Electronic Commerce Research and Applications 10 (2011) 595–604
Contents lists available at ScienceDirect
Electronic Commerce Research and Applications
journal homepage: www.elsevier.com/locate/ecra
Do males and females differ in how they perceive and elaborate
on agent-based recommendations in Internet-based selling?
Her-sen Doong a,⇑, Hui-chih Wang b
a
b
National Chiayi University, Chiayi, Taiwan, ROC
National Chiao Tung University, Hsinchu, Taiwan, ROC
a r t i c l e
i n f o
Article history:
Received 24 July 2009
Received in revised form 12 December 2010
Accepted 12 December 2010
Available online 21 December 2010
Keywords:
Advertising hierarchy of effects model
Empirical research
Perceived usefulness
Product involvement
Recommendation agents
Shopping task uncertainty
a b s t r a c t
Managers are eager to know whether they should use gender-specific strategies to communicate with
male and female consumers differently to recommendation services in Internet-based selling, as both
groups are worthwhile and profitable to target. This issue has been overlooked in the literature, despite
its importance and the fact that recommendations have become a major online advertising tactic in
recent years. This study is diverse from existing studies in three ways: testing recommendation agent
acceptance in the information context, revealing gender effects and applying a new lens of cognitionconation. By proposing a theoretical model based on the advertising hierarchy of effects model, this study
investigated consumers’ elaboration processes towards recommendation agent advices. Based on 432
members randomly selected from a database of a well-known Internet-based sellers, the study finds that
women consider perceived usefulness of recommendation agent advices to a greater extent than men
while making decisions about the usefulness of recommendations. Further, consumers’ perceived usefulness is more strongly influenced by their perceived shopping task uncertainty than by involvement, while
more-involved consumers have lower levels of perceive shopping task uncertainty.
Ó 2010 Elsevier B.V. All rights reserved.
1. Introduction
Recommender systems (sometimes also called recommender
agents or RAs, for short) are an important facet of Internet-based
sellers’ advertising and promotion strategies due to their ability
to deliver shopping advice, stimulate consumers’ purchase desires
and boost sales (Hung 2005). However, when managers introduce
recommendation services, do they need to take into account that
men and woman may react differently to the advice? This is important because advertising is the major way in which marketers
communicate with different target segments in the traditional
market, and recommendations have played a similar role in the
e-commerce market. Moreover, for many years, gender has been
considered to be the most useful basis for market segmentation because both segments are profitable and easy to target (Darley and
Smith 1995). Consequently, managers must understand whether
there are decision-processing differences between men and
women in order to produce effective recommender system advice
for each segment.
Gefen and Straub (1997) argued that the individual acceptance
literature has ignored the effects of gender, and since then, a
growing number of studies have revealed the impact of gender
⇑ Corresponding author.
E-mail addresses: [email protected] (H.-s. Doong), hcwang1@gmail.
com (H.-c. Wang).
1567-4223/$ - see front matter Ó 2010 Elsevier B.V. All rights reserved.
doi:10.1016/j.elerap.2010.12.005
on users’ acceptance of technology. Significant differences have
been found between men and women in these studies. For example, Venkatesh and Morris (2000) revealed that men consider perceived usefulness to a higher extent than women when making
decisions about using a new technology. Ahuja and Thatcher
(2005), in a post-adoption study, also reported that gender moderates the associations between the perceptions toward work environment factors and trying to be innovative. However, Awad and
Ragowsky (2008) contend that the literature regarding gender differences in e-commerce is still lacking. The current study also
argues that past studies have mainly focused on technology
acceptance at the system level, whilst individuals’ technology
acceptance at the information level, such as elaboration on
recommender system advice, is being neglected. In support of this,
Wixom and Todd (2005) also indicated that the acceptance of
information should be treated differently from the acceptance of
systems.
Consequently, to verify the effects of gender on consumers’
acceptance of recommender system advice at the information
level, this study has developed a model based on the advertising
hierarchy of effects model (Lavidge and Steiner 1961), which proposed that consumers’ advertising information processing can be
delineated as a cognitive-conation process. Distinct from the existing literature in terms of testing recommender system acceptance
in the information context, revealing gender effects, and applying
the new lens of cognition-conation, this study’s findings provide
596
H.-s. Doong, H.-c. Wang / Electronic Commerce Research and Applications 10 (2011) 595–604
interesting and fruitful insights for researchers and practitioners
regarding the process of consumers’ information processing of recommendations. In the following sections, we propose theoretical
model, its support from the literature, and a set of related research
hypotheses. We also will discuss the research methodology, design, statistical analysis and results, and the conclusions and implications of our work.
2. Theory
Shopping advice supported by recommender systems is accepted to be able to assist consumers’ purchase decision-making.
This study proposes that consumers’ cognitive processes can be
tested via several perceptual constructs related to their existing
knowledge. These include the current event, environmental stimuli
in terms of product involvement, perceived shopping task uncertainty and perceived usefulness of recommender system advice.
More specifically, we will test consumer intention to elaborate
on recommender system advice from the perspective of cognition-conation based on the advertising hierarchy of effects model.
Further, we will attempt to verify the effects of gender in decisions
about elaborating on advice and consumers’ recommender system
acceptance at the information level. The model is illustrated in
Fig. 1.
Fig. 1. Research model for recommender system advice service (RA).
2.1. Recommendation agents and consumer acceptance
Many online consumers are eager for the same assistance that is
available in a conventional environment combined with the convenience of shopping online (Holzwarth et al. 2006). Recommender
system advisory services, which involve software agents that
collect the interests and product preferences of individual consumers, and offer shopping advice (Xiao and Benbasat 2007), have been
developed. In practice, popular sellers such as Amazon.com and
Lands’ End have embedded collaborative-filtering systems, which
apply the opinions of like-minded people to produce recommendations. For example, immediately after introducing a product,
Amazon.com provides shopping advice. It asks: ‘‘What do customers ultimately buy after viewing this item? 80% buy the item
featured on this page . . ..’’ The advice offered by Land’s End is in
the form ‘‘You might also like . . .’’ Fig. 2 presents shows the kind
of recommender advice offered at Amazon.com.
As recommender system advisory services are believed to stimulate consumers’ shopping desires and boost sales, research in this
area has gained considerable attention in the past few years. The
computing techniques used to develop recommender systems are
among the most popular issues for research. For example, Liu
and Shin (2005) proposed an approach integrating group decision-making and data mining techniques to evaluate consumer
lifetime values and establish associated rules for giving shopping
advice. Interface design is another central topic. Al-Natour et al.
(2006), for instance, reported that the design of recommender systems can manifest personalities that attract consumers. Individual
acceptance of recommender systems is also vital. Komiak and
Benbasat (2009) used Amazon.com as an example. They demonstrate how the content of advice influences consumer trust in the
recommender system and affects their willingness to use a recommender system to automate their shopping process or as a decision
aid. Qiu and Benbasat (2009) employed a social relationship perspective to the interface design of recommender systems and
investigated the effects of applying human forms and voices on
users’ perceived relationship with a technology artifact. However,
most of the existing research mainly examines acceptance at the
system level. It explores how and why consumers adopt a recommender system as a technological system.
Knowledge of individual consumers’ acceptance of recommender systems at the information level – a form of elaboration
Fig. 2. Recommender system agent (RA) advice service offered at Amazon.com.
H.-s. Doong, H.-c. Wang / Electronic Commerce Research and Applications 10 (2011) 595–604
on the recommender system’s advice – is absent from the existing
studies though. A small but growing number of studies have proposed and empirically verified that, while using technology-based
systems, individuals’ perceptions of the system and information
are distinct constructs and should be validated separately (DeLone
and Mclean 2003, Ahn et al. 2004, Wixom and Todd 2005, Roca
et al. 2006). More importantly, Sussman and Siegal (2003) have
demonstrated that its perceived usefulness will lead to adoption
of this information. In the recommender systems field, the work
of Wang and Benbasat (2005) is the first to shed light on this issue,
and they have noted that their study focuses on ‘‘consumer intentions to adopt recommendation agents to get shopping advice.’’
Even so, they still refer to the construct ‘‘consumers’ intentions
to adopt the agent,’’ which does not differ from the studies investigating recommender system acceptance at the system level. To
further clarify this issue, the current study formulates the construct as ‘‘intention to elaborate on recommender system advice,’’
so as to address consumer intentions to consider recommendations
in their online purchase decision-making process.
2.2. The advertising hierarchy of effects model
Lavidge and Steiner (1961) imitated consumers’ responses to
advertising in the marketplace and proposed that advertising
may be thought of as a force that moves receivers up through a series of steps. (See Fig. 3.) The first two steps of ‘‘awareness and
knowledge’’ are associated with information or ideas, followed by
‘‘liking and preference,’’ which relate to favorable attitudes or feelings toward the product. The final two steps of ‘‘conviction and
purchase’’ serve to generate purchase action. Lavidge and Steiner
further link these six advertising functions directly to the classic
psychological model, which splits behavior into three dimensions:
(1) the cognitive dimension is the intellectual, mental, or ‘‘rational’’
states of consumers’ minds; (2) the affective dimension refers to the
‘‘emotional’’ or ‘‘feeling’’ states; and (3) the conative dimension
indicates the ‘‘striving’’ states, relating to the tendency to treat objects as positive or negative goals.
Although these authors proposed three psychological stages of
behavior, researchers have contended that consumers may undergo only cognitive responses or only affective responses before
reaching the conative stage (Vakratsas and Ambler 1999, Park
et al. 2008). Further, as individuals actively process the environmental stimulus, the cognitive stage preceding the conative stage
refers to the fact that individuals bring their past knowledge and
prejudices to bear in working out how they perceive and identify
with the present events (Shohov 2003). In this process, three perceptual attributes were identified: individual knowledge, the current event and the environmental stimulus. Ultimately, the
results generated from the cognitive stage will shape the conative
stage, which relates to intentions, behaviors and actions (Ray
1973).
To further explore the cognitive stage, this study frames the
knowledge factor as ‘‘product involvement,’’ because it previously
has been shown that higher involvement precedes greater product
information acquisition and improved product understanding
(Hess et al. 2006). Baker et al. (2002) indicate that the involvement
construct and consumer product knowledge are associated, such
Related Behavioral Dimension
Movement Toward Purchase
Cognitive Stage
Awareness, Knowledge
Affective Stage
Liking, Preference
Conative Stage
Conviction, Purchase
Fig. 3. Advertising hierarchy of effects model (Lavidge and Steiner 1961).
597
that consumers who are more involved with one particular product
class also tend to have more knowledge about this product class.
This is also in line with Lavidge and Steiner’s (1961) proposition,
which asserts that the stage of ‘‘knowledge’’ relates to consumers’
understanding of a product’s attributes or benefits, as well as the
ways in which a firm’s specific product or brand differs from its
competitors’ products or brands. Consistent with this, Bian and
Moutinho (2009) pointed out that when consumers’ product
involvement is high, they are more able to distinguish between
one specific brand and its competitors.
The current event factor was framed as ‘‘perceived shopping
task uncertainty’’, to capture the degree of complexity of the ongoing shopping task. This is appropriate because Daft and McIntosh
(1981) indicate that when the present event is unfamiliar, novel
and unexpected, the task variability is high. In support of this,
Chang et al. (2003) assert that task uncertainty is a contingency
factor, and examine this uncertainty in the context of using an
accounting information system. Similarly, Blili et al. (1998) also
examined task uncertainty with the current event of end-user
computing.
Finally, the environmental stimulus factor, which addresses
consumers’ responses to recommendations, was framed as
‘‘perceived usefulness of recommender system advice,’’ since for
consumers, perceptions of advice are only situational, and represent the response to the purchase environment offered by the
Internet-based seller (Wang and Doong 2010). This is because
the importance of the advice is only relevant to the consumer
while shopping on site, and such relevance quickly disappears
when the consumer leaves the seller’ site. In support of this,
Holzwarth et al. (2006) asserted that recommender systems services have been developed to meet consumers’ demand for a shopping environment online, as they feel helpless when facing an
array of complex product categories on an impersonal Web site
and many want the assistance that is available in a conventional
environment. More specifically, Komiak et al. (2005) indicated that
recommender systems advice services functions like a virtual
salesperson, who communicates with consumers and stimulates
their desire to make a purchase.
Together, these three perceptual factors will be able to delineate
consumers’ cognitive processes with respect to elaboration on recommender advice while shopping online. That is, to deal with the
shopping task they are undertaking (current event), consumers
may instantly use their existing knowledge frame (product
involvement) to analyze the product to be purchased. As soon as
they go online to shop, they will interact with the shopping advice
offered by the recommender system (environmental stimulus).
Finally, the interactions between these perceptions will lead to
the intention to elaborate on recommender system advice (conative stage). The supporting literature for each factor is discussed
below in order to build this study’s hypotheses.
2.3. Perceived usefulness
Perceived usefulness has been an important construct in verifying IS adoption since the work of Davis (1989). For example, both
Koufaris (2002) and Gefen et al. (2003) indicated that the more
useful consumers perceived a particular Internet-based seller to
be, the higher their intentions to make purchases from that seller.
In the context of the IS services offered online, Lederer et al. (2000)
revealed that perceived usefulness significantly influenced consumers’ use intentions towards Internet newsgroups. More specifically, in the domain of recommender systems, Kamis and Davern
(2005) reported that perceived usefulness increased consumers’
online decision quality and confidence. Wang and Benbasat
(2005) also indicated that perceived usefulness would enhance
individuals’ intentions to adopt recommender systems to get
598
H.-s. Doong, H.-c. Wang / Electronic Commerce Research and Applications 10 (2011) 595–604
shopping advice. However, the object of perceived usefulness
investigated in past studies is primarily the ‘‘system’’ or the ‘‘technology.’’ Only recently have Sussman and Siegal (2003) proposed
that this relationship should be examined at the information level
and verified that users’ perceived information usefulness would
lead to information adoption. As such, we hypothesize that:
Hypothesis 1. (The Usefulness of Advice Hypothesis). The perceived
usefulness of recommender system advice will
positively affect consumers’ intentions to elaborate
on this advice.
2.4. Gender differences and information processing
For decades, researchers from different disciplines have attempted to delineate the fundamental similarities and differences
between men and women. Bem (1981) argues that men and women use different socially cognitive attributes in information
encoding and problem-solving: hence, their perceptions are also
diverse. In other words, men and women may take unconscious
or internalized actions due to their gender. These differences are
particularly important for advertising and promotion strategies,
and gender has long been utilized as a basis for market segmentation. This is because gender meets the requirements for successful
implementation that not many other variables are able to meet: (1)
men and women are simple to identify; (2) gender segments are
easily accessible, given that they are included within the top few
data that most retail channels collect; and (3) gender segments
are large enough to be profitable (Darley and Smith 1995).
Gender effects are also becoming an increasing crucial topic in
the IS field since the work of Gefen and Straub (1997), who
asserted the importance of examining how men and women may
differ in IS usage. (See Table 1 for a summary.) For example, compared to men, women were found to be more aware of non-verbal
social cues in a computer-mediated environment (Dennis et al.
1999) and to perceive a greater social presence in electronic communication (Gefen and Straub 1997), and women’s decisions
regarding technology acceptance are less influenced by their perceptions of its usefulness (Venkatesh and Morris 2000). Although
these studies have found insightful diversity between men and women in their communication patterns and perceptions with regard
to technology, there has been little investigation of gender-based
differences in accepting information in terms of seller recommendations in the e-commerce environment.
A variety of studies indicate that while processing advertising
information, men and women may use considerably different elaboration strategies to deal with its content. For example, compared to
men, women may be more accurate in decoding non-verbal cues
(Rosenthal and DePaulo 1979), engage in more detailed elaboration
of message content and react to more subtle cues in messages
(Meyers-Levy 1989). Men, in contrast, look for overall message schemas and tend to employ highly available and particularly salient
cues for more detailed processing (Hess et al. 2006). As men are also
more task-oriented (Minton and Schneider 1980), the goal of information processing is to use tactics that aim to maximize efficiency.
In the online shopping context, task orientation may refer to the
accomplishment of shopping tasks. That is to say, men are more
likely than women to apply recommender system advice without
detailed consideration of its content because they inherently prefer
to apply efficient tools to complete the task of shopping. In contrast, while making judgments, women are more comprehensive
processors and exhibit superior information sensitivity (MeyersLevy and Maheswaran 1991). Consequently, women are likely to
consider more carefully whether the recommender advice is useful
for the shopping task, as compared with men. In support of this,
Holbrook (1986) also suggests that gender is a major factor in
moderating consumers’ evaluative judgments. So we assert:
Hypothesis 2. (The Gender Moderation of Usefulness and Advice
Elaboration Hypothesis). Gender will moderate the
association between the perceived usefulness of
recommender system advice and intentions to
elaborate on the advice that is offered.
2.5. Personal involvement
Involvement is widely recognized as the lynchpin of the consumer decision-making process. Zaichkowsky (1986), one of the
most well-cited scholars specializing in personal involvement, indicated that, since 1965, there have been three types of involvement
research. Some have utilized involvement in studying the effectiveness of advertising. Some have examined involvement with consumer purchase decisions or actions such as time spent searching
or number of stores visited. Still others have applied involvement
to examine the link between an individual and a product: in this
area, the research focus is how relevant or important the person
perceives the product category to be. Zaichowsky further asserted
that although each is a different domain of research, the rule of
Table 1
Differences between men’s and women’s conations about IT applications.
Author
Topic
Findings
Zhang et al.
(2009)
SanchezFranco
et al.
(2009)
Hess et al.
(2006)
Awad and
Ragowsky
(2008)
Ahuja and
Thatcher
(2005)
Venkatech
and Morris
(2000)
Gefen and
Straub
(1997)
The role of gender in bloggers’ switching behavior
Female bloggers, rather than males, are more sensitive to satisfaction and less
responsive to attractive alternatives
The influence of trust on commitment and of commitment on loyalty was significantly
stronger for females than for males, while the effects of satisfaction on commitment
and of trust on loyalty were significantly stronger for males
The moderating effect of gender on relationship quality and
loyalty toward Internet service providers
Explores how multimedia vividness and computer-based
social cues can influence involvement with technology
Cultural effect of gender on the relationship between online
word-of-mouth and trust in e-commerce
The influence of the work environment and gender on trying
to innovate with IT
Gender differences in the overlooked context of individual
adoption and sustained usage of technology in the workplace
Testing gender differences that might relate to beliefs and
use of computer-based media
Women report higher levels of involvement with the decision aid
The effect of trust on intention to shop online is stronger for women than for men.
Further, men value their ability to post content online, whereas women value the
responsive participation of other consumers to the content they have posted
The findings confirm that autonomy interacts with overload to determine trying to
innovate with IT and that these relationships vary by gender
Men’s technology usage decisions were more strongly influenced by their perceptions
of usefulness. In contrast, women were more strongly influenced by perceptions of
ease of use and subjective norms
Women and men differ in their perceptions but not use of e-mail
H.-s. Doong, H.-c. Wang / Electronic Commerce Research and Applications 10 (2011) 595–604
thumb between these studies is the construct of involvement is
defined as ‘‘personal relevance.’’ For example, in product class
research, involvement refers to the ‘‘relevance’’ of the product to
the consumer’s needs and values, and consequently there is continuing interest in product information (Zaichkowsky 1985).
In fact, little is known regarding the direct effects of product
involvement on consumers’ conation processes regarding recommender system-based shopping advice. Zaichkowsky (1986)
argues that involvement is more than just importance. Involvement
affects consumers’ information processing at a primary level, as
higher involvement can result in greater information acquisition,
improved understanding, and increased effort (Hess et al. 2006).
When individuals are involved, they pay attention, perceive the
matter to be of importance and behave in a different manner than
when they are not involved (Zaichkowsky 1985). More-involved
consumers will have ongoing interests toward the specific product
(Mittal and Lee 1988). Hence, they will be motivated to gather more
product information and over time will become increasingly
knowledgeable about the product (Zaichkowsky 1986). Given that
more-involved consumers also purchase or use the product more
frequently (Foxall and Bhate 1991), their product experience will
be extensive, so they will not ignore the general shopping advice
proposed by recommender systems. In support, Kramer (2007)
indicated that more-involved consumers are able to evaluate a
product better when shopping online because they continuously
pay greater attention to the specific product. Thus, we propose:
Hypothesis 3. (The Product Involvement and Usefulness of Advice
Hypothesis). Consumers’ product involvement will
negatively affect their perceived usefulness of
recommender system advice.
2.6. Shopping task uncertainty
Task uncertainty is the ‘‘degree to which work to be performed
is difficult to understand and complex’’ (Alexander and Randolph
1985). A particular task may be difficult because uncertainty constitutes the lack of sufficient information to describe a current
state, predict a future situation or estimate the actions needed to
achieve the task (Raiffa 1968). In other words, while shopping online, task uncertainty may refer to the ‘‘degree to which the shopping task to be undertaken is difficult to understand and complex.’’
Such uncertainty, however, may be a frequent occurrence for online buyers. Pavlou et al. (2007), for instance, proposed four antecedents of perceived uncertainty for buyers and verified how
these factors may influence online purchase intention. In support,
Bunn (1993) revealed that shopping task uncertainty was an
important factor determining which purchase decision-making
process consumers would use.
Bruner (1973) argued that uncertainty can be tolerated by using
existing knowledge to predict, infer, estimate, or assume facts in
place of missing information, with some resulting level of confidence and reliability. In other words, consumers who are able to
use their existing information frame to proceed with the current
purchase decision may feel more confident and experience less
shopping task uncertainty (Cowley and Mitchell 2003). Further,
more-involved consumers also will purchase and use the specific
product more frequently. Hence, if compared with less-involved
consumers, they will perceive less difficulty in performing the
shopping task related to that specific product. So we offer:
Hypothesis 4. (The Product Involvement and Shopping Task Uncertainty Hypothesis). Consumers’ product involvement
will negatively affect their perceived shopping task
uncertainty.
599
Zack (2007) asserted that uncertainty can be reduced by acquiring additional information. Thus, consumers who perceive higher
shopping task uncertainty may look for more information to help
them to make purchase decisions with confidence. In the context
of online purchases, consumers who perceive high shopping task
uncertainty may generally respond in two ways. They may look
beyond the Internet-based seller to collect essential product, price
or brand information for this shopping task (Mandel and Johnson
2002). Alternatively, they may browse the handy advice provided
by the Web site (Swaminathan 2003; Kramer 2007). Considering
that more uncertain consumers tend to regard purchase decisions
as more difficult and recommender system services have been
shown to improve the quality of purchase decisions (Wang and
Benbasat 2007), consumers may therefore rely on online recommendations more and find recommender system advice more useful. Thus we hypothesize that:
Hypothesis 5. (The Shopping Task Uncertainty and Usefulness of
Advice Hypothesis). Consumers’ perceived shopping
task uncertainty will positively affect the perceived
usefulness of RA advice.
3. Methods
3.1. Measurement
Four constructs were developed and investigated in this study:
product involvement, perceived shopping task uncertainty, perceived usefulness of recommender system advice and elaboration
intention toward recommender system advice. Constructs were
measured via seven-point Likert scales ranging from ‘‘strongly disagree’’ to ‘‘strongly agree,’’ with the exception of product involvement and perceived shopping task uncertainty, which utilized
seven-point semantic differential scales. Scale items were drawn
from previously validated IS or marketing literature and appropriately reworded to fit the context of recommender systems and online shopping. For each construct, at least three items were
included for adequate reliability.
Product involvement was defined as the perceived relevance of
the products based on individuals’ inherent needs, values and interests. Involvement is usually measured using terms that express
importance, concern or interest associated with the attitude, object,
issue or action. The validity and reliability of Zaichkowsky’s
Personal Involvement Inventory (PII) has also been confirmed by
various studies across different product categories in past decades
(Foxall and Pallister 1998). More recently, Olsen (2007) has adapted
four items from the revised PII to evaluate consumers’ product
involvement. These four items are important/unimportant, boring/interesting, insignificant/significant and irrelevant/concerning.
Findings reported that the validity and reliability were acceptable.
Specifically, Zaichkowsky (1985) indicated that ‘‘important,’’
‘‘relevant’’ and ‘‘significant’’ were cognitive dimensions. ‘‘Interesting,’’ although it was proposed to be an affective dimension by
Zaichkowsky, was argued to be a cognitive dimension in the service
context by Stafford and Day (1995), Celuch and Taylor (1999), and
Bienstock and Stafford (2006). This study thus employed Olsen’s
(2007) four items to evaluate consumers’ product involvement.
Perceived shopping task uncertainty refers to the degree to
which the shopping task to be performed lacks information, is difficult to understand and is complex. It has been measured using a
scale adapted from Barki et al. (1993), and appropriately reworded
to fit the context of online shopping.
The perceived usefulness of recommender system advice was
measured using a four-item Likert scale that examined subjects’
perceptions of the performance, productivity, effectiveness and
600
H.-s. Doong, H.-c. Wang / Electronic Commerce Research and Applications 10 (2011) 595–604
overall usefulness of the advice. The original scale was developed
by Bhattacherjee (2001) and was modified to fit the context of recommender systems. Finally, this study defined elaboration intention towards recommender system advice as the extent to which
consumers are willing to use the recommendation as an aid to help
with their decisions about which product to buy, and the measurement approach used was adapted from Wang and Benbasat (2005).
The initial version of the survey instrument was pre-tested by
ten consumers and two professors who served as expert judges.
The wording of each item was examined and ambiguous items
were rephrased based on the responses and suggestions of these
pilot testers. Table 2 provides the definitions of the various scaled
responses for these constructs.
3.2. Sample
An international company with a well-known brand name in
the retailing channel, operating an e-tailing Web site using the
same brand name, was selected as the survey target. This brand’s
Web site primarily sells information technology-related equipment (e.g., personal computers, projectors, printers, scanners,
monitors and digital cameras) and implements a recommender
system. The recommender system advice service offered by this
seller also applies an embedded collaborative-filtering approach,
which is similar to what Amazon and Land’s End do.
In this study, the sample frame included individuals from the
seller’s consumer database. Two thousand consumers who had
made purchases from the seller within the past month were randomly invited to participate in an online survey. An e-mail invitation with a hyperlink was sent to each of them. The hyperlink
enabled potential respondents to connect directly to the relevant
questionnaire Web page set up on the brand’s Web site. Respondents were asked to fill out the questionnaire according to their
latest purchase experience on the brand’s Web site. Respondents
who completed the online questionnaire were awarded a coupon
worth NT$100 (about US$3.00) that could be redeemed against
their next purchase from the seller’s Web site. To determine
whether participants had the knowledge or experience necessary
to participate in this study, the questionnaire’s introductory section illustrated the recommender system with screenshots from
the brand’s Web site and contained one filter question: ‘‘I am
aware of the recommendation mechanism built into the brand’s
Web site to help me with decisions about which product to purchase.’’ The survey was completed in two weeks and 560 responses
were collected. There were 72 respondents who failed to finish the
questionnaire and 56 respondents who answered ‘‘no’’ to the filter
question. They were removed from the valid sample. Thus, the
effective response rate was 21.6% and 432 useful responses were
harvested. Of the respondents, 69.7% were male and 30.3% were female; 12.7% were students and 87.3% were working people; 80.50%
were at least university-educated. The average age of the respondents was 32.32 (with a standard deviation of 8.31) and the average length of Internet shopping experience was 4.57 years (with a
standard deviation 0.76).
Non-response bias was examined by comparing the study
constructs between early and late respondents (Armstrong and
Overton 1977). No significant differences were found (t-test at
p = 0.05 level), suggesting that non-response bias was not a problem in this study.
4. Results
4.1. Scale validation
Scale validation was assessed by applying confirmatory factor
analysis (CFA) using the partial least squares (PLS) technique. The
software package employed was SmartPLS. According to Fornell
and Larcker (1981), three criteria must be met to conclude that
construct validity is acceptable. First, all indicator factor loadings
should exceed 0.7. Second, the composite reliability should not
be less than 0.8. Third, the average variance extracted (AVE) should
exceed 0.5. In this study, all item loadings exceeded 0.7. (See Table
3.) The composite reliabilities ranged from 0.91 to 0.96 and AVE
ranged between 0.77 and 0.87. Hence, this study’s results met all
requirements of convergent validity. Moreover, Fornell and Larcker
(1981) argued that the square root of AVE for each construct
should exceed the correlation between that and any other construct. The factor correlation matrix shows that the lowest squared
root of AVE was 0.88, while the highest correlation between any
pair of constructs was 0.60 (perceived usefulness and elaboration
intention). (See Table 4.) Therefore, discriminant validity was also
demonstrated.
4.2. Hypothesis assessment
Next, the path significance of each association in the research
model and the variance explained by each path were assessed
Table 2
Summary of scale items.
Construct
Measure items
Please use the series of descriptive words listed below to indicate your level of interest in the product you bought on the brand’s Web site:
Product involvement
PI1
Important . . .Unimportanta
PI2
Boring . . .Interesting
PI3
Insignificant. . .Significant
PI4
Irrelevant. . .Relevant
Perceived shopping task
uncertainty
TC1
TC3
For me, it is ____ what product attributes I should be concerned with and the hierarchy of attributes in order to perform this
shopping task successfully. Hard to identify. . .Easy to identifya
When I feel that I need a well-defined body of information to execute the shopping task, I feel that this information ____.
Exists. . .Does not exist
Overall, I feel that it is ____ to understand the shopping task. Simple . . .Complex
Perceived usefulness of RA
advice
PU1
PU2
PU3
PU4
The recommendation on the brand’s Web site enhanced my effectiveness in finding suitable products
The recommendation on the brand’s Web site greatly enhanced the quality of my shopping judgments
The recommendation on the brand’s Web site enabled me to find suitable products more quickly
Overall, I found the recommendation on the brand’s Web site useful in finding suitable products
Elaboration intention
toward advice
IA1
IA2
IA3
I am willing to use the recommendation on the brand’s Web site as an aid to help with my decisions about which product to buy
I am willing to let the recommendation on the brand’s Web site assist me in deciding which product to buy
I am willing to use the recommendation on the brand’s Web site as a tool to suggest a number of products from which I can
choose
TC2
a
Reverse coded.
601
H.-s. Doong, H.-c. Wang / Electronic Commerce Research and Applications 10 (2011) 595–604
Table 3
Validity, reliability and variance extracted scores.
Construct
AVE
CR
Alpha
Item
Item mean
Standard deviation
Item loading
Product involvement
0.80
0.94
0.91
PI1
PI2
PI3
PI4
4.53
4.49
4.51
4.45
0.85
0.85
0.96
0.82
0.89
0.91
0.86
0.92
Perceived shopping task uncertainty
0.77
0.91
0.84
TC1
TC2
TC3
5.08
5.01
5.08
1.06
1.07
1.04
0.85
0.88
0.89
Perceived usefulness of recommender system advice
0.85
0.96
0.94
PU1
PU2
PU3
PU4
4.65
4.47
4.54
4.51
1.03
1.10
1.05
1.06
0.91
0.92
0.92
0.94
Elaboration intention toward recommender system advice
0.87
0.95
0.93
IA1
IA2
IA3
4.77
4.58
4.81
1.11
1.07
1.08
0.94
0.93
0.93
Legend: AVE, average variance extracted; CR, composite reliability; Alpha, Cronbach’s alpha.
Table 4
Scale properties and correlations.
a
Construct
Mean
Standard deviation
Product involvement (1)
Perceived shopping task uncertainty (2)
Perceived usefulness of recommender system advice (3)
Elaboration intention toward recommender system advice (4)
4.50
5.05
4.54
4.72
0.77
0.93
0.98
1.01
Factor correlationsa
(1)
(2)
(3)
(4)
0.89
0.41
0.45
0.59
0.88
0.51
0.49
0.92
0.60
0.93
Square root of AVE for each construct are show on the main diagonal.
using the PLS approach. First, a main effects model was used to
assess the four associations between product involvement,
perceived shopping task uncertainty, perceived usefulness of
recommender system advice and elaboration intention towards
recommender system advice as specified in Hypotheses 1, 3, 4
and 5. Then, multi-group PLS analysis was performed by comparing differences in the coefficients of the corresponding paths (the
effect of perceived usefulness of recommender system advice on
elaboration intention towards recommender system advice) for
different genders (Chin 2000; Sia et al. 2009).
Fig. 4 shows the standardized path coefficient and path significance for the main effects model; all the hypothesized associations
were found to be strongly significant at p < 0.01. Consumers’ perceived shopping task uncertainty was found to be the strongest
Path
H2: Perceived usefulness of
recommender system advice
elaboration intention towards
recommender system advice.
**
Significant at 0.01 level
predictor of their perceptions of the usefulness of recommender
system advice (b = 0.39), followed by their product involvement
(b = 0.29). These two constructs were able to explain 33% of the
variance in the perceived usefulness of recommender system
advice. In other words, consumers who are more product-involved
would perceive recommender system advice to be less useful while
shopping at the seller’s Web site. Hence, the Product Involvement
and Usefulness of Advice Hypothesis (H3) was supported. Moreover, the higher the consumers’ perceived shopping task uncertainty is, the more useful they perceive recommender system
advice to be. Thus, the Shopping Task Uncertainty and Usefulness
of Advice Hypothesis (H5) was supported.
Consumer product involvement significantly affected their perceived shopping task uncertainty (b = 0.41), explaining 17% of the
tspooled
Female
Path Coefficient
Male
Path Coefficient
Hypothesis
Support
3.66**
0.77
0.50
Supported
Fig. 4. PLS analysis: main and moderating effects of recommender system (RA) advice.
602
H.-s. Doong, H.-c. Wang / Electronic Commerce Research and Applications 10 (2011) 595–604
variance. In other words, the higher the consumer involvement
with the product, the less shopping task uncertainty they
perceived while shopping online. Consequently, the Product
Involvement and Shopping Task Uncertainty Hypothesis (H4) was
supported. Last, consumers’ perceptions of the usefulness of recommender system advice significantly affected their elaboration
intentions towards this advice (b = 0.60), and the former explained
37% of the variance of the latter. Thus, the more useful consumers
perceive the recommender system advice to be, the higher their
elaboration intention towards the recommender system advice.
The Usefulness of Advice Hypothesis (H1) was ultimately
confirmed.
The moderating effect described in the Gender Moderation of
Usefulness and Advice Elaboration Hypothesis (H2) was evaluated
using multi-group PLS analysis. This analysis was conducted by
comparing the corresponding paths for each sub-group (male vs.
female). The t-statistic with the pooled estimator for the variance
was computed as follows (Chin 2000):
Spooled ¼
stronger for women than for men. In the minds of men and women,
the importance of perceiving recommender system advice as useful in aiding a particular online shopping task is different. Our findings also revealed that the psychological lens of cognitive and
conative processing explains the process of consumers’ decisionmaking regarding elaboration on recommender system advice.
Further, our argument that product involvement, perceived usefulness of recommender system advice and perceived shopping task
uncertainty can be used to imitate the perceptual status of consumers’ existing knowledge, current event and environmental
stimulus while the person shops online at the seller’s Web site.
Specifically, perceived usefulness precedes consumers’ elaboration
intentions toward online recommendations. Both involvement and
perceived shopping task uncertainty significantly affect consumers’ perceptions of the usefulness of recommender system advice,
and the latter demonstrates a stronger effect. Further, moreinvolved consumers tend to perceive less shopping task uncertainty than do less-involved consumers. Hess et al. (2006) have
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
f½ðN1 1Þ2 =ðN1 þ N2 2Þ SE21 þ ½ðN2 1Þ2 =ðN1 þ N2 2Þ SE22 g
t spooled ¼ ðPC 1 PC 2 Þ=½Spooled pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ð1=N1 þ 1=N2 Þ
where Spooled is the pooled estimator for the variance, tspooled refers
to the t-statistic with (N1 + N2 2) degrees of freedom, Ni is the
sample size of the dataset for gender i, SEi is the standard error of
path in the structural model of gender i, and PCi is the path coefficient in the structural model of gender i.
The results showed that the path coefficient from perceived
usefulness of recommender system advice to elaboration intention
towards recommender system advice for the female model (path
coefficient = 0.77) is significantly stronger than that for the male
model (path coefficient = 0.50). So, compared to men, women’s
intentions to elaborate on recommender system advice are more
strongly affected by their perceptions of the usefulness of this advice. Thus, the Gender Moderation of Usefulness and Advice Elaboration Hypothesis (H2) was supported (tspooled = 3.66, significant
at p = 0.01 level).
This is an interesting finding, since Venkatesh and Morris
(2000) reported that men’s decisions regarding technology usage
were more strongly influenced by perceived usefulness compared
to those of women. Our study’s findings provide a different viewpoint on recommender system shopping advice. We found that
women, rather than men, are more strongly influenced by perceived usefulness when deciding whether to consider recommender system advice as a shopping aid. This may result because
recommender system acceptance occurs at the information level.
This is different from recommender system acceptance at the
new technology level discussed in past studies. Consequently,
our study’s findings are more in line with studies examining men’s
and women’s information adoption than men’s and women’s technology adoption. For example, Garbarino and Strahilevitz (2004)
also report that women’s online purchase decisions are more
strongly affected by a friend’s recommendation than are those of
men. This finding supports the current study’s results: a friend’s
recommendation is also a type of information.
5. Conclusion
For advertisers and managers of Internet-based sellers, the crucial facet of gender may be differences in how men and women
process promotional information. The current research has revealed that the effect of perceived usefulness on intention to elaborate on recommender system advice while shopping online is
argued that past studies on involvement exploring computermediated technology have seldom utilized a decision-making
viewpoint. Thus, our findings represent new knowledge.
The theoretical contributions of this research include the development of a theoretical model to address and verify two nascent
issues: (1) the consumer decision-making process towards recommender system advice acceptance at the information level via the
cognitive-conative approach, and (2) gender effects on the information processing of recommender system advice. The latter is
important because Venkatesh and Morris (2000) have reported
that in the context of system acceptance, men consider perceive
usefulness to a greater extent than women in making their decisions about new technology use. However, the current study has
empirically demonstrated that at the information acceptance level,
women, rather than men, feel that it is more important to perceive
recommender system advice to be useful before they are willing to
consider this advice in their current shopping. The contrasting
findings indicate that more research is essential to clarify whether
the effects of perceived usefulness on adoption intention between
men and women may also be different depending on whether the
focus is on is a system or information to accept.
Practical implications of this study may be applied to marketing
segmentation in terms of providing more involving recommender
system designs to users based on gender. Today’s advanced technologies, such as cookies, make it easy for Internet-based sellers
to recognize the identity of current visiting members. By identifying the gender of each consumer, sellers will be able to provide
suitable shopping recommendations based on the gender preferences revealed by this study’s findings. For example, perceived usefulness has more influence on women’s intentions to elaborate on
recommender system advice and their nature is to consider more
comprehensive information than men. Consequently, shopping
recommendations offered by recommender systems to female consumers should be carefully formulated to contain rich product
information with supporting evidence so as to enhance their perceptions of argument quality. In support of this, Wang and Doong
(2010) empirically revealed that by designing recommender system advice with different argumentation forms and spokesperson
types, consumer perceptions of argumentation quality and source
credibility may increase and ultimately lead to stronger online
purchase intentions. By following these recommendations, this
recommender system advice strategy is likely to provide appropriate shopping assistance based on gender characteristics.
H.-s. Doong, H.-c. Wang / Electronic Commerce Research and Applications 10 (2011) 595–604
This research is not without its limitations. Only one Internetbased seller case is examined in this study. Future research should
compare multiple online sellers, product categories and other cultural factors to confirm and extend our understanding of the issue
addressed in this study. To verify the effects of these various
factors on the perceived usefulness of recommender system
advice, an experiment could be designed to manipulate the level
of product involvement and task uncertainty to demonstrate their
influence on perceived usefulness.
Acknowledgements
An earlier version of this paper was presented at the 2009 International Conference on Electronic Commerce, where it was initially
discussed and developed. The authors are grateful to chairs, discussant and participants at ICEC 2009, the Editor in Chief of Electronic
Commerce Research and Applications, Rob Kauffman, the Area Editor,
and the three anonymous reviewers for offering constructive
comments. This research was supported by the National Science
Council in Taiwan under grant numbers NSC98-2410-H-415-014MY3 and NSC96-2416-H-194-005-MY2.
References
Ahn, T., Ryu, S., and Han, I. The impact of the online and offline features on the user
acceptance of Internet shopping malls. Electronic Commerce Research and
Application, 3, 4, 2004, 405–540.
Ahuja, M. K., and Thatcher, J. B. Moving beyond intentions and toward the theory of
trying of trying: Effects of work environment and gender on post-adoption
information technology use. MIS Quarterly, 29, 3, 2005, 427–459.
Alexander, J. W., and Randolph, W. A. The fit between technology and structure as a
predictor of performance in nursing subunits. Academy of Management Journal,
28, 4, 1985, 844–859.
Al-Natour, S., Benbasat, I., and Cenfetelli, I. R. The role of similarity in e-Commerce
interactions: The case of online shopping assistants. Journal of the Association for
Information Systems, 7, 12, 2006, 821–861.
Armstrong, J. S., and Overton, T. S. Estimating nonresponse bias in mail surveys.
Journal of Marketing Research, 14, 3, 1977, 396–402.
Awad, N. F., and Ragowsky, A. Establishing trust in electronic commerce through
online word of mouth: An examination across genders. Journal of Management
Information Systems, 24, 4, 2008, 101–121.
Baker, T. L., Hunt, J. B., and Scribner, L. L. The effect of introducing a new brand on
consumer perceptions of current brand similarity: The roles of product
knowledge and involvement. Journal of Marketing Theory & Practice, 10, 4,
2002, 45–57.
Barki, H., Rivard, S., and Talbog, J. Toward an assessment of software development
risk. Journal of Management Information Systems, 10, 2, 1993, 203–225.
Bem, S. L. Gender schema theory: A cognitive account of sex typing. Psychological
Review, 88, 4, 1981, 354–364.
Bhattacherjee, A. Understanding information systems continuance: An expectationconfirmation model. MIS Quarterly, 25, 3, 2001, 351–370.
Bian, X., and Moutinho, L. An investigation of determinants of counterfeit purchase
consideration. Journal of Business Research, 62, 3, 2009, 368–378.
Bienstock, C. C., and Stafford, M. R. Measuring involvement with the service: A
further investigation of scale validity and dimensionality. Journal of Marketing
Theory & Practice, 14, 3, 2006, 209–221.
Blili, S., Raymond, L., and Rivard, S. Impact of task uncertainty, end-user
involvement, and competence on the success of end-user. Information &
Management, 33, 3, 1998, 137–153.
Bruner, J. S. Beyond the Information Given; Studies in the Psychology of Knowing. W.W.
Norton and Company, New York, NY, 1973.
Bunn, M. D. Taxonomy of buying decision approaches. Journal of Marketing, 57, 1,
1993, 38–56.
Celuch, K., and Taylor, S. A. Involvement with Services: An empirical replication and
extension of Zaichkowsky’s personal involvement inventory. Journal of
Consumer Satisfaction, Dissatisfaction, & Complaining Behavior, 12, 1999,
109–122.
Chang, R. D., Chang, Y. W., and Paper, D. The effect of task uncertainty,
decentralization and AIS characteristics on the performance of AIS: An
empirical case in Taiwan. Information & Management, 40, 7, 2003, 691–703.
Chin, W. W. Frequently Asked Questions – Partial Least Squares & PLS-Graph, 2000.
<http://disc-nt.cba.uh.edu/chin/plsfaq.htm>.
Cowley, E., and Mitchell, A. A. The moderating effect of product knowledge on the
learning and organization of product Information. Journal of Consumer Research,
30, 3, 2003, 443–454.
Daft, R. L., and McIntosh, N. B. A tentative exploration into the amount and
equivocality of information processing in organizational work units.
Administrative Science Quarterly, 26, 2, 1981, 207–224.
603
Darley, W. K., and Smith, R. E. Gender differences in information processing
strategies: An empirical test of the selectivity model in advertising response.
Journal of Advertising, 24, 1, 1995, 41–56.
Davis, F. D. Perceived usefulness, perceived ease of use, and user acceptance of
information technology. MIS Quarterly, 13, 3, 1989, 319–339.
DeLone, W., and McLean, E. The DeLone and McLean model of information system
success: A ten year update. Journal of Management Information Systems, 19, 4,
2003, 9–30.
Dennis, A., Kinney, S., and Hung, Y. Gender differences in the effects of media
richness. Small Group Research, 30, 4, 1999, 405–438.
Fornell, C., and Larcker, D. Structural equation models with unobservable variables
and measurement error. Journal of Marketing Research, 18, 1, 1981, 39–50.
Foxall, G. R., and Bhate, S. Cognitive style, personal involvement and situation as
determinants of computer use. Technovation, 11, 3, 1991, 183–199.
Foxall, G. R., and Pallister, J. G. Measuring purchase decision involvement for
financial services: Comparison of the Zaichkowsky and Mittal scale.
International Journal of Bank Marketing, 16, 4/5, 1998, 180–194.
Garbarino, E., and Strahilevitz, M. Gender differences in the perceived risk of buying
online and the effects of receiving a site recommendation. Journal of Business
Research, 57, 7, 2004, 768–775.
Gefen, D., and Straub, D. W. Gender differences in the perception and use of e-mail:
An extension to the technology acceptance model. MIS Quarterly, 21, 4, 1997,
389–400.
Gefen, D., Karahanna, E., and Straub, D. W. Inexperience and experience with online
stores: The importance of TAM and trust. IEEE Transactions on Engineering
Management, 50, 3, 2003, 307–321.
Hess, T. J., Fuller, M. A., and Mathew, J. Involvement and decision-making
performance with a decision aid: The influence of social multimedia, gender,
and playfulness. Journal of Management Information Systems, 22, 3, 2005-6,
15–54.
Holbrook, M. B. Aims, concepts, and methods for the representation of individual
differences in esthetic responses to design features. Journal of Consumer
Research, 13, 3, 1986, 337–347.
Holzwarth, M., Janiszewski, C., and Neumann, M. M. The influence of avatars on
online consumer shopping behavior. Journal of Marketing, 70, 4, 2006, 19–36.
Hung, L. P. A personalized recommendation system based on product taxonomy for
one-to-one marketing online. Expert Systems with Applications, 29, 2, 2005,
383–392.
Kamis, A. A., and Davern, M. J. An exploratory model of decision quality and its
antecedents for category novices using multiple-stage shopping engines. EService Journal, 4, 1, 2005, 3–27.
Komiak, X. S., and Benbasat, I. Understanding customer trust in agent-mediated
electronic commerce, Web-mediated electronic commerce, and traditional
commerce. Information Technology and Management, 5, 1–2, 2009, 181–207.
Komiak, S., Wang, W., and Benbasat, I. Trust building in virtual salespersons versus
in human salespersons: Similarities and differences. E-Service Journal, 3, 3, 2005,
49–63.
Koufaris, M. Applying the technology acceptance model and flow theory to online
consumer behavior. Information Systems Research, 13, 2, 2002, 205–223.
Kramer, T. The effect of measurement task transparency on preference construction
and evaluations of personalized recommendations. Journal of Marketing
Research, 44, 2, 2007, 224–233.
Lavidge, R. J., and Steiner, G. S. A model for predictive measurements of advertising
effectiveness. Journal of Marketing, 25, 4, 1961, 59–62.
Lederer, A. L., Maupin, D. J., Sena, M. P., and Zhuang, Y. The technology acceptance
model and the World Wide Web. Decision Support Systems, 29, 3, 2000,
269–282.
Liu, D. R., and Shih, Y. Y. Integrating AHP and data mining for product
recommendation based on customer lifetime value. Information &
Management, 42, 3, 2005, 387–400.
Mandel, N., and Johnson, E. J. When web pages influence choice: Effects of visual
primes on experts and novices. Journal of Consumer Research, 29, 2, 2002, 235–
245.
Meyers-Levy, J. Gender differences in information processing: A selectivity
interpretation. In P. Cafferata, A. Eind, and M. Tybout (eds.), Cognitive and
Affective Responses to Advertising, Lexington, Massachusetts, MA, 1989.
Meyers-Levy, J., and Maheswaran, D. Exploring differences in males’ and females’
processing strategy. Journal of Consumer Research, 18, 1, 1991, 63–70.
Minton, H. L., and Schneider, F. W. Differential Psychology. Waveland Press, Prospect
Heights, IL, 1980.
Mittal, B., and Lee, M. S. Separating brand-choice involvement from product
involvement via consumer involvement profiles. Advances in Consumer Research,
15, 1, 1988, 43–49.
Olsen, S. O. Repurchase loyalty: The role of involvement and satisfaction. Psychology
& Marketing, 24, 4, 2007, 315–341.
Park, J., Leslie, S., and Sharron, L. J. Cognitive, affective and conative responses to
visual simulation: The effects of rotation in online product presentation. Journal
of Consumer Behaviour, 7, 1, 2008, 72–87.
Pavlou, P. A., Liang, H., and Xue, Y. Understanding and mitigating uncertainty in
online exchange relationships: A principal-agent perspective. MIS Quarterly, 31,
1, 2007, 105–136.
Qiu, L., and Benbasat, I. Evaluating anthropomorphic product ecommendation
agents: A social and relational perspective to designing information systems.
Journal of Management Information Systems, 25, 4, 2009, 145–181.
Raiffa, H. Decision Analysis: Introductory Lectures on Choices under Uncertainty.
Addison Wesley, Reading, MA, 1968.
604
H.-s. Doong, H.-c. Wang / Electronic Commerce Research and Applications 10 (2011) 595–604
Ray, M. L. Marketing communication and hierarchy of effects. In P. Clarke and
Beverly Hills (eds.), New Models of Mass Communication Research, Sage
Publications, Thousand Oaks, CA, 1973, 147–176.
Roca, J. C., Chiu, C. M., and Martínez, F. J. Understanding e-learning continuance
intention: an extension of the technology acceptance model. International
Journal of Human-Computer Studies, 64, 8, 2006, 683–696.
Rosenthal, R., and DePaulo, B. M. Sex differences in accommodation in nonverbal
communication. In R. Rosenthal and G. Oelgeschlager (eds.), Skill in Nonverbal
Communication: Individual Differences, Hain, Publishers, Inc., Cambridge, MA,
1979, 68–103.
Sanchez-Franco, M. J., Ramos, A. F. V., and Velicia, F. A. M. The moderating effect of
gender on relationship quality and loyalty toward Internet service providers.
Information & Management, 46, 3, 2009, 196–202.
Shohov, S. P. Topics in Cognitive Psychology. Nova Science publishers, Inc., New York,
NY, 2003.
Sia, C. L., Lim, K. H., Leung, K., Lee, M. K. O., Huang, W., and Benbasat, I. Web
strategies to promote internet shopping: Is cultural-customization needed? MIS
Quarterly, 33, 3, 2009, 491–512.
Stafford, M. R., and Day, E. Retail services advertising: The effects of appeal,
medium, and service. Journal of Advertising, 24, 1, 1995, 57–71.
Sussman, S. W., and Siegal, W. S. Informational influence in organizations: An
integrated approach to knowledge adoption. Information Systems Research, 14, 1,
2003, 47–65.
Swaminathan, V. The impact of recommendation agents on consumer evaluation
and choice: The moderating role of category risk, product complexity, and
consumer knowledge. Journal of Consumer Psychology, 13, 12, 2003, 93–101.
Vakratsas, D., and Ambler, T. How advertising works: What do we really know?
Journal of Marketing, 63, 1, 1999, 26–43.
Venkatesh, V., and Morris, M. G. Why don’t men ever stop to ask for directions?
Gender, social influence, and their role in technology acceptance and usage
behavior. MIS Quarterly, 24, 1, 2000, 115–139.
Wang, W., and Benbasat, I. Trust in and adoption of online recommendation agents.
Journal of the Association for Information Systems, 6, 3, 2005, 72–101.
Wang, W., and Benbasat, I. Recommendation agents for electronic commerce:
Effects of explanation facilities on trusting beliefs. Journal of Management
Information Systems, 23, 4, 2007, 217–246.
Wang, H. C., and Doong, H. S. Online customers’ cognitive differences and their
impact on the success of recommendation agents. Information & Management,
47, 2, 2010, 109–114.
Wixom, B. H., and Todd, P. A. A theoretical integration of user satisfaction and
technology acceptance. Information Systems Research, 16, 1, 2005, 85–102.
Xiao, B., and Benbasat, I. E-commerce product recommendation agents: Use,
characteristics, and impact. MIS Quarterly, 31, 1, 2007, 137–209.
Zack, M. H. The role of decision support systems in an indeterminate world. Decision
Support Systems, 43, 4, 2007, 1664–1674.
Zaichkowsky, J. Conceptualizing involvement. Journal of Advertising, 15, 2, 1986,
4–34.
Zaichkowsky, J. L. Measuring the involvement construct. Journal of Consumer
Research, 12, 3, 1985, 341–352.
Zhang, K. Z. K., Lee, M. K. O., Cheung, C. M. K., and Chen, H. Understanding the role of
gender in bloggers’ switching behavior. Decision Support Systems, 47, 4, 2009,
540–546.
This article appeared in a journal published by Elsevier. The attached
copy is furnished to the author for internal non-commercial research
and education use, including for instruction at the authors institution
and sharing with colleagues.
Other uses, including reproduction and distribution, or selling or
licensing copies, or posting to personal, institutional or third party
websites are prohibited.
In most cases authors are permitted to post their version of the
article (e.g. in Word or Tex form) to their personal website or
institutional repository. Authors requiring further information
regarding Elsevier’s archiving and manuscript policies are
encouraged to visit:
http://www.elsevier.com/copyright
Author's personal copy
International Journal of Information Management 30 (2010) 493–501
Contents lists available at ScienceDirect
International Journal of Information Management
journal homepage: www.elsevier.com/locate/ijinfomgt
Argument form and spokesperson type: The recommendation
strategy of virtual salespersons
Hui-Chih Wang a,1 , Her-Sen Doong b,∗
a
b
Department of Information Management, National Chung Cheng University, 168 University Rd., Min-Hsiung, Chiayi 621, Taiwan
Department of Management Information Systems, National Chiayi University, 580 Sinmin Rd., Chiayi 60054, Taiwan
a r t i c l e
i n f o
Keywords:
Recommendation agents
Spokesperson
Argument quality
Source credibility
a b s t r a c t
A recommendation agent (RA) at a Webstore is believed to replace the functions of communicating and
facilitating purchases for which conventional salespersons are responsible in traditional stores. Seeing
the RA as a virtual salesperson, this study proposed a model to demonstrate the online persuasion process between RAs and customers. Specifically, by integrating Toulmin’s model of argumentation with the
spokesperson strategy from marketing theory, this study developed a model illustrating that Webstore
managers may be able to manipulate the argument form and spokesperson type used in RAs’ recommendations to strengthen customers’ online purchase intentions through influencing their perceived
argument quality and source credibility. Based on data derived from 270 subjects who participated in
a 3 × 3 laboratory experiment, findings supported five out of six hypotheses. This study has broadened
the scope of RA studies by utilizing different theoretical foundations other than the trust-assuring issues
that were widely discussed in the past. Findings were able to provide crucial practical implications for
both scholars and Webstore managers.
© 2010 Elsevier Ltd. All rights reserved.
1. Introduction
The sheer volume of material now available on the Internet –
the total number of Websites surpassed 230 million in October
2009 (Netcraft, 2009) – means that customers are simply overwhelmed by the amount of information they are being presented
with (Aggarwal & Vaidyanathan, 2003), leading to browsing problems such as getting lost and spending excessive time searching
for the information they need (Edmunds & Morris, 2000). The convenience of accessing useful information online has unfortunately
become a heavy duty, particularly for those who want to expend
minimal effort when shopping (Hansen, Jensen, & Solgaard, 2004;
Heinze & Hu, 2006). In particular, customers feel helpless when
shopping online, since they may face complex product categories
at impersonal Webstores, and many want the customer assistance
that is available in a conventional environment combined with the
convenience of shopping in a Webstore (Holzwarth, Janiszewski, &
Newmann, 2006). To meet such needs, online retailers have developed the concept of the recommendation agent (RA), viewed as a
sort of ‘virtual salesperson’.
∗ Corresponding author. Tel.: +886 5 2732892; fax: +886 5 2732893.
E-mail addresses: [email protected] (H.-C. Wang),
[email protected] (H.-S. Doong).
1
Tel.: +886 5 2720411x34613; fax: +886 5 2721501.
0268-4012/$ – see front matter © 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijinfomgt.2010.03.006
RAs use product knowledge – either hand-coded knowledge
provided by experts or “mined” knowledge learned from users’
behavior (Kim, Kim, Oh, & Ryu, 2009) – to create recommendations and help customers to make their purchase decisions
(Frias-Martinez, Chen, & Liu, 2009; Frias-Martinez, Magoulas, Chen,
& Macredie, 2006). As more and more B2C Webstores are using RAs
in an attempt to offer customers a better online purchase experience and induce them to make purchase decisions, answers to
the following questions are essential for Webstore managers: Do
RA recommendations truly have an impact on customers’ purchase
intentions at Webstores, as managers expect? If they do, how can
this impact be strengthened so as to generate higher purchase
intentions among online customers? If not, how can such impact
be created to avoid wasted efforts in developing a RA?
Some researchers have addressed these questions via the lens of
facilitating online customers’ trust in RAs and have reported fruitful results (e.g. Komiak & Benbasat, 2006; Lim, Sia, Lee, & Benbasat,
2006; Wang & Benbasat, 2005). However, investigations from different perspectives regarding how RAs may influence customers’
online purchase intentions remain scarce. In fact, popular Webstores such as Barnesandnoble.com have made great efforts to
strengthen RAs’ influence on customers’ purchase intentions via
different types of recommendation, from “Customers who bought
this also bought. . .” to “Ten bestsellers in 2009”. A single recommendation alone, such as “We recommend this item!”, which is
commonly used at various Webstores, may not convince online cus-
Author's personal copy
494
H.-C. Wang, H.-S. Doong / International Journal of Information Management 30 (2010) 493–501
Fig. 1. Argument analysis.
tomers enough to browse the recommended product immediately
or make an instant purchase. Rather, customers may be keen to
know how the recommendation has been developed and why they
should trust it, because persuasion relies on the receiver’s understanding and belief in the message (Hovland, Janis, & Kelly, 1953).
The extent to which customers are able to distinguish good message arguments from bad ones is defined as the perceived argument
quality (Petty, Cacioppo, & Schumann, 1983) and is reported to be
a major factor in changing customers’ attitudes (Petty & Cacioppo,
1986). Further, the literature has also indicated that the strategy
used in the claim “Customers who bought this also bought. . .” was
one of the most useful customer spokesperson tactics (e.g. Ohanian,
1990).
To examine the persuasion process that exists between online
RAs and customers, this study has incorporated the model of
argumentation (Toulmin, 1958) with the concept of spokesperson
strategy from the marketing discipline and proposed that managers
may be able to shape customers’ online purchase intentions via the
strategic use of different argument forms and diverse spokesperson types to affect customers’ perceptions of the argument quality
and source credibility of RAs’ recommendations. Building on a
close inspection of the RA strategies applied at leading Webstores
and sound theoretical foundations, the current study will be able
to provide rich insights that are distinct from the trust-assuring
viewpoints of prior RA studies. The following sections discuss the
theoretical foundation, research methodology, data analysis processes and finally the findings and the conclusions drawn from
them.
• The warrant is the statement authorizing the movement from the
data to the claim (Mason & Mitroff, 1981). In the aforementioned
example, a warrant can be integrated into the message so as to
link the claim and data: “Every person born in the United States
is a legally citizen of the United States”.
• The backing provides additional lawful or statistical evidence
to support the argument. It usually refers to established principles, such as statistical principles or laws (Mason & Mitroff,
1981). In the foregoing example, a backing statement can be
added to enhance the argument quality of the message: “According to Amendment 14 of the Constitution of the United States of
America, proposed on June 13, 1866, and ratified on July 9, 1868,
respectively, ‘Every person born or naturalized in the United
States, and subject to the jurisdiction thereof, is a citizen of the
United States and of the State wherein s(he) resides”’.
Decades later, Petty et al. (1983) referred to argument quality as the distinction between good and bad message arguments
and Petty and Wegener (1999) further operationalized it as the
argument content relative to argument strength. Argument quality is suggested to be closely related to the measures of persuasion
(Krishnamurthy & Sivaraman, 2002) and is a primary determinant
of informational influence when customers are keen to purchase
(Petty & Cacioppo, 1986). The importance of customers’ perceived
argument quality of salespeople’s recommendations in influencing
their purchase intentions is thus established.
In fact, a similar persuasion process also occurs in online shopping contexts, where RAs have replaced the salesperson’s functions
of providing recommendations. As there is no face-to-face contact
online, argument forms that describe what a system does, how it
works and why its actions are appropriate are crucial to facilitate
the interaction between customers and RA recommendations. Only
recently have a small number of researchers started to apply Toulmin’s model of argumentation to examine customers’ Webstore
purchasing. For example, Kim and Benbasat (2006) have used this
model to test the effects of three forms of trust-assuring arguments
(claim only, claim plus data, and claim plus data and backing) on
customer trust in a B2C Webstore in a laboratory experiment. Findings indicated that the latter two forms (claim plus data and claim
plus data and backing) both increased customers’ trust towards the
Webstore, while the former (claim only) did not. In addition, the
most detailed argument form, which included claim plus data and
backing, led to the highest level of trust towards the Webstore.
2.2. Spokesperson strategy and source credibility
2. Theoretical foundation
2.1. Toulmin’s model
Toulmin (1958) proposed four interrelated argument elements
that commonly appear in daily communication: the claim, data,
warrant, and backing. He further argued that a good quality of argument can be created by providing strong justification for a claim.
Fig. 1 shows how a good argument quality can be formed using
these four elements.
• The claim, which refers to “assertions or conclusions put forward
for general acceptance” (Ye & Johnson, 1995), provides valuable
conclusions. For example, if a person tries to convince others that
s(he) is an American citizen, the claim will be “I am an American
citizen.”
• The data is the “grounds” or facts appealed to as a foundation
for the claim (Mason & Mitroff, 1981). In the above example, the
claim’s support can be enhanced if the following sustaining data
is added: “Given that I was born in Boston, Massachusetts.”
Spokespersons have become prevalent in advertising (Agrawal
& Kamakura, 1995), and are widely used to attract attention
to brands and products. The popularity of spokesperson advertising is due to the advertisers’ belief that messages delivered
by well-known people may achieve a higher level of brand and
advertisement recall (Stafford, Stafford, & Day, 2002) and the
product-image tends to be more favorable when celebrities are
used (Atkin & Block, 1983). When a message is delivered by the
right spokesperson, a higher degree of attention may be achieved
(Ohanian, 1991), and the persuasion effect is also more likely to
occur (Fabrigar, Priester, Petty, & Wegener, 1998).
The spokesperson is also reported to be one of the sources in
credibility research (Wathen & Burkell, 2002). Source credibility
refers to the extent to which the source is perceived to possess
expertise relevant to the topic of communication and is believed
to give an objective opinion on the topic (Ohanian (1990). Source
credibility has been widely examined in advertising and marketing
studies to delineate consumers’ purchase decision-making processes. Recently, information system studies have also applied this
factor to examine its effect on users’ adoption intentions towards
Author's personal copy
H.-C. Wang, H.-S. Doong / International Journal of Information Management 30 (2010) 493–501
495
Fig. 2. Research model.
an information system. For example, Bhattacherjee and Sanford
(2006) have proposed and empirically verified that the source credibility of messages has a positive influence on potential users’
attitudes toward the acceptance of informational technology.
H1 . Different argument forms will lead to different customer
perceptions of the argument quality of the RA recommendation
provided at the Webstore.
3. Research model and hypotheses
McKnight and Kacmar (2006) have asserted that information
credibility is essential to Webstore operation: customers who do
not believe the credibility of the Webstore’s information are not
likely to act on the RA’s recommendations. As a system’s credibility can be established via evidence of how a recommendation has
been derived (Wang & Benbasat, 2007), customers’ perceptions of
the source credibility of a RA may be strengthened if additional
justification of how the recommendation has been developed and
why customers should believe in it were added to the claim proposed in the RA recommendation. For example, if the claim “We
recommend book A” can incorporate supporting justification such
as “because customers who have bought the book you are currently viewing also bought book A”, customers’ perceived source
credibility may be enhanced, since the additional evidence of how
the recommendation is generated is revealed. In support, Ye and
Johnson (1995) reported that the more detailed the explanation,
the higher the user’s level of belief in the recommendation. They
also found that different levels of detail in arguments may influence users’ confidence in advice generated by the knowledge-based
system. Accordingly, it is hypothesized that:
The major function of salespeople in the conventional market
is to provide recommendation arguments to stimulate customers’
purchase intentions and convince them to buy (Simintiras &
Cadogan, 1996). As (1) arguments that involve more detailed reasons for an action are found to facilitate the recall of that action
better than those without a detailed rationale (Stein, Littlefield,
Bransford, & Persampieri, 1984), and (2) customers who have a
higher message understanding are persuaded more easily than
those who have a lower message understanding (Wood & Kallgren,
1988), skilful salespeople recognize that detailing the benefits of
specific products is a useful strategy to facilitate customers’ purchase interests.
In the information systems and e-marketplace fields, Benbasat
and his colleagues have also reported similar results. For example,
Mao and Benbasat (2000) have indicated that a knowledge-based
system may supply three different explanation types (i.e. why,
how, and strategic), which will lead to different system adoption intentions. Moreover, Wang and Benbasat’s (2007) empirical
findings revealed that different explanation types might influence customers’ trust beliefs in RAs to different levels. However,
these studies have focused on the trust issue: that is, what are the
antecedents and results of customers’ trust in RAs? This study, in
contrast, would like to demonstrate the process by which, by acting as virtual salespersons, RAs may persuade online customers to
take up their purchase recommendations via the lenses of argument quality and source credibility. Furthermore, by applying
Toulmin’s (1958) model of argumentation and the statement that
spokesperson type will affect source credibility from the marketing discipline (e.g. Ohanian, 1990), this study proposed the research
model depicted in Fig. 2.
3.1. Effects of argument form
According to Toulmin (1958), by providing stronger and more
supporting justification for the claim offered in a RA recommendation, better argument quality will be created. To verify this issue,
the current study tested three forms of argument: claim only (CO),
claim plus data and warrant (CDW), and claim plus data and backing
(CDB). CO is the simplest RA recommendation: for example, “We
recommend item A”. CDW is a RA recommendation that integrates
further data and warrant to support the claim. CDB is a RA recommendation that includes additional statistical evidence to support
the claim. It is worth mentioning that a warrant can be left unexpressed when backing is used (Kim & Benbasat, 2006). Based on the
preceding discussion, it is hypothesized that:
H1a . Customer perceptions of the argument quality will be arrayed
as CDB > CDW > CO.
H2 . Different argument forms will lead to different customer perceptions of the source credibility of the RA recommendation.
H2a . Customer perceptions of the source credibility will be arrayed
as CDB > CDW > CO.
3.2. Effects of spokesperson
While using the right spokesperson to deliver the message, attitude formation may be more strongly influenced by the argument
quality (Krishnamurthy & Sivaraman, 2002). For example, Wheeler,
Petty, & Bizer (2005) indicated that customers who read messages
that matched their self-schemata elaborated more on the persuasive message than those who read messages that mismatched
their self-schemata. That is to say, in the former circumstances,
customers are likely to perceive a higher extent of argument quality and be more willing to process the messages. In support of
this, a celebrity-endorsed advertisement for a brand was found
to generate consistently higher ratings for the advertisement and
product than a non-celebrity-endorsed advertisement (Atkin &
Block, 1983). In other words, customers may perceive a higher level
of argument quality for the celebrity-endorsed advertisement and
hence give a higher rating to these advertisements than to those
without the celebrity. It is thus hypothesized that:
H3 . Using a spokesperson will lead to higher perceived argument
quality of the RA recommendation than not using a spokesperson.
Author's personal copy
496
H.-C. Wang, H.-S. Doong / International Journal of Information Management 30 (2010) 493–501
Table 1
Number of subjects in each experimental treatment.
Spokesperson type
Webstore itself
Customer
Expert
Argument form
Claim only (CO)
Claim plus data and warrant (CDW)
Claim plus data and backing (CDB)
30
30
30
30
30
30
30
30
30
In the context of the e-marketplace, Lim et al. (2006) empirically verified that a customer spokesperson will significantly shape
online customers’ trust, which was asserted to be one of the most
common dimensions of credibility (Hovland et al., 1953). More
specifically, Wathen and Burkell (2002) asserted that source characteristics, such as the spokesperson’s expertise, will effectively
increase the source credibility of RA recommendations. Senecal
and Nantel (2004) further reported that using expert and customer spokespersons will effectively increase online customers’
perceived credibility of recommendations. Accordingly, based on
the foregoing discussion, it is hypothesized that:
H4 . Using a spokesperson will lead to higher customer perceptions
of the source credibility of the RA recommendation than not using
a spokesperson.
3.3. Effects of argument quality
Sussman and Siegal (2003) investigated the evaluation process
by which knowledge workers assess the usefulness of received
information and how this process may influence their adoption of
this information. Empirical findings indicated that the accepters’
professional knowledge and their degrees of involvement would
shape their evaluation. If knowledge workers perceive an argument
to be of high quality, they will feel that this information is useful,
which will ultimately affect their intentions to adopt the information. Further, Bhattacherjee and Sanford (2006) examined the
acceptance of document management systems and reported that
argument quality was a significant determinant of the intention to
adopt and use the information system. Furthermore, Tam and Ho
(2005) tested the interactive process between the Webstore and
customers and revealed that different types of persuasion arguments ultimately lead to different levels of adoption of personal
services provided by the Webstore. As Komiak and Benbasat (2006)
argued that the intention to adopt RAs in reality is reflected by the
intention to deliberate upon the recommendations provided by RAs
(i.e. purchase the items recommended by them), it is hypothesized
that:
H5 . Customer perceptions of the argument quality of the RA recommendation will positively affect their purchase intentions at the
Webstore.
3.4. Effects of source credibility
The faceless and impersonal nature of the Webstore makes it
more difficult to build source credibility with regard to product
information, as online products cannot be handled and checked in
person (Dellarocas, 2003). As previously discussed, customers are
not likely to act on the RA’s recommendation if they do not regard
it as credible (McKnight & Kacmar, 2006). Customers who are convinced by the credibility of the recommendation will be encouraged
to look differently at the subject to which it refers (Alexander, Fives,
Buehl, & Mulhern, 2002). In fact, the customer’s perceptions of the
source credibility will affect their intention to accept the recommendation (Tseng & Fogg, 1999). Further, a highly credible source is
more effective than a less credible source in producing positive attitude changes towards the position advocated and inducing more
behavioral changes (Sternthal, Phillips, & Dholakia, 1978). That is,
the source credibility of the RA recommendation will affect consumers’ willingness to use the RAs as an aid to help with their
decisions about which product to buy. Thus, it is hypothesized that:
H6 . Customer perceptions of the source credibility of the RA recommendation will positively affect their purchase intentions at the
Webstore.
4. Research methodology
4.1. Subjects
A 3 × 3 between-subject full-factorial design was employed.
The two independent variables were argument form (claim only,
claim plus data and warrant, and claim plus data and backing) and
spokesperson type (Webstore itself, expert, and customer). A total
of 270 volunteer undergraduate students at the Business School of
a National University in Taiwan were recruited for the study. Their
average age was 21.9 years, and 51.5% were male. More than half
of the subjects (50.7%) had purchased books online in past three
months. Subjects were randomly assigned to the nine experimental
treatments (see Table 1). Only one set of the nine RA recommendations was presented to each subject.
4.2. Experimental task
In this experiment, subjects were assigned a task involving
information collection activities in relation to a book purchase they
may have experienced in their daily life – buying a book for a friend
whose birthday was approaching. Before the experiment started,
each subject was told that this friend was presently reading a book
entitled “How Would You Move the Eiffel Tower?” and loved to read
books on business and investing. Subjects were asked to browse a
Web bookstore to see whether there was any book they would consider to be a suitable birthday gift for this friend. If subjects did not
find anything they thought suitable, they were allowed to complete
the session without buying a book.
Books were chosen as the context of the experiment for three
reasons. First, the book is one of the most popular products in
B2C commerce. Second, the price range of books was affordable
to the subjects in this study (i.e. students) and buying books from
Webstores is a common activity among students in Taiwan. Third,
previous studies have also used books as the experimental product (e.g. Koufaris, 2002; Lai & Yang, 2000; Lim et al., 2006). Based
on these factors, it was believed that the behaviors of the subjects
would mirror those of actual online customers.
4.3. Experimental product and Webstore
A fictional book entitled “How Would You Move the Eiffel Tower?”
was used in this experiment to avoid possible bias or prior perceptions of an existing book. Moreover, in order to assure realism
and the ability to generalize findings, a fictional Web bookstore,
eBooks.com.tw, was created and used for this experiment only. The
basic book Webpage (see Fig. 2) comprised basic book information
(including the title of the book, author, publisher, publication date,
Author's personal copy
H.-C. Wang, H.-S. Doong / International Journal of Information Management 30 (2010) 493–501
497
Argument form: Three forms of argument (claim only, claim
plus data and warrant, and claim plus data and backing), offering
different levels of information, were used to create the RA recommendation.
• Claim only (CO): “We recommend The Art of Thinking: Conceptual
and Creative Thinking into a Practical Business Plan”.
• Claim plus data and warrant (CDW): We recommend “The Art of
Thinking: Conceptual and Creative Thinking into a Practical Business Plan” because customers who have bought (browsed) the
book you are viewing now, i.e. “How Would You Move the Eiffel
Tower” also have bought (browsed) the book entitled “The Art of
Thinking”.
• Claim plus data and backing (CDB): “We recommend The Art of
Thinking: Conceptual and Creative Thinking into a Practical Business
Plan” because 30% of customers who have bought and 60% who
have browsed the book you are viewing now, i.e. “How Would You
Move the Eiffel Tower?” have also bought and viewed the book
entitled “The Art of Thinking”, respectively.
In addition, three types of spokesperson were employed to generate the RA recommendation in each treatment:
Fig. 3. The basic book Webpage of eBooks.com.
and price), an introduction to the content and a picture of the book.
All the Web pages used in the experiment had the same format
and consistent book information, font size, image size and color
scheme.
The book entitled “How Would You Move the Eiffel Tower?” was
categorized under Business & Investing. On its description Webpage, a RA recommendation was displayed for another fictional
book entitled “The Art of Thinking: Conceptual and Creative Thinking into a Practical Business Plan”, which was also categorized under
Business & Investing. It should be noted that a RA recommendation
was displayed only on the Webpage of the “Eiffel Tower” book and
not on any of the other book web pages created for the fictional
eBook.com site developed for the experiment. In other words, only
when subjects browsed the “Eiffel Tower” Webpage would they see
a RA recommendation for “The Art of Thinking” presented in different forms according to the different experimental treatments (see
Fig. 3 for an example of an experimental treatment). A monitoring system was set up to ascertain whether all subjects had visited
this specific Webpage of the “Eiffel Tower” book. Results indicated
that all 270 subjects had browsed this specific Webpage before they
completed the subsequent questionnaire.
4.4. Experimental environment
The system used in the experiment was designed explicitly for
this study. To avoid potential confounding biases, the experiment
was conducted in a controlled laboratory setting. Fifty identical
Pentium IV personal computers with 17-in. LCD monitors were set
up in a laboratory. Each computer was pre-assigned to run one of
the nine experimental treatments. To ensure a consistent and high
network speed for all subjects, the Webstore was installed on a
Win2003 server in the same local area network as the personal
computers in the laboratory. Internet Explorer 6.0 was set as the
constant Internet browser for all subjects.
4.5. Independent variables
The nine experimental treatments of a RA recommendation
were designed according to two independent variables: argument
form and spokesperson type.
• Webstore itself: “eBooks.com strongly recommends” (see Fig. 3).
• Expert: in this experiment, the book reviewer of The Chinatimes
(equivalent to The New York Times in the USA) was regarded as the
expert spokesperson. Thus, the spokesperson message in Fig. 3
was “Book reviewers from The Chinatimes strongly recommend.”
• Customer: in this experiment, readers at eBooks.com were
regarded as the customer spokesperson. Thus, the spokesperson
message in Fig. 3 was “Readers at eBooks.com strongly recommend.”
4.6. Dependent variables
Three dependent variables (argument quality, source credibility,
and purchase intention) were measured using a Likert scale with
anchors from 1 = strongly disagree to 7 = strongly agree. Argument
quality is a broad term that distinguishes between good and bad
message arguments (Petty et al., 1983). Source credibility refers to
the extent to which the source is perceived as possessing expertise relevant to the communication topic and can be trusted to give
an objective opinion on the subject (Ohanian, 1990). Bhattacherjee
and Sanford’s (2006) scales were adapted to measure argument
quality and source credibility. Purchase intention was used to measure the subject’s intention to purchase the recommended book,
“The Art of Thinking: Conceptual and Creative Thinking into a Practical Business Plan”, at eBooks.com. Four items adapted from Lim
et al. (2006) were used to measure the purchase intention. Table 2
shows the scale items for the three dependent variables.
4.7. Experimental procedures
Upon arrival, each subject was randomly directed to one of the
computers in the laboratory. These computers were arranged such
that subjects could not see other subjects’ computer screens, which
were showing other treatments. Next, an information sheet introducing the study and explaining the experimental procedure was
given to each subject, and adequate time was allotted for the subject
to understand the specified task. Before starting the task, subjects
were asked if they had any questions regarding the information
sheet, and following questions, were explicitly told that (1) there
were no right or wrong answers and that their judgments should
be based on their personal preferences; (2) they should not discuss
their thoughts and decisions with the people seated next to them;
(3) all their answers would be treated confidentially and only used
Author's personal copy
498
H.-C. Wang, H.-S. Doong / International Journal of Information Management 30 (2010) 493–501
Table 2
Scale items for the dependent variables.
Table 3
Exploratory factor analysis results.
Argument quality
The recommendation provided by the virtual
salesperson
AQ1 . . .was informative.
AQ2 . . .was helpful.
AQ3 . . .was valuable.
AQ4 . . .was persuasive.
Construct
Scale item
Factor 1
Factor 2
Factor 3
Argument quality
AQ1
AQ3
AQ4
AQ2
0.91
0.90
0.90
0.89
0.12
0.12
0.16
0.21
0.22
0.21
0.20
0.19
Source credibility
The virtual salesperson providing the book
recommendation
SC1 . . .was knowledgeable on this topic.
SC2 . . .was trustworthy.
SC3 . . .was credible.
SC4 . . .appeared to be an expert on this topic.
Purchase intention
PI3
PI1
PI2
PI4
0.14
0.18
0.15
0.12
0.94
0.90
0.89
0.84
0.20
0.22
0.20
0.16
Source credibility
SC4
SC2
SC1
SC3
0.19
0.22
0.19
0.21
0.13
0.23
0.26
0.18
0.86
0.85
0.85
0.81
6.21
51.78
4.34
1.40
2.24
18.65
4.46
1.22
1.66
13.81
5.06
0.84
Purchase intention
PI1: I am considering purchasing “The Art of Thinking:
Conceptual and Creative Thinking into a Practical Business
Plan” now at eBooks.com.
PI2: I would seriously contemplate buying “The Art of
Thinking: Conceptual and Creative Thinking into a Practical
Business Plan” at eBooks.com.
PI3: It is likely that I am going to buy “The Art of Thinking:
Conceptual and Creative Thinking into a Practical Business
Plan” at eBooks.com.
PI4: I am likely to make future purchases of “The Art of
Thinking: Conceptual and Creative Thinking into a Practical
Business Plan” at eBooks.com.
in this academic research; and (4) after completing the experiment,
a NT$70 (roughly equivalent to US$ 2.50) cash reward would be
provided in gratitude for their participation.
After the briefing, subjects then started browsing eBooks.com
and were requested to raise their hands when they had completed
the task. A questionnaire was subsequently passed to each subject. The returned questionnaire was double checked to make sure
every question had been answered. Upon confirmation of this, the
promised cash was given. Subjects took an average of 26 min to
complete the experimental procedures. All the experimental procedures were administered by the same experimenter following a
standard protocol.
4.8. Manipulation check
The manipulation check for argument forms and spokesperson
type was concluded in our research design. Each subject had to
answer the three questions listed below and indicate the extent
of his or her agreement with each question measured by the
7-point Likert scale. These questions were: (1) eBooks.com provided a book recommendation; (2) the recommendation offered at
eBooks.com has offered a recommendation foundation helping you
to understand why the specific book is being recommended; (3)
the recommendation offered at eBooks.com has offered additional
statistical evidence to support the argument. Analyzed results
showed that there was no significant difference between the CO
(mean = 6.78), CDW (mean = 6.77) and CDB (mean = 6.73) groups
for question 1. However, for question 2, while the means of CDW
(mean = 5.93) and CDB (mean = 6.13) groups did not differ from each
other statistically, both of them were significantly greater than that
of CO (mean = 2.88). Finally, for question 3, the mean for CDB (5.91)
was greater than that for CO (2.21) and CDW (2.41), although the
means of the latter two did not differ from each other statistically.
As regards manipulation of spokesperson type, each subject has
to choose one answer from the given choices to indicate which
spokesperson had recommended the book at eBooks.com. Findings indicated that each subject, no matter which treatment he
or she was in, was able to give the correct answer with regard to
the spokesperson. Consequently, the results of the manipulation
Eigenvalue
Percentage of variance
Mean
Standard deviation
Factor loadings
Note: BMO (Kaiser–Meyer–Olkin) = 0.87; Bartlett’s test was significant at p < 0.01.
checks indicated that experimental manipulations were successful
for all conditions in this study.
5. Results and analysis
The analysis was conducted using SPSS for Windows Version
15.0. All the constructs demonstrated adequate reliability and construct validity. The Cronbach alpha value was 0.95 for argument
quality, 0.91 for source credibility, and 0.94 for purchase intention. The scales were also tested for internal consistency and a
specified factor structure using factor analysis. Table 3 shows that
factor loadings on the intended constructs were all above 0.80, and
cross-loadings were no higher than 0.40, indicating that they were
unidimensional. Item scores for each variable were aggregated for
further analysis.
To verify the research hypotheses, multivariate analysis of
variance (MANOVA) was applied to test for the effects of argument form and spokesperson type on the argument quality and
source credibility. Results indicated that both the main effects of
argument form (Wilks’s = 0.42, F(4, 520) = 71.90, p < 0.001) and
spokesperson type (Wilks’s = 0.81, F(4, 520) = 14.66, p < 0.001)
were significant. However, the interaction effect between argument form and spokesperson type was not significant (Wilks’s
= .991, F(8, 520) = 0.29, p = 0.97). Thus, univariate analysis of variance (ANOVA) was performed to explore the main effects on each
dependent variable. When ANOVA yielded a significant result,
the Tukey HSD method was employed to investigate differences
between the effect of argument forms and spokesperson types on
each dependent variable.
As shown in Table 4, argument forms significantly affected subjects’ perceptions of argument quality (see Table 4 and Fig. 4). Thus,
post hoc tests were applied to determine which means differed.
Tukey HSD tests indicated that the differences between treatments
were all significant (see Table 5). That is to say, different argument forms led to different perceptions of argument quality among
subjects. Table 5 shows that subjects’ perceptions of the argument
quality of the RA recommendation were ranked as CDB > CDW > CO.
Therefore, Hypotheses 1 and 1a were supported.
Hypothesis 2 posited that different argument forms would lead
to different perceptions of source credibility. The analysis (see
Table 4 and Fig. 5) indicated that different recommendation argument forms resulted in differences in subjects’ perceptions of
source credibility. Tukey HSD tests (see Table 6) also showed that
subjects’ perceptions of the source credibility of the RA recom-
Author's personal copy
H.-C. Wang, H.-S. Doong / International Journal of Information Management 30 (2010) 493–501
Table 6
Effects of argument forms on subjects’ perception of source credibility.
Table 4
ANOVA results.
Argument quality
Argument form
Spokesperson type
Spokesperson × argument
Error
Source credibility
Argument form
Spokesperson type
Spokesperson × argument
Error
**
DF
MS
F
Significance
2
2
4
261
149.46
1.56
0.33
0.88
169.66
1.77
0.37
0.000**
0.17
0.83
39.32
30.89
0.17
**
2
2
4
261
18.36
14.42
0.08
0.47
0.000
0.000**
0.95
p < 0.01.
Fig. 4. Subjects’ perception of argument quality in each treatment.
Table 5
Effects of argument forms on argument quality.
Group I
Group J
Mean difference (I–J)
Significance
CO
CDW
CDB
−0.93
−2.55
0.00**
0.00**
CDW
CO
CDB
0.93
−1.61
0.00**
0.00**
CDB
CO
CDW
2.55
1.61
0.00**
0.00**
**
499
p < 0.01.
Group I
Group J
Mean difference (I–J)
Significance
CO
CDW
CDB
−0.28
−0.88
0.02*
0.00*
CDW
CO
CDB
0.28
−0.61
0.02*
0.00**
CDB
CO
CDW
0.88
0.61
0.00**
0.00**
*
**
p < 0.05.
p < 0.01.
(see Table 4 and Fig. 4) revealed that this assertion was not supported at the 0.05 significance level. Hence, Hypothesis 3 was not
supported.
Hypothesis 4 posited that using a spokesperson would lead to
perceptions of higher source credibility of the RA recommendation
than not using a spokesperson. Table 4 shows that ANOVA yielded a
significant result. Consequently, Tukey HSD tests were undertaken
and the differences between treatments were found to be significant (see Table 7 and Fig. 5). Subjects exhibited significantly higher
perceptions of the source credibility of the RA recommendation
when a spokesperson was used than when a spokesperson was not
used. Moreover, using an expert spokesperson had more significant
effects than using a consumer spokesperson. Therefore, Hypothesis
4 was supported.
Hypotheses 5 and 6 examined whether customers’ perceptions
of the argument quality and source credibility of the Webstore positively affected their purchase intentions at that Webstore. To test
H5 and H6 , purchase intention was regressed onto both argument
quality and source credibility. The analysis showed that the proposed model had a high and significant F ratio (F(2, 267) = 40.54,
p < 0.001), with an acceptable fit (adjusted R2 = 0.23). Both independent variables were statistically significant in explaining purchase
intention: the beta coefficient for argument quality was 0.19
(t = 3.12, p < 0.01), whilst that of source credibility was 0.37 (t = 6.06,
p < 0.001). Accordingly, Hypotheses 5 and 6 were both supported.
Thus, with data analysis completed, the research aims of this study
were achieved.
6. Conclusion and implications of the study findings
6.1. Discussion of findings
mendation differed across the three treatments, and was ranked
as CDB > CDW > CO. Thus, Hypotheses 2 and 2a were confirmed.
Hypothesis 3 asserted that using a spokesperson would lead to
perceptions of higher argument quality of the RA recommendation
than not using a spokesperson. However, the results of the analysis
Based on theories of spokesperson and Toulmin’s model of
argumentation, the current study has found empirical evidence
to support five out of six proposed hypotheses. First, customers’
perceptions of the argument quality and source credibility of RA
recommendations were found to effectively influence their purchase intentions at the Webstore. Second, customers’ perceptions
of argument quality and source credibility differed significantly as
a result of the varied argument forms (i.e. CO, CDW and CDB).
Third, although the varied spokesperson types (i.e. expert, conTable 7
Effect of spokesperson type on subjects’ perception of source credibility.
Fig. 5. Subjects’ perceptions of source credibility in each treatment.
Group I
Group J
Mean difference (I–J)
Significance
Webstore itself
Customer
Expert
−0.43
−0.80
0.00**
0.00**
Customer
Webstore itself
Expert
0.43
−0.37
0.00**
0.00**
Expert
Webstore itself
Customer
0.80
0.37
0.00**
0.00**
**
p < 0.01.
Author's personal copy
500
H.-C. Wang, H.-S. Doong / International Journal of Information Management 30 (2010) 493–501
sumer and Webstore itself) generated significantly different levels
of source credibility, using a spokesperson as opposed to not using a
spokesperson did not lead to customer perceptions of higher argument quality with regard to RA recommendations.
This may be due to the presentation of the spokesperson’s message in this study’s experiment. In the current study, the phrase
“spokesperson recommends this item” was used to simulate an
actual Webstore’s RA system (e.g. at Amazon.com, “we recommend
this item” was the phrase being used). Accordingly, while subjects may have perceived higher source credibility due to the effect
of a spokesperson, their perceived argument quality remained
unchanged, as there was no additional information in the RA recommendation revealing why the spokesperson wanted to recommend
this item. In other words, if the argument forms of CDW and CDB
had been used in the experiment, findings would probably have
been different, as this study has empirically disclosed that these
two forms will generate significantly higher argument quality than
the CO form.
6.2. Theoretical contributions
This study has added value to the existing RA literature and
its major theoretical contribution to research is twofold. First, this
study has successfully integrated the multi-discipline theories of
the spokesperson and the model of argumentation to investigate
to the process by which online customers assess the argument
quality and source credibility of RA recommendations and how
this process may influence their online purchase intentions. This
is important because past RA studies have mainly used customer
trust to explain customers’ online purchase intentions or RA acceptance. This study has broadened the knowledge of determinants of
customers’ RA acceptance and purchase intention from the existing
scope of trust assurance to encompass the new breadths of argument quality and source credibility. Second, this study has further
verified which marketing mix of argument forms and spokesperson type would best strengthen the online persuasion process. This
is crucial in the sense that argument quality and source credibility
are individuals’ underlying perceptions that cannot be controlled
by Webstore managers. Consequently, understanding the effects of
manipulating the argument forms and spokesperson types used in
RA recommendations has not only brought new insights, but also
indicated new directions for future research.
6.3. Practical implications
This study’s findings are of use for Webstore managers who are
seeking to build an effective strategy for RA recommendations over
their rival Webstores. By developing RA recommendations containing different argument forms and spokespersons, customers’ online
purchase intentions can be stimulated to different levels according
to their perceptions of argument quality and source credibility. As
only a handful of Webstores have initiated strategies to develop RA
recommendations comprising both argument form and spokesperson, this could become a useful sales tactic over rival Webstores that
only provide single messages in their RA recommendations.
The leading Webstore Amazon.com, for example, has utilized
a combined strategy. When recommending a book, Amazon.com
has introduced “average customer ratings”, which show how customers have rated that book by means of the number of stars
beneath it. This is similar to the strategy of customer endorsement
in recommending a book. Amazon.com also provides a warrant
in the RA recommendation (i.e. customers who bought this item
also bought . . .) to persuade readers who are browsing to purchase the recommended book. However, the argument element
of “backing” had not been used in developing an RA recommendation when this study was conducted. Nevertheless, Amazon has
initiated a similar strategy in providing “backing” information to
convince customers to make a purchase on the Web page for a
particular book: “What do customers ultimately buy after viewing this item? 65% buy the item featured on this page..”. Based on
this study’s findings, if Amazon.com could integrate the backing
element and the expert spokesperson strategy into a RA recommendation, customers’ purchase intentions might be effectively
stimulated, to Amazon’s undoubted advantage.
6.4. Limitations and future research areas
It is worth mentioning that this study’s results may have been
due to the task design and product type used in this study (i.e. “purchase a book as a gift for a friend”). To improve the generalization of
the findings, future studies might explore different product types
(e.g. luxury goods) and tasks (e.g. purchase for personal use) to
further verify the persuasive power of the RA systems provided at
Webstores. Further, a comparative study across different nations
regarding customers’ trust in RA recommendations could also be
informative and useful for both scholars and practitioners. Furthermore, due to the limited availability of subjects, this exploratory
study was designed to examine three argument forms, namely CO,
CDW and CDB. Future studies may follow Toulmin’s model and
examine all four treatments – (1) claim only, (2) claim plus data,
(3) claim plus data and warrant, and (4) claim plus data, warrant
and backing – to provide more interesting results.
6.5. Conclusion
For more than two decades, individuals’ adoption of information
technology has been a central and continuing issue in the discipline
of information systems (Kim, Chun, & Song, 2009). By incorporating Toulmin’s model and spokesperson strategy, this study has
extended this research theme in the context of RAs with new
theoretical and marketing insights. As RAs have become a prevailing strategy across different Webstores to assist customers’ online
purchase decisions, understanding the determinants of accepting
RA recommendations is thus central to scholars’ and practitioners’ interests. This is vital, as the expected benefits of developing
a RA at Webstore – to effectively boost sales, make customers
spend more money, and thereby ultimately increase the Webstore’s
profit – will be harmed if customers do not accept the RA in the
first place. By empirically demonstrating that: (1) customers perceived the highest argument quality with the CDB argument form,
followed sequentially by CDW and CO; (2) customers perceived
a higher source credibility with a spokesperson than without a
spokesperson and (3) both argument quality and source credibility can effectively predict online purchase intentions, this study’s
contributions to the field are addressed.
Acknowledgements
This research was supported by the National Science Council
in Taiwan under grant numbers NSC96-2416-H-194-005-MY2 and
NSC98-2410-H-415-014-MY3.
References
Aggarwal, P., & Vaidyanathan, R. (2003). The perceived effectiveness of virtual shopping agents for search vs. experience goods. Advances in Consumer Research,
30(1), 347–348.
Agrawal, J., & Kamakura, W. A. (1995). The economic worth of celebrity endorsers:
An event study analysis. Journal of Marketing, 59(3), 56–62.
Alexander, P. A., Fives, H., Buehl, M. M., & Mulhern, J. (2002). Teaching as persuasion.
Teaching and Teacher Education, 18(7), 795–813.
Atkin, C., & Block, M. (1983). Effectiveness of celebrity endorsers. Journal of Advertising Research, 23(1), 57–61.
Author's personal copy
H.-C. Wang, H.-S. Doong / International Journal of Information Management 30 (2010) 493–501
Bhattacherjee, A., & Sanford, C. (2006). Influence process for information technology acceptance: An elaboration like-hood model. MIS Quarterly, 30(40), 805–
825.
Dellarocas, C. (2003). The digitization of word-of-mouth: Promise and challenges of
online feedback mechanisms. Management Science, 49(10), 1407–1424.
Edmunds, A., & Morris, A. (2000). The problem of information overload in business organizations: A review of the literature. International Journal of Information
Management, 20(1), 17–28.
Fabrigar, L. R., Priester, J. R., Petty, R. E., & Wegener, D. T. (1998). The impact of attitude accessibility on elaboration of persuasive messages. Personality and Social
Psychology Bulletin, 24(4), 339–352.
Frias-Martinez, E., Chen, S. Y., & Liu, X. (2009). Evaluation of a personalized digital library based on cognitive styles: Adaptivity vs. Adaptability. International
Journal of Information Management, 29(1), 48–56.
Frias-Martinez, E., Magoulas, G., Chen, S. Y., & Macredie, R. (2006). Automated user
modeling for personalized digital libraries. International Journal of Information
Management, 26(3), 234–248.
Hansen, T., Jensen, J. M., & Solgaard, H. S. (2004). Predicting online grocery buying intention: A comparison of the theory of reasoned action and the theory
of planned behavior. International Journal of Information Management, 24(6),
359–550.
Heinze, N., & Hu, Q. (2006). The evolution of corporate web presence: A longitudinal study of large American companies. International Journal of Information
Management, 26(4), 313–325.
Holzwarth, M., Janiszewski, C., & Newmann, M. M. (2006). The influence of avatars
on online consumer shopping behaviour. Journal of Marketing, 70(4), 19–36.
Hovland, C. I., Janis, I. L., & Kelly, H. H. (1953). Communication and persuasion. New
Haven, CT: Yale University Press.
Kim, D., & Benbasat, I. (2006). The effect of trust assuring arguments on customer
trust in Internet stores. Information System Research, 17(3), 286–300.
Kim, J. K., Kim, H. K., Oh, H. Y., & Ryu, Y. U. (2009). A group recommendation system for online communities. International Journal of Information Management,
doi:10.1016/j.ijinfomgt.2009.09.006
Kim, Y. J., Chun, J. U., & Song, J. (2009). Investigating the role of attitude in technology acceptance from an attitude strength perspective. International Journal of
Information Management, 29(1), 67–77.
Komiak, S. Y. X., & Benbasat, I. (2006). The effects of personalization and familiarityon trust and adoption of recommendation agents. MIS Quarterly, 30(4),
941–960.
Koufaris, M. (2002). Applying the technology acceptance model and flow theory to
online consumer behaviour. Information Systems Research, 13(2), 205–223.
Krishnamurthy, P., & Sivaraman, A. (2002). Counterfactual thinking and advertising
responses. Journal of Consumer Research, 28(4), 650–659.
Lai, H., & Yang, T. C. (2000). A system architecture for intelligent browsing on the
web. Decision Support Systems, 28(3), 219–239.
Lim, K. H., Sia, C. L., Lee, M. K. O., & Benbasat, I. (2006). Do I trust you online, and
if so, will I buy? An empirical study of two trust-building strategies. Journal of
Management Information Systems, 23(2), 233–266.
Mao, J. I, & Benbasat, I. (2000). The use of explanations in knowledge-based systems:
Cognitive perspective and a process-tracing analysis. Journal of Management
Information Systems, 17(2), 153–179.
Mason, R. O., & Mitroff, I. I. (1981). Challenging strategic planning assumptions: Theory,
cases, and techniques. New York: Wiley-Interscience.
McKnight, H., & Kacmar, C. (2006). Factors of information credibility for an Internet
advice site. In Proceedings of the 39th Hawaii international conference on system
sciences.
Netcraft. (2009). Available http://news.netcraft.com/archives/2009/10/17/october
2009 web server survey.html (Access 2009/11/04).
501
Ohanian, R. (1990). Construction and validation of a scale to measure celebrity
endorsers’ perceived expertise, trustworthiness, and attractiveness. Journal of
Advertising, 19(3), 39–52.
Ohanian, R. (1991). The impact of celebrity spokespersons’ perceived image on consumers’ intention to purchase. Journal of Advertising Research, 31(1), 46–54.
Petty, R. A., & Cacioppo, J. T. (1986). Communication and persuasion: Central and
peripheral routes to persuasion. New York: Springer-Verlag.
Petty, R. E., & Wegener, D. T. (1999). The elaboration likelihood model: Current status
and controversies in dual-process. In S. Chaiken, & Y. Trope (Eds.), Theories in
social psychology (pp. 37–72). New York: Guilford Press.
Petty, R. E., Cacioppo, J. T., & Schumann, D. (1983). Central and peripheral routes
to advertising effectiveness: The moderating role of involvement. Journal of
Consumer Research, 10(2), 135–147.
Senecal, S., & Nantel, J. (2004). The influence of online product recommendations on
consumers’ online choices. Journal of Retailing, 80(2), 159–169.
Simintiras, A. C., & Cadogan, J. W. (1996). Behaviourism in the study of
salesperson–customer interactions. Management Decision, 34(6), 57–64.
Stafford, M. R., Stafford, T. F., & Day, E. (2002). A contingency approach: The effects of
spokesperson type and service type on service advertising perceptions. Journal
of Advertising, 31(2), 17–35.
Stein, B. S., Littlefield, J., Bransford, J. D., & Persampieri, M. (1984). Elaboration and
knowledge acquisition. Memory & Cognition, 12(55), 522–529.
Sternthal, B., Phillips, L., & Dholakia, R. R. (1978). The persuasive effect of source credibility: Tests of cognitive response. Journal of Consumer Research, 4(4), 252–260.
Sussman, S. W., & Siegal, W. S. (2003). Informational influence in organizations:
An integrated approach to knowledge adoption. Information Systems Research,
14(1), 47–65.
Tam, K. Y., & Ho, S. Y. (2005). Web Personalization as a persuasion strategy: An
elaboration likelihood model perspective. Information Systems Research, 16(3),
271–291.
Toulmin, S. E. (1958). The use of argument. UK: Cambridge University Press.
Tseng, S., & Fogg, B. J. (1999). Credibility and computing technology. Communications
of the ACM, 42(5), 39–44.
Wang, W., & Benbasat, I. (2005). Trust in and adoption of online recommendation
agent. Journal of the Association for Information Systems, 6(3), 72–101.
Wang, W., & Benbasat, I. (2007). Recommendation agents for electronic commerce:
Effects of explanation facilities on trusting beliefs. Journal of Management Information System, 23(4), 217–246.
Wathen, C. N., & Burkell, J. (2002). Believe it or not: Factors influencing credibility on
the web. Journal of the American Society for Information Science and Technology,
53(2), 134–144.
Wheeler, S. C., Petty, R. E., & Bizer, G. Y. (2005). Self-schema matching and attitude change: Situational and dispositional determinants of message elaboration.
Journal of Consumer Research, 31(4), 787–797.
Wood, W., & Kallgren, C. A. (1988). Communicator attributes and persuasion: Recipients’ access to attitude-relevant information in memory. Personality and Social
Psychology Bulletin, 14(1), 172–182.
Ye, L. R., & Johnson, P. E. (1995). The impact of explanation facilities on user acceptance of expert systems advice. MIS Quarterly, 19(2), 157–172.
Hui-Chih Wang is an assistant professor at Department of Information Management, National Chung Cheng University, Taiwan. Her research interests are in
Consumer Behavior, e-commerce and Internet Marketing Strategies.
Her-Sen Doong is an associate professor at Department of Management Information
Systems, National Chiayi University, Taiwan. His current research interests are in the
areas of negotiation support systems, human–computer interaction, and electronic
commerce.