Keywords

1 Introduction

In recent years, product reviews have taken an important role in helping consumers make online purchasing decisions [3, 8]. According to statistics in [25], 84% of Americans rely on product reviews in their online purchasing process. Consulting reviews can effectively increase users’ understanding for products [29, 32], reduce their decision uncertainty [16, 24] and make users’ online purchasing process more enjoyable [18].

Given the important role of product reviews, merchants have tried to induce consumers to “spread the word” [17]. Actually, only a small proportion of consumers actively post their reviews online because consumers are social loafers rather than contributor [1]. It has been pointed out that consumers are more inclined to write reviews when they are very satisfied or disgruntled with the products, so that the reviews for a single item usually follow a bimodal distribution [20]. It can be inferred that reviews of a product sometimes can not reveal its true quality, which would mislead users’ purchasing decisions.

With the purpose to proactively induce consumers to ‘spread the word’, researchers have utilized different approaches to sparking positive community participation, e.g., using uniqueness and goal setting [2, 26]. In addition, some systems have been developed to provide authors real-time suggestions on what they may wish to write in terms of previous reviews [5, 12, 13, 21]. The results show that users feel they get support while writing reviews and would like to incorporate the suggestions into reviews. Furthermore, attention has been devoted to the forces that motivate consumers to write product reviews, e.g., economic incentives, concern for other consumers and social benefits, etc. [11, 14, 30, 31]. Hennig-Thurau et al. pointed out that consumers can be divided into distinct groups in terms of their motivations for posting electronic word-of-mouth (eWOM) [19]. Accordingly, it was suggested that different strategies should be developed based on different motivation groups. However, existing studies relied on traditional survey methodology to explicitly acquire consumers’ eWOM motivations. From consumers’ perspective, they may be unwilling to answer the survey for the sake of saving efforts or protecting their privacy.

In this paper, we focus on how to implicitly derive consumers’ motivation for writing product reviews online from their behavior data. Specifically, we have identified a set of behavioral features that are significantly correlated with users’ motivations through experimental validation.

In the following, we first introduce related work on WOM (word-of-mouth) motivations in Sect. 2. We then present the details of motivation and behavioral feature identification process in Sect. 3. In Sect. 4, we describe the correlations between consumers’ motivations and their behavioral features. We finally conclude the work and indicate its future direction in Sect. 5.

2 Related Work

Considering that understanding the antecedents of online reviews is the foundation for enhancing users’ willingness to write reviews [17], researchers have devoted attention to the drivers of contribution to online product review. For traditional WOM communication, Dichter first proposed four main motivations of providing positive WOM: product-involvement, self-involvement, other-involvement, and message-involvement [11]. Based on Dichter’s findings, Engel et al. introduced dissonance reduction as an additional motivation [14]. Then, Sundaram et al. pointed out the motivations for generating positive and negative reviews should be different. According to 390 critical-incident reviews, they identified four motivations for positive WOM (i.e., altruism. product-involvement, self-enhancement, and helping the company) and the other four motivations for negative reviews (i.e., altruism, anxiety reduction, vengeance, and advice seeking) [30].

With the development of internet technology, users are provided with a variety of ways for sharing information and opinion (e.g., email, virtual community, web-based opinion paltforms, discussion forums, etc.) [4]. The communication of product reviews has been extended from the word-of-mouth between relatives to the electronic word-of-mouth among publics. Dholakia et al. pointed out the main motivations of virtual community participation are purposive value, self-discovery, maintaining interpersonal connectivity, social enhancement, and entertainment value [10]. Ma and Agarwal also indicated that knowledge contribution in online commnities is strongly linked to people‘s perceived identity verification [27]. Hennig-Thurau et al. investigated the motivations consumers may have in engaging in eWOM communication on web-based opinion paltforms. Through analyzing the responces of more than 2000 consumers, they extracted eight motivations: venting negative feelings, self-enhancement, concern for other consumers, helping the company, advice seeking, platform assistance, social benefits, and economic incentives [19]. In addition, some researchers pay attention to users’motivation for a specific type of product (e.g., movies [9], travel [33]). Table 1 summarizes the motivations of word-of-mouth communication identified in the literatures.

Table 1. Motivations of word-of-mouth communication identified in the literatures [9]
Full size table

Furthermore, Hennig-Thurau et al. analyzed how the eWOM motivations differ among consumers. The results show that consumers are not a homogeneous group in terms of their eWOM motivations, but can be divided into different groups [19]. For example, some customers are referred to as ‘true altruism’, as they are strongly motivated by helping other consumers and companies; while, some customers tend to be ‘self-interested helpers’, as they appear to be driven by economic incentives.

However, existing studies mainly rely on traditional questionnaires to measure users’ motivations, which unavoidably demand a large amount of user efforts and hence impede the development of adaptive design solutions for encouraging eWOM participation. We are thus interested in exploring proper behavioral features to infer users’ motivation for writing online reviews.

3 Method

3.1 Motivation Identification

In order to understand the motivations of writing online product reviews, our review of the literature has led us to suggest eight motivations [19], including: (1) economic incentives, (2) concern for other consumers, (3) helping the seller, (4) “advice seeking”, (5) expressing positive feelings, (6) social benefits, (7) showing connoisseurship, and (8) obtaining prestige.

Then, through an informal discussion with consumers, we added the ninth and tenth dimension of motivations: (9) hard to refuse request and (10) reducing anxiety. Specifically, people usually find it hard to refuse a request made by a relative or friend, even though the benefit may remain unknown, lest being criticized as “lacking human feeling” [7]. Besides, some people also proposed that they feel anxiety or uncomfortable if they have unfinished work. Thus, when such people are informed that there are uncompleted reviews, they would write comments to prevent or reduce distress.

In total, a list of 10 distinct motivations for writing online reviews was developed. To measure these motivations, a set of questions were pre-designed from existing studies, where they were tested and found to have strong content validity and reliability (see Table 2). Each question was responded on a five-point Likert scale ranging from 1 ‘strongly disagree’ to 5 ‘strongly agree’. The higher the score, the stronger the motivation is.

Table 2. Questions to measure consumers’ motivations for writing reviews online
Full size table

3.2 Feature Identification

In terms of literature review, the behavioral features were assessed by five constructs: (1) demographic properties, (2) the valence of reviews, (3) the form of product reviews, (4) whether to answer potential consumers’ questions, and (5) the likelihood of appending reviews.

Users’ Demographic Properties.

There are several demographic properties have been shown related to users’ motivation. Young users are more likely to proactively interact with a social networking site and to be online content creators [22]. Males are traditionally driven by self-efficacy, self-assertion, and achievement orientation [28]. Yoo et al. found that females are more motivated by helping the company and experiencing enjoyment/positive self-enhancement. Also low-income groups are more motivated by desires to vent negative feelings and concerns for other consumers [33].

The Valence of Reviews.

One of the most researched topics in WOM literature is the valence or polarity of messages. Researchers proposed that type of consumer’s motivation does indeed influence the dissemination of the different valence of reviews (i.e., positive and negative) [23]. Specifically, a consumer with a utilitarian orientation (e.g., economic incentives, advice seeking) is more likely to spread negative information. In contrast, one with a hedonic orientation (e.g., social benefits) is more likely to spread positive information.

The Form of Product Reviews.

Based on the findings of [6], it was concluded that motivation also influences the form of reviews (e.g., rating, textual comment or picture). Users with more of an other-directed (i.e., helping other consumers) motivation are more inclined to express themselves by combined use of text and rating. While those with a primarily self-directed motivation express by text only.

Whether to Answer Potential Consumers’ Questions.

In some online shopping websites (e.g., Amazon.com), potential consumers of a product can ask for help from people who have already bought it. We assume that whether to provide advice may vary with different motivations. For example, people with social sharing motivation are more inclined to feel a sense of urgency to connect with others and proactively answer the questions.

The Likelihood of Appending Reviews.

Since the quality problem of some products will not appear in a short time, some online shopping websites (e.g., TaoBao.com) provide consumers with a second chance to post additional reviews after using a period of time. We can argue that consumers with different motivations will have different likelihood to append reviews.

3.3 User Survey Setup

To investigate the relation between motivations and behavioral features, we performed a user survey to collect consumers’ motivations and behavior data. Of the 187 respondents, only those 110 individuals (54 females) who had previously written online reviews were included in data analysis. Table 3 gives the demographic profile of the final sample. All of them are Chinese, with different education backgrounds (51.6% with Bachelor, 35.9% with Master, and 12.5% with PhD) and age ranges (33.6% in the range of 20–30, 46.4% in the range of 30–40, and 20% in the range of above 40).

Table 3. Demographic profile of participants in the user study
Full size table

The behavioral features were concretely determined in the following ways. With respect to the valence of reviews, participants reported the percentages of positive, neutral and negative reviews they have given. Similarly, the form of product reviews was assessed via the average frequencies of posting rating, text and pictures. As for whether to answer potential consumers’ questions, users were required to indicate “Have you ever provided advice to other consumers online?”. The likelihood of appending reviews was obtained by user’s answer to the question “I would like to provide additional comments after using a product.” (given on a five-point Likert scale from 1 ‘strongly disagree’ to 5 ‘strongly agree’).

4 Results

We employed Spearman’s rank coefficient to validate whether the correlations between consumers’ motivations and their behavioral features are significant correlated, because Spearman’s rank coefficient can be applied to both numerical and ordinal variables [15]. The results are shown in Table 4.

Table 4. Correlations between consumers’ motivations and behavioral features (*p < 0.05 and ** p < 0.01)
Full size table

Among users’ demographic properties, on the contrary to literature, we observe that age is significantly correlated with ‘Social benefits’ in a positive way, which suggests that old people are more likely to share experience about a product with others than young people. Education is negatively correlated with users’ motivation for economic incentives, indicating that people with higher education level show less interest in financial incentives. Consistent with the results in literatures, females are more motivated by experiencing enjoyment.

As for the valence of reviews, the proportion of neutral reviews is significantly positively correlated with ‘Hard to refuse requests’, which implies that people who are ashamed to refuse others’ requests tend to preserve an attitude of neutrality. The proportion of negative reviews is negatively correlated with ‘Concern for other consumers’, suggesting that people possess altruistic motivation are not likely to provide positive reviews.

In terms of the form of product reviews, people who give ratings tend to possess the motivation of ‘helping the seller’ and ‘obtaining social benefits’. Moreover, those who further post textual comments and pictures are more inclined to express delight, show connoisseurship and reduce anxiety.

Furthermore, people who once actively answered potential consumers’ questions tend to be more motivated by ‘Concern for other consumers’, ‘Expressing positive feelings’, and ‘Reducing anxiety’. With respect to the likelihood of appending reviews, it shows that people who post additional reviews after a period of time are subject to ‘Helping the seller’, ‘Social benefits’, ‘Obtaining prestige’, ‘Hard to refuse requests’ and ‘Reducing anxiety’.

5 Conclusion

In this paper, we presented an approach to implicitly deriving users’ eWOM motivation from their behavior in online community. Specifically, we first identified a set of behavioral features. Then we have validated the significant correlations between multiple features and users’ eWOM motivation through user survey.

In the future, we will combine all of these above significant features into a unified inference model to demonstrate that a consumer’s motivation for writing online reviews can be inferred by his/her behavioral features. Concretely, we can compare three regression models: Gaussian Process, Pace Regression, and M5 Rules. These findings will lay solid foundation to develop more effective design solutions, which adaptively and accurately stimulate different consumers’ motivations to encourage eWOM participation.