance of our research is that based on our findings, merchants or platforms can take countermeasures to reduce the negative impact of conflicting reviews by circumventing them in advance, thereby reducing the loss of profit due to increased operational costs associated with returns and refunds (Chen & Xie, 2008).

The existing literature has examined online reviews extensively. In studies addressing the perceptions of review readers, researchers have found that the disclosure of the reviewers' identities, as well as the quality of online reviews, such as the accuracy, relevance, and timeliness of the information provided, positively influence perceptions of reviews and subsequent decisions based on information comprehension (Liu & Park, 2015); and also found the ways in which other consumers' online reviews influence or help viewers to make decisions (Mudambi & Schuff, 2010; Luo et al., 2021; Liu & Hu, 2021).

Research has shown that prospective customers often consult multiple reviews when in the process of searching for a product (Liu, Lee, & Srinivasan, 2019). Therefore, if there are conflicts between product reviews, then this conflicting information can affect the informativeness of the online reviews that potential customers obtain and the subsequent purchase decisions, such as purchase deferral arising from reduced attitude certainty (Park et al., 2007). While limited scholarly attention has been devoted to exploring the influence of conflicts on online market performance from various angles, there is still a significant need for further investigation to tackle numerous research inquiries. These inquiries encompass identifying the categories of conflicts within online consumer reviews, examining how distinct types of conflicting information impact the helpfulness of these reviews, and determining whether the perceived helpfulness affects consumers' perception of making the correct purchase. Providing answers to these queries holds the potential to empower consumers to enhance the precision of their buying choices while concurrently mitigating the additional expenses associated with erroneous purchases.

There are some studies in the existing literature that note the phenomenon of conflicting reviews. These studies have found that conflicting ratings affect the credibility of reviews and their usefulness to potential customers in understanding the product (Siddiqi et al., 2020), as conflicts, i.e. the lack of consensus opinion, can cause discomfort (Matz & Wood, 2005; Angelidis & Lapata, 2018) and uncertainty; or affect the informativeness the reader obtains (Park et al., 2007). It has also been suggested that when both positive and negative comments are present, the impact of negative comments is greater than that of positive comments, according to negativity bias (Cui, Lui, & Guo, 2012). However, these studies focus on only a single type of conflicting review, such as conflicting star ratings. Our study investigates multiple dimensions of review conflicts and their impact on potential customers' perceived informativeness. The three types of conflicting reviews are described below.

People have different tastes, and thus such different preferences lead reviewers to comment on the same attributes differently. Here is a sample review from the Google Shopping website showing conflicting opinions on the same attributes of a mechanical pencil (Vali, 2016):

While past literature has noted that potential customers will consider conflicting reviews regarding product attributes (Von Helversen, Abramczuk, Kopeć, & Nielek, 2018), they have only examined which positive and negative reviews are more influential for observers, what biases elicit positive and negative sentiments and thereafter, the conflicting reviews (Cheng & Jin, 2019), as well as when facing conflicting reviews what the observer's focus is (Gavilan, Avello, & Martinez-Navarro, 2018). While these studies all convey the findings that potential customers do consider conflicting reviews about product attributes from past consumers on multiple levels before consumption behaviors occur, to our knowledge, we are the first to explicitly examine the effect of conflicting reviews about product attributes on the informativeness potential customers will embrace.

Conflicting ratings are the second type of conflict to be investigated in this study. Conflicting ratings refer to the fact that different customers rate the same item with significantly different scores. Compared with conflicting opinions on specific product attributes in written text format, conflicting ratings focus on the different opinions on a product as a whole in numeric format. Although this type of conflict has been studied by a few researchers (Mudambi & Schuff, 2010; Bigne et al., 2020), so far no study has analyzed the impact of this type of conflicting information when other types of dissonance are also present, since different types of conflict may have different impacts when other conflict types are present, according to Cognitive Dissonance Theory (Festinger, 1962).

Furthermore, we present the third type of conflicting review: the intensity of disagreement between reviewers' opinions. For instance, consider a scenario where there are five reviews of a product. Among these, two reviews express similar opinions regarding a specific product attribute, while the remaining three reviews collectively share similar opinions among themselves, but these opinions sharply contrast with those of the first two reviews. This cluster of reviews exhibits a higher degree of conflict compared to a situation where four reviews are in agreement, and only one expresses an opposing view (Yao et al., 2009). This form of conflict has also been recognized as disagreement versus consensus. Disagreement can induce discomfort as it poses informational challenges by potentially undermining the credibility of individuals' attitudes or social challenges by jeopardizing the shared social identity of the group (Matz & Wood, 2005). Research in decision-making has shown that a lack of consensus opinion creates uncertainty for consumers (Wu, Ngai, Wu, & Wu, 2020). Previous research has shown that consumers react negatively to such uncertainty (Xu et al., 2020).

When customers read a set of reviews, their initial goals are to learn as much as possible about the target product by learning from the contents. While multiple factors within a collection of reviews can influence the learning experience and the perceived level of informativeness, one prominent theory in the study of effective factors during learning is Cognitive Load Theory (Sweller, 1994). This theory posits that the challenges individuals encounter when acquiring new intellectual skills can vary significantly, spanning from straightforward to intricate tasks. Consequently, the perceived informativeness can also vary widely, ranging from minimal to extensive.

Humans only have a limited working memory to process incoming information. Hence, when individuals experience an excessive load on their working memory, it has a detrimental impact on the learning process (Xu, Benbasat, & Cenfetelli, 2014). Within the framework of Cognitive Load Theory, task complexity emerges as a pivotal factor. Additionally, research by Wood (1986) posits the existence of a curvilinear connection between task complexity and productivity. We consider the process of potential customers learning about an item by reading and browsing reviews to be a learning process, but the conflict between reviews actually brings more complex information, which affects the brain's cognitive load and, thus, their processing and eventual perception of the information. More specifically, while browsing more reviews for more information can help potential customers understand the item better (Locke, Shaw, Saari, & Latham, 1981), due to the limited cognitive load of the human brain, when the complexity of the information exceeds a certain amount resulting in brain overload, human learning behavior will be affected in the opposite way (Roetzel, 2019), i.e. learning effectiveness and information usefulness will decline (Xu et al., 2014). Kamis, Koufaris, and Stern (2008) also found that as complexity increases, usefulness follows an inverted U-shaped curve. Therefore, as illustrated in Figure 1, before the brain is cognitively overloaded (Point A), higher information complexity leads to higher information cognitive effectiveness; after the brain is cognitively overloaded (Point A), higher information complexity leads to lower information cognitive effectiveness.

In addition to information complexity, another factor that affects cognitive load is the structure of the information. According to the schema mechanism of learning, when the information is pertinent to the subject and the information itself is easily captured, the learning process is likely to be uncomplicated. Conversely, if the information lacks relevance to the subject and proves challenging to grasp, the learning difficulty is expected to be elevated (Sweller, 1994).

We, therefore, propose to classify information according to two criteria: product diagnosticity (Byun, Ma, Kim, & Kang, 2021) and ease of capture (Sweller, 1994; Xu et al., 2014; Hanjaya, Kenny, & Gunawan, 2019). When product information is less complicated, it is easy to learn and understand. If the information is complicated, the difficulty of learning is high. According to Cognitive Load Theory, if a review is highly diagnostic of the product and provides information that is easily captured, then product information is easily learned, even if this information is inconsistent; conversely, if the information is either irrelevant or hard to capture, the learning task becomes arduous, and the inconsistency becomes even more detrimental. In our case of online reviews, the relevance to the topic depends on whether there is a direct description of the product to help the potential customer determine if the product meets expectations; and otherwise, the ease of capturing information is determined by whether it is possible to tell at a glance what prior customers' opinion of the product is (Table 1). If the learning difficulty for absorbing information is high, humans will receive relatively little informativeness about it; Conversely, if the learning difficulty is low, humans may receive more informativeness about it. To summarize, only when product information provides high product diagnosticity and ease of capture, perceived informativeness would be high. The absence of one of them will have a negative effect on perceived informativeness.

Three types of conflicting reviews in our research context are now being marked up in the cognitive load theory parabola. Because conflicting attribute reviews positively affect informativeness, it follows that they should exhibit greater learning effectiveness compared to the other two types of reviews that carry a negative effect. Alternatively, between conflicting ratings (which are just numbers at a glance) and the intensity of conflicting reviews (requiring the thorough reading of captions in all reviews), it becomes apparent that while both possess low learning effectiveness, the intensity of conflicting reviews carries significantly greater information complexity for potential consumers. At the same time, comparing conflicting attribute reviews and the intensity of conflicting reviews, the latter exhibits greater information complexity. This complexity arises due to the inherent difficulty in capturing the full scope of these reviews and the substantial cognitive effort required by potential consumers to process the extensive information. Consequently, owing to the simultaneous fulfillment of these conditions, we propose that these three categories of conflicting reviews be positioned within cognitive load theory, akin to the arrangement depicted in Figure 2. The rationale behind the intensity of conflicting reviews surpassing the cognitive load capacity limit lies in the challenge of extracting sufficient resources from all reviews to make informed judgments. This intensive process of review analysis surpasses the human cognitive load capacity threshold, leading to an overwhelming cognitive burden.

When facing a conflict of opinion about an item's attributes, the information provided is relatively complex but does not exceed the cognitive load of the human brain, so human cognition of information is very effective. For instance, in the case of pencils, the reviews might highlight excellent lead performance and comfort, but not the eraser quality of the pencils. At the same time, conflicts over item attributes are easily captured and provide information that is highly relevant to the item, i.e. there is a direct description (positive or negative) of the item and rationale to help the potential customer decide, so the learning process of the item is relatively easy and therefore the conflict of opinion over item attributes is informative.

Settle and Golden (1974) conducted a study to investigate how disparities in advertising affect the reactions of prospective consumers to these disparities. Their research revealed that maintaining consistency in advertisements does not always lead to heightened assurance. Put differently, when conflicting messages appear in advertisements, they can actually boost a consumer's perceived confidence in a product, thereby conveying information that seems more related to the product itself rather than the seller's intention to promote it. However, the role of this attribute conflict in today's e-commerce market, especially in consumer reviews, has not been noticed adequately. We want to verify whether this effect also exists in the informativeness revealed by consumer reviews, meaning that review conflicts about product attributes may positively affect consumer perceptions of informativeness. We can incorporate the discoveries of Settle and Golden into our research when a prospective purchaser has encountered conflicting attributes. The consumers will have a better understanding of product attributes from a different perspective with relatively less effort. Based on this argument, we hypothesize:

Alternatively, when faced with conflicting ratings of an item, the information provided is completely plain, and only a little bit of brain capacity will even be used to interpret this information. For example, when a potential customer is confronted with a mug sold online and some people give it a full rating out of five, while some people leave only the lowest rating, though the observer can easily capture this information, they do not derive any useful helpfulness as to why it was given a full rating by some people and the lowest rating by others, leaving no special information about the mug at all. According to the inverted U-shaped curve of cognitive load, human perception of such low-complexity information is quite ineffective. Also, since conflicting product ratings do not have product diagnosticity, this indicates that there is little helpfulness in making decisions (Mudambi & Schuff, 2010). In other words, despite conflicting ratings being easily captured, it does not provide information that is highly relevant to the item itself, i.e. there is no direct description of the item to help potential customers make a decision, so the learning process for the item is not easy and therefore conflicting ratings of an item are weakly informative (Lo & Yao, 2019). Therefore, we hypothesize:

Similarly, when confronted with an item possessing high-intensity conflicting reviews, the efficiency of human cognition of such information is greatly reduced and the information learning effectiveness is low as the information provided is overly complex and even exceeds the boundary line of the human brain's capacity to process information (i.e. point A of our inverted U-shaped curve of cognitive load). For instance, in the same example of the mug, half of past consumers leave negative reviews, while the other half brag about it. Although such a review set allows potential customers to aggregate and process a lot of information from various sources to help evaluate the product, their brain memory cognition is often overloaded with it, making it difficult to effectively process this diverse information, not even mention that it is not easy to capture the key points from the various reviews (Su, Yang, Swanson, & Chen, 2022). At the same time, the intensity of the conflicting reviews is difficult to capture, even though it provides a high amount of information related to the item, i.e. there is a direct description of the item to help the potential customer make a decision, so the learning process for the item is not easy and therefore the conflicting reviews of an item are weakly informative. Therefore, we hypothesize:

Next, we investigated the effects of the co-presence of different types of conflicts. When intensive review conflicts and conflicting opinions on attributes, or intensive review conflicts and conflicting ratings, co-present, observers' attention is focused on the attributes or ratings over anything else, because either conflicting attributes or conflicting ratings, respectively, possess a high ease of capture, so the observer intuitively relies more predominantly on these two for informativeness. However, the situation becomes nuanced when it comes to the simultaneous presence of conflicting opinions on attributes and conflicting ratings. This is because conflicting opinions on attributes already possess both ease of capture and product diagnosticity, and it already effectively provides the observer with adequate informativeness. The inclusion of ratings increases the cognitive load of consumers, which means it increases the difficulty of processing information and prevents consumers from studying or understanding the meaning of the review information readily. Therefore, adding inconsistencies in ratings to the processing of attribute information may lead to cognitive overload for consumers (Lo & Yao, 2019; Bigne et al., 2020), making them unable to make accurate judgments, which in turn leads to an imbalance in perceived informativeness. Hence, compared to the presence of conflicting opinions on attributes alone, the informativeness perceived by potential customers is rather reduced when it is present simultaneously with conflicting ratings. Therefore, we hypothesize:

Perceived correct purchase revolves around elucidating the dynamics between online shoppers and a product review system. It encompasses how online shoppers gather information through perusing and scrutinizing product reviews, and how this information subsequently shapes their perception of a product. It's important to emphasize that greater informativeness doesn't have a direct, overwhelming impact on the potential customer's buying decision. Instead, it works by boosting the perceived informativeness of online shoppers. This, in turn, reduces their uncertainty about the product from various angles and ultimately improves their perception in making a more accurate purchasing decision. To illustrate, the official product description provided by the merchant serves to diminish product description uncertainty; feedback from fellow consumers acts to alleviate the product's performance uncertainty; additionally, subjective assessments of the product's attributes by other consumers from diverse standpoints assist potential buyers in assessing whether the product aligns with their individual needs, thus mitigating product's fit uncertainty - these dimensions collectively encapsulate the spectrum of product uncertainty (Xu et al., 2020). Consequently, this reduction in overall product uncertainty augments their perception of a more correct purchase decision. Based on the above deductions, we hypothesize in this study that:

The entire research model is shown in Figure 3.

To test our proposed hypotheses, we used an experiment with a 2 × 2 × 2 factorial between-subject design for each conflict type comprising two levels. Table 2 presents illustrations corresponding to the high and low levels of each of the three types of conflict. In our experiment, a class of students was presented with five online reviews of a pencil. In terms of the attributes of the item, i.e. the pencil, we provided online reviews containing four attributes: ease of use, eraser, refillability, and grip handling. Before commencing the experiment, participants received notification that their involvement would entitle them to receive an additional 1% bonus point, alongside the opportunity, as is common in many other experiments, to be entered into a drawing for a USD $50 Best Buy gift card. For the purpose of our study, we created 8 different scenarios: 2 (low vs high conflicting opinions on an item's product attributes) × 2 (low vs high conflicting star ratings of an item) × 2 (low vs high intensity of conflicting review of an item). Although each scenario contained five online reviews, these scenarios had different levels of varying conflicts. For example, in the low*low*low combination, the five reviews are presented in such a way that the reviews only have conflicting opinions on one attribute, all five reviews are rated roughly the same, and only one of the reviews is inconsistent with the attitudes of the rest of the reviews. Subjects were randomly assigned to any one of these eight scenarios, and they would indicate their perceptions about informativeness and correct purchase after viewing the online reviews.

We adopted measures from the existing literature. A sample item for perceived informativeness is “In general, these online reviews provided me with high-quality information for the pencil selection task” (Xu, Benbasat, & Cenfetelli, 2012; Vali, 2016). One sample item for perceived correct purchase is framed as follows: “This product will be the correct pencil for me” (Balabanis & Craven, 1997). To serve as a manipulation check for each conflict type, we requested participants to gauge the extent of conflict on a 7-point scale. The results, displayed in Table 3, demonstrate that participants were able to effectively differentiate between low and high levels of conflict for each type.

In our study, we recruited 194 student participants from a public university in the Midwest of the United States with an average age of 24.2 years, including 152 males and 42 females; 6 of them were graduate students and the rest were undergraduates; about 65% of them were between 18 and 24 years old, 21% between 25-29 years old, and 14% were over 30 years old. These subjects came from three colleges and ten majors and hence represented a diversity in individual backgrounds. We compared the average amount of perceived informativeness between all groups, utilizing a 7-point scale for measurement. Table 4 displays the mean level of informativeness across the eight subject groups.

We used a variety of statistical models, including ANOVA and regression, to analyze our proposed research model. In order to assess the internal consistency of the constructs, we computed Cronbach's alpha for each of them: 0.858 for perceived informativeness and 0.857 for perceived correct purchase. All items met the recommended tolerance (>0.70, Fornell & Larcker, 1981). Factor analysis showed that each item exhibited a stronger association with its respective construct compared to any other constructs assessed concurrently, indicating good discriminant validity.

We conducted a regression model to examine H1a, H1b, and H1c. The regression model showed that conflicting opinions on attributes positively influence perceived informativeness (β = 0.25, p < 0.05), but star ratings negatively influence perceived informativeness (β = −0.27, p < 0.05), and intensity of conflicting reviews also negatively affects perceived informativeness (β = −0.28, p < 0.05), supporting H1a, H1b, and H1c. In addition, the ANOVA analysis conducted for H2 indicates a significant interaction effect between conflicting ratings and conflicting opinions on attributes (p < 0.05). Figure 4 depicts these results. When the level of conflicting ratings changes from low to high, the positive effect of conflicting opinions on attributes on perceived informativeness decreases. In addition, regression results showed a significant relationship between perceived informativeness and the correct purchase measure (p-value <0.001), supporting H3.

This study has several theoretical contributions. Firstly, we established a theoretical connection between the perception of product informativeness and two key criteria: product diagnosticity and ease of access. This theoretical contribution not only spares us the tedium of reiterating existing research on categorizing conflicting reviews and discussing their individual impacts on informativeness, but, more importantly, it offers insight into how the variations between distinct types of conflicting reviews generate diverse impacts on informativeness. In particular, the distinction lies in the intuitive nature and descriptive narratives of conflicting opinions on attributes versus conflicting ratings and review intensity. When reviewers express attribute-based opinions, they describe the attribute's positivity or negativity, aiding potential customers in making informed judgments. This contrasts with less intuitive conflicting ratings and intense reviews, which lack such descriptive clarity and effectiveness in conveying meaningful information to customers. Thus, our study advances the literature by taking conflicting comments beyond the mere categorization of past research, and going deeper and linking their disparities to their impact on perceived informativeness.

Second, in previous research, two out of the three conflict types (conflicting ratings and intensity of conflict) have been identified and studied independently. However, no study has comprehensively investigated all three conflict types collectively to assess their combined effects and their influence on customers' perception of informativeness. Through the simultaneous examination of these conflict types within a single study and the comparison of their respective effects on perceived informativeness, we uncover the varying degrees of influence that different types of conflicting reviews hold. Additionally, we also uncover how conflicting ratings and conflicting attribute opinions interact to affect this perception. Through this comprehensive analysis, we introduce a fresh perspective: the lack of either ease of capture or information diagnosticity is detrimental to users' perceived informativeness.

Third, our study holds the distinction of being the first to evaluate how various forms of conflicting information within reviews influence perceived informativeness. Our discovery that the inclusion of contradictory attributes unexpectedly enhances consumers' perceptions of informativeness challenges the conventional belief stemming from prior research, which predominantly indicates that conflicting reviews invariably lead to adverse consumer perceptions. This illuminates diverse pathways through which conflicting reviews can exert influence, thereby addressing a previously unexplored dimension in research. In other words, this finding suggests that loadings from information that appears contradictory and intricate at first glance may not be as high as expected. Consequently, certain fundamental principles and assumptions of the cognitive load theory may be questioned, necessitating more thorough situational refinement for further exploration and discussion.

Assessing when and what reviews should be posted to improve vendor sales has been a discussion topic for a while (Chen & Xie, 2008). In all these research endeavors, the primary emphasis has revolved around enhancing immediate sales, without due consideration for the augmented expenses linked to erroneous purchases. Rather, our study proposes to start with the information that potential customers obtain about the product, identify the key factors that enhance the informativeness they perceive, and then use these to improve the customer's purchase accuracy to increase multilevel sales success and thus assure profit boost. Our research has helped merchants better understand consumers' perceptions of conflicting reviews, which can guide their design of review displays. In other words, the findings of this study suggest that online vendors can improve perceived correct purchase by managing consumer reviews with applications that ensure both ease of capture and product diagnosticity with proper complexity. For example, when there are conflicting opinions on product attributes, merchants would prefer to display these reviews in a more conspicuous spot throughout the review area, whereas when there are conflicting ratings and intensive conflicting reviews, merchants should find a way to place these in a secondary, less noticeable spot. Some online sales platforms have recognized the need for a more effective approach to sharing information and, as a result, have revamped how they present product reviews. For instance, Taobao now showcases the counts of positive and negative reviews at the beginning of the review section. This helps potential customers quickly grasp the extent of conflicting opinions about the product from previous consumers. Moreover, with the option to apply filters, it becomes easier to discern which specific features and aspects of the product the positive and negative reviews focus on, allowing customers to access essential information in a more direct and efficient manner. The findings of this study have important implications for companies that sell online, particularly in the context of managing online reviews. Some companies have invested thousands of dollars in their scenarios to generate profits and engage returning customers by providing the proper product information. Online sellers can use the results of our research to understand which set of online reviews generates the greatest amount of perceived informativeness for their customers.

This study is not without limitations, which may open avenues for future research. First, additional factors may help to elucidate how varying forms of conflicting reviews impact perceived informativeness. For instance, factors such as familiarity with the product, understanding of customer preferences, and individual relevance can influence the impact of conflicting reviews on their informativeness. Second, only one product was used in our experiment, and yet having a more expensive product for the experiment might induce different customer perceptions, so additional supplementary research may be required to assess the resilience of our conclusions by experimenting with various product categories and varying price ranges in the model. Finally, only five reviews were provided to each subject in our experiment, which is usually the number of reviews that customers at least need to consider (Park et al., 2007). Perhaps increasing the number of reviews would have had a different impact on the subjects and led to different experimental results. However, we also need to be aware that if more review quantities are included, the resulting information conflict and overload may produce even more negative effects on perceived informativeness.

The model proposed in this study considers conflicting reviews of product attributes, conflicting ratings, and the intensity of conflicting reviews among commenters. We highlight the influence of informativeness on perceived correct purchase. Based on cognitive load theory, we provide a classification for the three conflicting review categories with their effect of ease of capture and product diagnosticity. We also reveal that conflicting ratings and the intensity of conflicting reviews among commenters negatively affect perceived informativeness, but conflicting review opinions about product attributes have a positive influence. We also observed that conflicting ratings could mitigate the positive effects of conflicting reviews about product attributes. More specifically, perceived informativeness is negatively affected when there are conflicting ratings and conflicting opinions on attributes among commenters.

The work described in the paper was partially supported by grants from the National Natural Science Foundation of China (Project No. 72271210), Strategic Research Grants of the City University of Hong Kong (Projects No. 7005595), and the Digital Innovation Laboratory of the Department of Information Systems at the City University of Hong Kong.