The influence of online reviews on consumers' decision-making is a very subtle and complicated process (Chen et al., 2009; Li he. Lee, 2009), the concept of usefulness theory captures the persuasive effect and potential value of online reviews on purchasing decisions, and there are also some documents that are effective (Cheung &; Thadani, 20 12), persuasion (,20 1 1) and other similar expressions, in the empirical research, even the second-hand objective data (Cao et al., 20 1 1) are used, and the experimental or questionnaire data are also collected. Mudabi, 20 1 1). The research on the usefulness of online reviews focuses on "What kind of online reviews do consumers think are more useful?" This core problem (Korfiatis et al., 2012; Pan & Zhang, 20 1 1), the research results will help to understand the theoretical mechanism of the influence relationship between online reviews and consumer decisions (Ghose &; Ipeirotis，20 1 1； Zheng et al, 20 1 1).

On the basis of literature review and in-depth analysis, this paper holds that the existing research on the usefulness of online comments is mainly discussed from the following two aspects.

First of all, similar to the existing research on the aggregation effect of online reviews and sales, most of the research on the usefulness of online reviews focuses on the content characteristics of reviews, including star rating ②/extreme (Korfiatis et al., 2011; Mudanbi and. Shufu, 2065438+00; , 2008), comment length (Zhang et al., 2010; Mudanbi and. Schuff, 20 10), affirmative/negative statements (Cheung et al., 2009; Connors Corp. Mudabi, 2011; Schlosser, 20 1 1), the above characteristics are relatively objective and easy to measure. Mudambi and Schuff(20 10) clearly defined the concept of the usefulness of online comments in their research papers published by MISQ. Based on the theoretical framework of information economics and decision uncertainty, the econometric model analysis based on Amazon data shows that the extreme value (star rating) and the length of comments have a direct impact on the usefulness of online comments, and the moderating effect based on search and empirical products is investigated. Zhang et al. (20 10) used econometric model analysis to show that under the condition of controlling consumption purpose, positive/negative sentences (measured by star rating) and comment length have significant influence on online comment persuasion ability (measured by Amazon usefulness voting). Hao Yuanyuan and Ye Qiang (20 10), on the basis of expanding the research of Ghose and other scholars, discussed the factors affecting the usefulness of online reviews from the perspective of text features, and made an empirical analysis with film review data.

On the other hand, few literatures have found that besides studying the content characteristics of comments, it is necessary to further investigate the question of "who published online comments is more useful" (Connors &: Mudabi, 2011; Foreman et al., 2008; Hu, 2008; Pan & Zhang, 2011; Sun Chunhua and Liu, 2009). The empirical research of Forman et al. (2008), Ghose and Iperotis (20101) based on online data shows that the characteristics of reviewers-identity information, ranking of reviewers, number of comments published, etc. , will positively affect consumers' evaluation of the usefulness of online reviews, and more significantly affect product sales. Hu (2008) used the theory of transaction cost economics to study the effectiveness of online reviews, and found that consumers not only pay attention to the star rating of online reviews, but also are influenced by the expert status and activity of commentators in the network. Connors and Mudambi(20 1 1) reached a similar conclusion through experimental research, and the expert status of commentators in the network would affect the usefulness evaluation of their comments. The above preliminary empirical results are consistent with the persuasion theory of social psychology, that is, the source of information will affect the perceived value and acceptance of information recipients (Bhattacherjee &; Sanford, 2006; Stephen & Lyman, 2012; Foreman et al, 2008).

Table 1 collates and analyzes the above-mentioned important documents about the usefulness of online reviews, and then clearly understands the key factors that affect consumers' evaluation of the usefulness of online reviews.

To sum up, the existing research focuses on the content characteristics of comments, but pays insufficient attention to the factors of reviewers' perspective, especially the social factors of commentators. However, e-commerce and third-party comment websites are constantly enriching the socialization function, which highlights that the existing research ignores the important role of social mechanism in the influence of comment usefulness, which is both a blank point in the existing research and a theoretical contribution of this study; On the other hand, most of the existing literatures choose heuristic analysis method to comment on the influencing factors of usefulness (Wang et al., 2011; Connors Corp. Mudabi, 2011; Pan & Zhang, 20 1 1), lacks a reference theory with * * * cognition. In view of this limitation, this paper chooses the theory of information dissemination and persuasion in the field of social psychology as the theoretical analysis framework. The following (2) will briefly describe the dual-process model of persuasion and related research.

(B) the dual-process model of persuasion

Persuasion is one of the most basic forms of information dissemination and processing, and it is defined as the change of attitude or behavior caused by individuals accepting other people's information, which has always been a hot field of social psychology research (Bhattacherjee &; Sanford, 2006; Guo Zhenzhi and Xu Peixi, 2006). Among many persuasion process models, Heuristic-System Model (HSM) and Refined Likelihood Model (ELM) are the two most representative dual-process models (Chaiken,1980; Patty & ampCacioppo, 1986), widely used in advertising marketing, consumer behavior, knowledge management and other fields. HSM and Elm in information processing and attitude change (Angst &; Agarwal, 2009; Cheung et al., 2009): "Systematic Treatment of HSM" and "The Central Path of Elm" show that individuals will carefully and systematically analyze the received information and carefully consider whether to accept the views contained in the information, which will lead to a change in attitude; The heuristic processing of HSM and the edge path of elm show that individuals use simple decision rules to form judgments according to some situational clues of information, and then decide whether to accept persuasion.

In the field of information system research, Sussman and Siegal(2003) applied the ELM model to the online information exchange situation earlier and put forward the information adoption model. This model defines the theoretical concept of "information usefulness" for the first time, and holds that the quality of information content and the reliability of information sources are the direct factors that affect the receiver's perception of information usefulness.

After that, many literatures began to use the dual process model of persuasion to analyze and understand the process of users processing information and then influencing decision-making in the Internet environment (Cheung et al., 2012; Chu & Kamal, 2008; Zhang et al, 2010; Zhang Zikan, 20 10), which shows that the dual-process model of persuasion can provide an applicable reference theoretical framework for discussing the usefulness of online comments.

Third, theoretical models and assumptions.

According to the literature review, the existing research mainly focuses on the content characteristics of online reviews, including objective and quantitative characteristics, qualitative text factors and so on. , and lack of a more consistent theoretical analysis framework. Based on the two-process model of persuasion, consumers' evaluation of the usefulness of online comments can be regarded as a process of information processing and attitude change. This paper investigates the factors influencing the usefulness of comments from two aspects: the characteristics of comments and the characteristics of commentators, and emphatically analyzes the influence effect of social factors. Therefore, in addition to considering the content characteristics of comments, this paper also draws lessons from the theoretical perspective of conformity effect and social network, focusing on the social mechanism of the difference between average star rating and commentator factors affecting the usefulness of online comments (as shown in Figure 2), in order to make up for the shortcomings of existing research and expect to make greater theoretical contributions.

Length of comments

ELM and related research show that when consumers make purchase decisions, the information they can obtain such as product quality and business reputation is often incomplete, and they need to search information to reduce the uncertainty of consumption (Petty &: Cacioppo,1986; Sussman & Siegel, 2003). The empirical study of Chevalier and Mayzlin(2006) shows that the length of reviews means that consumers will read specific reviews, not just rely on the summary data of the review stars. The length of comments may stimulate readers to browse carefully, deepen or change their original attitude, improve their knowledge of products or services, and offset the uncertainty of users (Mudambi &; Shufu, 20 10). If information acquisition does not require additional search costs, the improvement of information diagnostic power brought by the length of comments will be more beneficial to mass consumers (Johnson &: Payne, 1985), thus enhancing consumers' evaluation of the usefulness of the review. Therefore, this paper puts forward the following research hypotheses.

Hypothesis 1a(H 1a): the duration of online reviews is positively correlated with consumers' evaluation of its usefulness.

On the other hand, cognitive load theory assumes that people's cognitive structure consists of working memory and long-term memory, and the capacity of working memory is limited, and only a little information can be stored and processed at a time (Sweller, 1988). In order to make the processing go smoothly, the amount of information currently entering the working memory cannot exceed the working memory capacity. Otherwise, it will lead to insufficient allocation of cognitive resources, thus affecting the individual's ability to learn or solve problems. This situation is called cognitive overload. When reading online reviews, consumers process information in working memory, and the length of reviews is one of the key indicators to characterize cognitive load (Gan et al., 20 12). Due to the limited capacity of working memory, although it is within a certain range, rich comment information will help consumers make shopping decisions. However, once the processing demand of comment information reaches the upper limit of working memory, it will cause consumers' cognitive overload and make it difficult for people to make better decisions. Therefore, we can draw the following research hypotheses.

Hypothesis 1b(H 1b): when the length of online reviews exceeds a certain range, it is negatively correlated with consumers' evaluation of its usefulness.

(B) with the average star difference-herding effect

From the literature review, the star rating of the review (including extremes) is an objective feature that will be discussed in almost all studies. Many scholars have conducted quite a lot of theoretical and empirical discussions from the theoretical perspective of information diagnostic power and information source reliability (Wu et al., 2011; Zheng et al. 20 1 1), and considered the adjustment function of product type (Mudambi &; Shufu, 20 10). From the perspective of information diagnostic power, many studies mostly think that online comments with unclear intermediate attitudes and opinions will be considered as lacking in added value of information, with limited influence on shopping decisions, and will be given a lower evaluation of usefulness by consumers (Forman et al., 2008; Bapulu & Dimoka, 2006). Therefore, this paper can put forward the following research hypotheses.

Suppose 2a(H2a): the average star difference (positive and negative) is positively correlated with consumers' evaluation of the usefulness of online reviews.

On the other hand, usefulness evaluation is actually to organize and apply online comments through social voting mechanism. Therefore, the biggest difference between this paper and the existing research lies in: based on the herd mentality of social psychology, this paper discusses how the difference between the star rating of a single review and the average star rating of a product will affect the usefulness evaluation of online reviews. Conformity effect is the result of social impact research in social psychology, which refers to the psychological process in which people consciously or unconsciously make judgments and change their attitudes and behaviors based on the opinions of the majority (Guo Zhenzhi, Xu Peixi, 2006). For the social voting mechanism of online comments, consumers will not only consider the content of comments, information sources and other factors when evaluating the usefulness of comments, but also be influenced by other consumers' opinions (Cheung et al., 2009). Therefore, when the star rating of a single comment is highly consistent with the average star rating of the product, consumers may give a higher evaluation of the usefulness of the comment. To sum up, a single comment with a large difference (positive or negative) from the average star rating does not represent the informed opinion of the reviewer, so consumers will choose to give a lower usefulness rating under the herd effect. Therefore, this paper puts forward the following research hypotheses.

Suppose 2b(H2b): the average star difference (positive and negative) is negatively correlated with consumers' evaluation of the usefulness of online reviews.

(C) the characteristics of commentators-social network effect

Literature review and analysis show that the existing research has not paid enough attention to the factors of reviewers, and mainly investigated the influence effect of reviewers' identity information (Forman et al., 2008), and hardly touched on the online social network of reviewers, but it should be paid attention to in the trend of Internet socialization (Smith, 2007). The dual-process model of persuasion shows that when the information receiver processes information along the edge path, the credibility of the information source is often the most important clue, which directly affects the basic judgment of the information receiver (Chu&: Kamal, 2008; Cheung，2009； Jin, 2008). At the conceptual level, the reliability of information sources can be characterized from two dimensions (Bhattacherjee &; Sanford, 2006; Zhang et al, 20 10): trustworthiness and expertise.

Accordingly, under the background of online comments, the professional ability and credibility of reviewers are the key factors that affect the evaluation of the usefulness of their comments (Wang et al., 2011; Yu Chunling et al., 20 1 1) is also an important heuristic rule that consumers will adopt when making evaluation and judgment (Cheung et al., 2012; Zhang & ampThadani, 20 12). From the perspective of social network, centrality is a key step to analyze the characteristics of nodes in the network and an important indicator to measure the influence of individuals in the social network (Smith et al., 2007), which reflects the social capital of individuals to some extent (Ellison et al., 2007; Yin et al., 2006). For example, an empirical study based on online data of Web2.0 video site (YouTube) shows that social networks play an important role in the diffusion and influence of user-created content. The more subscribers linked by video content creators, the greater the possibility that videos will get positive comments, which will further accelerate the spread and diffusion of video content (Susarla et al., 20 12). In addition, related research shows that there is a strong positive correlation between the network centrality of individual nodes and their credibility, which is also a reflection of the social capital of high school central nodes (Prell, 2003; Wang et al, 20 1 1). To sum up, the network centrality of reviewers will play a positive role in the process of consumers evaluating the usefulness of user reviews. Therefore, this paper puts forward the following research hypotheses:

Hypothesis 3a(H3a): The introversion of reviewers is positively related to the evaluation of the usefulness of their comments.

Hypothesis 3b(H3b): The extroversion of reviewers is positively related to the evaluation of the usefulness of their comments.

When the information receiver adopts the edge path of ELM model, he will process the information more according to social clues, and then form a judgment (Bhattacherjee &: Sanford, 2006). Network exposure is a key indicator to reflect the social clues of commentators, and it is a heuristic rule to guide the decision-making of information recipients, which will have a positive impact on the evaluation of the usefulness of their comments (ghost &: Ipeirotis, 2011; Hu et al., 2008). Groups are an important function of online communities, and each group has its own theme and hobbies. In an online community similar to Douban.com, users may join different interest groups, thus forming many different and overlapping social networks (Smith et al., 2007). According to the theory of "weak connection" in social networks, the online relationship between users in a group is "weak connection", and they also maintain a variety of relationship links and have certain influence in various social networks (Ellison et al., 2007). To sum up, it is reasonable to believe that the more commentators join the group, the more consumers will pay attention to and positively evaluate their online comments. Therefore, this paper puts forward the following research hypotheses.

Hypothesis 4(H4): There is a positive correlation between the relationship diversity of commentators and the usefulness evaluation of their comments.

(4) Control variables

In order to better verify the research hypothesis, this paper also introduces the timing of comments, the experience and ability of reviewers and other related factors as control variables, and the influence of film level will be considered in the econometric model.

The usefulness of online reviews will largely depend on the timing of reviews, that is, whether it can provide timely and up-to-date reviews for consumers who are collecting product information (Cheung et al., 20 12). From the perspective of information quality, many literatures think that timeliness is the key dimension of comment quality, which further affects consumers' evaluation of the usefulness of comment consumption (Otterbacher, 2009; Chen & Zeng, 20 1 1).

Another important clue of reviewers is the number of historical reviews, which will affect consumers' judgment on reviewers' experience, ability and reputation (ghost &: Ipeirotis，20 1 1； Wang et al, 20 1 1). From the analysis of the two-process model of persuasion, the experience and ability of commentators will enhance the credibility of their comments, be easily accepted by consumers, and give a high evaluation of their usefulness (Cheung et al., 2009; Foreman et al, 2008).