Let's Shop Online Together: An Empirical Investigation of Collaborative Online Shopping Support
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
Information Systems Research informs ®
Vol. 21, No. 4, December 2010, pp. 872–891
doi 10.1287/isre.1080.0218
issn 1047-7047 eissn 1526-5536 10 2104 0872
© 2010 INFORMS
Let’s Shop Online Together: An Empirical
Investigation of Collaborative Online
Shopping Support
Lei Zhu, Izak Benbasat
Sauder School of Business, University of British Columbia, Vancouver, British Columbia V6T 1Z2, Canada
{leizhumis@gmail.com, izak.benbasat@sauder.ubc.ca}
Zhenhui Jiang
School of Computing, National University of Singapore, Singapore 117417, Republic of Singapore,
jiang@comp.nus.edu.sg
P rior studies investigating business-to-consumer e-commerce have focused predominantly on online shop-
ping by individuals on their own, although consumers often desire to conduct their shopping activities
with others. This study explores the important, but seldom studied, topic of collaborative online shopping.
It investigates two design components that are pertinent to collaborative online shopping support tools, namely,
navigation support and communication support. Results from a laboratory experiment indicate that compared
to separate navigation, shared navigation effectively reduces uncoupling (i.e., the loss of coordination with one’s
shopping partner) incidents per product discussed and leads to fewer communication exchanges dedicated to
resolving each uncoupling incident, thereby enhancing coordination performance. Compared to text chat, voice
chat does not help reduce the occurrence of uncoupling, but likely increases the efficiency in resolving uncou-
pling. The results further show that shared navigation and voice chat can significantly enhance the collaborative
shoppers’ perceptions of social presence derived from their online shopping experiences. The interaction effect
on social presence implies that the benefit of shared navigation is higher in the presence of text chat than in the
presence of voice chat.
Key words: collaborative online shopping; shared navigation; common ground; media richness; uncoupling;
social presence; electronic commerce
History: Laurie Kirsch, Senior Editor; Dennis Galetta, Associate Editor. This paper was received on
May 28, 2006, and was with the authors 17 months for 3 revisions. Published online in Articles in Advance
May 12, 2009.
1. Introduction The need for social online shopping support is
Shopping is often a social process in which a shop- evident from prior studies (e.g., Tractinsky and
per is accompanied by friends or family members Rao 2001). A survey by Jupiter Communications has
(Evans et al. 1996). Tauber (1972) has argued that one found that 90% of online customers prefer some sort of
of the prime motives for shopping is the desire to human contact when they are conducting online trans-
communicate with others who have similar interests, actions (Gutzman 2000). Correspondingly, Rayport
to share ideas about particular products with shop- and Jaworski (2001) have suggested that the capac-
ping companions, to seek their feedback, and to enjoy ity for online consumers to communicate with one
leisure time with friends and family. Nevertheless, it another is critical to the success of Web stores.
is sometimes difficult to shop together simply because In this study, we use the term collaborative on-
of physical separation, e.g., two friends may reside line shopping to describe the activity in which a
in different cities. Fortunately, this constraint may be consumer shops at an online store concurrently
alleviated by online shopping, because in a virtual with one or more remotely located shopping part-
shopping mall friends need not be collocated. ners. Multiple techniques can be integrated to create
872Zhu et al.: Let’s Shop Online Together: An Empirical Investigation of Collaborative Online Shopping Support
Information Systems Research 21(4), pp. 872–891, © 2010 INFORMS 873
a collaborative online shopping experience. For for collaborative online shopping tools on shoppers’
example, Landsend.com has deployed a feature called coordination performance and their perceptions of
“shop with a friend™ ” that provides support for con- social presence. Coordination performance reflects the
sumers at different locations to synchronize their Web utilitarian perspective when shoppers coordinate their
navigation and to communicate online using text. product search and evaluation processes, whereas
These instant interaction functionalities have been social presence represents the social perspective, i.e.,
acknowledged by practitioners to bolster a company’s the relational nature of collaboration that is exempli-
Internet sales (Dukcevich 2002). fied by the feeling of intimacy and warmth (Kumar
Indeed, technology-mediated person-to-person and Benbasat 2006). Both perspectives are relevant
communication in organizational environments has and complement each other, because the goal of col-
been a subject of academic research for several laborative online shopping is not only to assist a
decades (Short et al. 1976). For example, studies shopper in navigating to the right product to seek
have investigated how people work collaboratively his companions’ opinions and suggestions, but also to
with the support of groupware technologies, such fulfill his desire to interact with others and socialize
as e-mail, bulletin boards, group schedules, group with them (Tauber 1972). To the best of our knowl-
support systems, workflow systems, and collabo- edge, this study is among the first in the IS literature
rative authoring tools (Ishii et al. 1994, Kayworth to evaluate the effectiveness of different techniques
and Leidner 2002, Limayem and DeSanctis 2000). for collaborative online shopping.
Additionally, a large number of empirical studies This paper is organized as follows. The next sec-
have compared computer-mediated communications tion reviews previous literature and discusses the the-
to face-to-face interactions (Bordia 1997, Hoffman oretical foundations. Section 3 identifies the two key
and Novak 1996). However, to date there has been technological components for designing collaborative
little research attention paid to the phenomenon of online shopping. A research model is then developed
collaboration in online shopping with new IT-enabled in §4. The experimental research method used in the
features, such as synchronized navigation and instant present study is described in §5. Section 6 discusses
communication. Because of the lack of knowledge data analysis procedure and corresponding results.
of these emerging collaborative technologies, as well The final section concludes with the findings, contri-
as the social nature of online shopping, it may be butions, and limitations of the study.
presumptuous to apply the previous findings on
the use and impact of collaborative technologies in
2. Theoretical Foundations
working environments to an online shopping context.
Therefore, additional research effort is needed to 2.1. Collaborative Work
analyze and evaluate collaborative online shopping Collaborative technologies, such as e-mail, group sup-
technologies theoretically and empirically to advance port systems, and video conferencing, are used by
the IS knowledge concerning this important and members of groups or organizations to communicate
expanding buying channel. with one another and coordinate their activities to
To address this deficiency, the present study inves- execute tasks (Carte and Chidambaram 2004). In gen-
tigates the design of a collaborative online shopping eral, collaborative technologies have been found to be
support tool by identifying its two primary fea- useful in enhancing the effectiveness of team collabo-
tures, namely, navigation support and communica- ration (Goodman and Darr 1998) in various contexts,
tion support. These two features are related to the such as distributed learning (Alavi et al. 2002), virtual
two fundamental processes of collaborative online communities (Bieber et al. 2002), and system/product
shopping, i.e., to help shopping companions nav- development (Scott 2000). For example, Easley et al.
igate to a particular product of potential interest (2003) found that the use of collaborative systems
and to allow for the exchange of ideas or opin- could significantly increase creative performance for
ions about that product. More specifically, this study team-based work. Banker et al. (2006) found that
evaluates the influence of different design choices the implementation of collaborative product designZhu et al.: Let’s Shop Online Together: An Empirical Investigation of Collaborative Online Shopping Support
874 Information Systems Research 21(4), pp. 872–891, © 2010 INFORMS
could improve product quality, reduce design cycle and interpretations of the situation, at times assum-
time, and lower product development costs. Similarly, ing incorrectly that the other speaks and under-
Gallupe et al. (1992) compared electronic brainstorm- stands on the basis of the same information and
ing with traditional verbal brainstorming and found interpretations. In an ethnographic study, for exam-
group members to be more satisfied with the former. ple, Bechky (2003) observed how engineers, techni-
Prior studies have also revealed that the processes cians, and assemblers on a product floor resolved
of collaboration encompass both the detection and misunderstandings among one another. He found
resolution of conflicts arising from collaboration that members of these communities overcame misun-
(Chu-Carroll and Carberry 2000) as well as the facil- derstandings by cocreating common ground, which
itation of social awareness among team members transformed their understanding of products and
(Burke 2001, Carroll et al. 2003). Therefore, two rele- production processes. He also observed that verbal
vant theories on common ground and media richness explanations alone did not suffice to create common
are discussed below to provide the theoretical founda- ground. Instead, members used demonstrations with
tions for the design of collaborative online shopping tangible and visible representations to establish com-
tools. mon ground.
Based on these findings, it can be inferred that
2.2. Common Ground Theory common ground could be useful in helping collab-
Research on situated cognition theorizes that peo- orative shoppers to coordinate their behavior; and
ple’s learning and cognition are highly dependent on that common ground could be established by show-
the contexts in which learning and cognition take ing the same Web contents to both participants
place (Lave 1988, Lave and Wenger 1990). For collo- simultaneously.
cated collaborative work, collaborators share the same
2.3. Media Richness Theory
working environment and are exposed to the same
Media richness theory is used to characterize a me-
contextual cues; hence, they are likely to be aware of
dium’s ability to change understanding within a spe-
one another’s concerns, opinions, and comments and cific time interval (Daft and Lengel 1986, Daft et al.
to reach unanimity through this mutual awareness, 1987). According to the theory, the richness of media
thereby improving productivity (Olson and Olson can be evaluated based on four criteria, namely, the
2000). In contrast, in distant collaboration, one person ability of a medium to transmit multiple cues, allow
often fails to anticipate which features of his local con- for immediacy of feedback, support language variety,
text differ significantly from those of his remote part- and provide personal focus. Based on these criteria,
ner, thereby leading to misunderstanding between the Daft and his colleagues propose that media can be
two (Cramton 2002). In both cases, the key to success- ranked along a “media richness continuum” ranging
ful collaboration is whether collaborators can estab- from very rich to very lean. Face-to-face communi-
lish common ground, defined as the knowledge held cation is considered to be the richest communication
in common by the collaborators, combined with their medium, followed by telephone, handwritten notes,
awareness that they have the knowledge in common addressed documents, and unaddressed documents.
(Clark and Brennan 1991, Olson and Olson 2000). Media richness theory divides information pro-
Common ground is considered to be vital for effec- cesses into two categories: reducing uncertainty
tive communication among collaborators, because it (i.e., overcoming the absence of information) and low-
provides them with a shared referential base for ering equivocality (i.e., removing ambiguity). Uncer-
discussion and ensures that the knowledge trans- tainty can be reduced by supplying more relevant
ferred connotes the same meaning for both the sender information, whereas equivocality can be lowered by
and the receiver (Clark 1996, Cramton 2002, Hanna using richer media. For example, in the context of
et al. 2003). In contrast, without common ground, interpersonal collaboration to interpret and resolve
people speak and understand things that are com- cognitively conflicting situations, richer media are
municated on the basis of their own information often preferred and used by managers, as comparedZhu et al.: Let’s Shop Online Together: An Empirical Investigation of Collaborative Online Shopping Support
Information Systems Research 21(4), pp. 872–891, © 2010 INFORMS 875
to lean media (Daft and Lengel 1986). Carlson and discussion refers to the same products or topics. Sec-
Davis (1998) have thus suggested that media richness ond, such a system must allow remote shoppers to
is closely tied to people’s social communication, inter- engage in synchronous conversations, so that they
pretation, and gain of consensus. Canessa and Riolo can discuss products and services with each other, to
(2003) further noted that “if the intrinsic communi- share and exchange opinions.
cation richness of the medium that members use is Corresponding to these two mechanisms, a collabo-
high, then the medium will effectively contribute to rative online shopping system can be designed using
creating the overall shared meaning.” two types of technological support: navigation sup-
The effects of richer media have been investigated port and communication support.
in numerous studies. Kahai and Cooper (2003), for
example, examined the effects of media richness on 3.1. Navigation Support
decision quality through three mediating constructs: Navigation support determines how collaborative
social perception, message clarity, and the ability to shopping companions navigate to the products of
evaluate others. In their study, subjects were asked their interests. For example, if two people who are
to perform two tasks under conditions having differ- physically separated would like to shop for an item
ent levels of media richness: face-to-face, electronic together on a website, they may first inform each
meetings, and electronic mail. Kahai and Cooper other what website they will be visiting and what
found that rich media enhanced social perception products they will be looking at. Next, the two shop-
and increased individuals’ perceived ability to eval- pers need to navigate to the specific website and look
uate others. Complementing these findings, Kraut for the products that they have agreed to explore.
et al. (1992) investigated media choice in collabora- Here, the common website and products displayed
tive writing. They found that richer media (e.g., face- that are visible to both parties serve as a referential
to-face interaction), as compared to leaner media context.
(e.g., computer/phone and computer only), signifi- The two companions could conduct separate naviga-
cantly alleviated coordination problems in collabora- tion, i.e., the paces of their navigation are independent
tive writing, e.g., when people performed equivocal and controlled by each individual. Alternatively, IT
tasks such as planning and constructing a long doc- support, such as the shared navigation technique,
ument. The results also revealed that spoken annota- enables two or more people to synchronously view
tions (i.e., voice) were preferred to, and were easier the same Web pages through their individual Web
to use than, written annotations (i.e., text) when com- browsers (Twidale 1995). Either one of the two shop-
municating complex and equivocal topics. Thus, their pers, but only one at a time, can control what
findings clearly support the media-task fit tenet pro- appears in both of their browsers, including the Web
posed by media richness theory. page content, navigation, and even mouse movement.
In other words, shared navigation enforces synchro-
nized browsing behavior. Similar applications can be
3. Support Technologies for found in work-related contexts, e.g., library represen-
Collaborative Online Shopping tatives assist customers in finding the resources that
Two facilitating mechanisms are important to design- they are looking for (Zou 2006); lecturers control the
ing an effective collaborative online shopping system. Web pages displayed on audiences’ monitors (Marais
First, a well-designed collaborative online shopping and Bharat 1997, Puglia et al. 2000).
interface should provide shoppers with a common
context for product selection. More specifically, it 3.2. Communication Support
should create a referential context that both shop- Communication support ensures that shopping part-
pers can access and comment on, such as web pages ners can communicate to share their interests, obser-
that display products (Kraut et al. 2003). Without a vations, and suggestions instantly. Two types of
common referential context, collaborative shopping is Web-based instant communication support, i.e., text
difficult because shoppers cannot ensure that their chat and voice chat, are investigated in the presentZhu et al.: Let’s Shop Online Together: An Empirical Investigation of Collaborative Online Shopping Support
876 Information Systems Research 21(4), pp. 872–891, © 2010 INFORMS
Figure 1 Research Model (Shen et al. 2002). Therefore, the key to successful col-
laborative information search is to coordinate shop-
Coordination performance
H1A ping companions’ browsing actions so as to accurately
Navigation
H1B
The number of uncoupling and efficiently locate product information of common
support incidents per product discussed
H2 interest (Diamadis and Polyzos 2004). If there is a lack
H4
H3 The number of communication of smooth coordination, one cannot easily locate and
exchanges used to resolve
each uncoupling incident examine the product that his companion is comment-
ing on; consequently, the primary purpose of collabo-
Communication
H5 Social presence rative online shopping cannot be achieved.
support H6
In this paper, we use the term uncoupling to describe
the state in which collaborative shoppers lose coor-
study. Both instant text chat and voice chat facilitate dination with their shopping companions. As such,
real-time communication between two users via the to improve collaborative online shopping, shoppers
Internet. In text mode, text submitted via a chat win- require a collaborative technology that helps them (1)
dow by one user appears instantly on another user’s reduce the occurrence of uncoupling; and (2) facilitate
computer screen. Voice chat uses Voice over Internet the resolution of uncoupling.
Protocol (VoIP) technologies to facilitate voice calls One factor relevant to the extent of uncoupling is
over the Internet instead of the traditional telephone the number of uncoupling incidents that occur in a
landline system. shopping task. Furthermore, in view of the previous
findings that it is easier and faster to speak than to
type (Kinney and Watson 1992, Walther 1992, Williams
4. Research Model and Hypothesis 1977), it is likely that collaborative shoppers discuss
Development more products using voice than using text.1 Discussing
and exchanging opinions on more products implies
4.1. Overview
that shoppers can perform a more thorough examina-
Prior research has suggested that collaboration in-
tion of displayed product alternatives, thereby poten-
volves action awareness and social awareness between
tially leading to a more-informed product decision
collaborators (Carroll et al. 2003). Correspondingly, the
(O’Keefe and McEachern 1998). On the other hand,
present study investigates the impact of navigation
the fact that more products are being discussed may
support and communication support on the coordi-
increase the number of uncoupling incidents in collab-
nation performance of online shopping companions as
orative shoppers’ communication. Therefore, to allevi-
a group and their perceptions of social presence (Fig-
ate this confounding effect, it was decided to calculate
ure 1). Two types of navigation support are stud-
the occurrence of uncoupling by dividing the number
ied, i.e., separate navigation versus shared navigation, of uncoupling incidents by the number of products
together with two types of communication support, that were discussed in a shopping task, thus represent-
i.e., text chat versus voice chat. ing the average number of uncoupling incidents per
product discussed.
4.2. Hypothesis Development
On the other hand, when uncoupling occurs, col-
4.2.1. Dependent Variable: Coordination Perfor- laborative shoppers usually resolve uncoupling by
mance. O’Keefe and McEachern (1998) have noted informing their partners of the product or the web
that an important stage for Web-based customer deci- page that they are looking at as well as their
sion making is information search. For collaborative navigation intentions to coordinate their collaborative
online shopping, because information search is a task behavior. Hence, the extent to which a collaborative
performed jointly by both parties, it is not uncom- technology facilitates the resolution of uncoupling is
mon that conflicts may occur when the two shoppers calculated by dividing the number of communication
follow divergent product search paths at times, thus
leading to their actions interfering with each other 1
Our experimental data confirms this conjecture (see §6.3).Zhu et al.: Let’s Shop Online Together: An Empirical Investigation of Collaborative Online Shopping Support
Information Systems Research 21(4), pp. 872–891, © 2010 INFORMS 877
exchanges dedicated to resolving uncoupling by the product examination and increases the chance of loss
total number of uncoupling incidents that occur in of coordination with his companion.3
a particular shopping task. In short, it refers to the In terms of overall coordination performance, the
average number of communication exchanges used to use of shared navigation is likely to alleviate the occur-
resolve each uncoupling incident. rence of uncoupling as compared to the use of separate
navigation. Shared navigation allows people to view
4.2.2. Effects of Navigation and Communication
the same Web pages synchronously and to share their
Support on Coordination Performance. Uncoupling
navigation. These shared visual and behavioral cues
can occur in both separate and shared navigation
enhance both shoppers’ awareness of each other’s sit-
conditions. With separate navigation, shopping part-
uations and their common ground (Kraut et al. 2003).
ners2 work with their own individual displays of the
Specifically, shared navigation enforces a temporal
Web pages (a privileged ground situation, see Hanna
and spatial match between the information accessed
et al. 2003). In such circumstances, it is likely that
by both shoppers, which enables them to under-
one shopper might assume incorrectly that the other
stand each other’s contextual cues concurrently and
is speaking about and understanding the situation on
is thus likely to reduce the occurrence of uncoupling.
the basis of the first shopper’s privileged ground. Con-
In addition, once uncoupling occurs, shared naviga-
sequently, uncoupling incidents may occur because
tion facilitates the resolution of uncoupling by allow-
they cannot easily locate the same Web page or
ing shoppers to consciously rely on synchronized page
because they do not refer to the same product on a par-
navigation and to use pointing devices to show others
ticular Web page. Furthermore, because of the lack of
the item one is looking at.
visible common ground, shoppers with separate nav-
Because shared navigation helps establish better
igation cannot resolve an uncoupling incident easily,
common ground between the two shoppers than sep-
but have to inform each other of their current location arate navigation, we posit:
and the product that they are looking at, and, based
on that, align their navigation with each other. Hypothesis 1A (H1A). Compared to separate naviga-
Uncoupling may occur in the shared navigation con- tion, shared navigation reduces the number of uncoupling
dition when both parties do not refer to the same prod- incidents per product discussed.
uct despite being on the same Web page. In addition,
Hypothesis 1B (H1B). Compared to separate naviga-
uncoupling can also be caused by poor coordination.
tion, shared navigation leads to fewer communication
For example, because two browsers are strictly syn-
exchanges used to resolve each uncoupling incident.
chronized, one’s full control over his own preferred
way of navigation may be interfered with or infringed Media richness theory suggests that voice-based
on by his companion’s unannounced act of moving to communication is ranked higher than text-based com-
a different page. For example, assume that both shop- munication along the media richness continuum (Daft
pers are looking at the same screen. Whereas shop- and Lengel 1986). This is because voice can deliver
per A is focusing on examining product X, shopper B multiple cues beyond text. People can use their voices
decides to navigate to a different Web page. Therefore, to emphasize important points, to reveal doubt or
shopper A may get confused by shopper B’s unan- uncertainty, to display acceptance, to invoke dom-
nounced act and thus suffer from the loss of coor- inance, and for other purposes, through nonver-
dination. Hence, if appropriate coordination is not bal cues such as inflection, pitch, tone, and pauses
developed, such interference leads to the unwanted (Williams and Cothrel 2000). Specifically, media rich-
outcome of disrupting one’s natural cognitive flow in ness theory also suggests media-task fit (McGrath
and Hollingshead 1993), i.e., a task is most effectively
2
Here we assume that two shoppers are physically separated but
3
perform collaborative online shopping at the same website. The We will elaborate this phenomenon in the discussion section and
logic is the same if more than two collaborative shoppers are show that shared navigation actually leads to more intrascreen nav-
involved. igational uncoupling.Zhu et al.: Let’s Shop Online Together: An Empirical Investigation of Collaborative Online Shopping Support
878 Information Systems Research 21(4), pp. 872–891, © 2010 INFORMS
performed in the “best-fitting” communication envi- other users (Short et al. 1976). It represents the capa-
ronment. In particular, equivocal messages are better bility of a medium to allow a user to experience oth-
communicated using rich media than lean media. This ers as being psychologically present (Fulk et al. 1987).
tenet was tested by Kraut et al. (1992), who found that In general, social presence is found to be important
in highly equivocal tasks such as collaborative writ- in the context of task collaboration. Burke (2001), for
ing, rich media led to better performance and fewer example, argues that social presence is an important
coordination problems than lean media. aspect of distant collaboration and that it is positively
With separate navigation, a coordination task re- related to users’ participation in a learning environ-
quires collaborative shoppers to explicitly inform each ment because the lack of social cues may lead to
other of the specific products that they are comment- feelings that the environment is cold and unfriendly.
ing on. Hence, it is necessary to convey information Other studies have also identified the important role
about product location, which often involves contex- of social presence in the context of Internet shopping.
tual information such as screen displays, landmarks, For example, Kumar and Benbasat (2006) indicate that
layout, and even salient product characteristics. Such social presence characterizes the relational nature of a
information is usually difficult to describe clearly, shopping experience, thus complementing the utilitar-
leading to ambiguity and conflict in coordination ian perspective. Gefen and Straub (2003) have found
(McGrath and Hollingshead 1993). Therefore, with that social presence affects consumers’ trust, which
separate navigation, the greater communicative needs in turn influences their purchase intentions. Because
and coordination difficulties make coordination tasks one of the main objectives of collaborative online
highly equivocal; consequently, voice is better than shopping is to fulfill people’s desire for social inter-
text in improving collaborative shoppers’ coordina- action (Schubert 2000, Tauber 1972), social presence is
tion performance. In contrast, with shared navigation, particularly important in the present context.
coordination tasks are minimally equivocal as collab-
orative shoppers are physically bound, i.e., are look- 4.2.4. Effects of Navigation and Communication
ing at the same screen, and can show each other a Support on Social Presence. Compared to separate
particular product by pointing their mice at the prod- navigation, shared navigation enables both shoppers
uct. Therefore, the use of voice versus text is unlikely to view the same screen contents synchronously, thus
to cause significant differences in coordination per- generating a visible common ground. This experience
formance. Thus, we predict the following interaction where one can see his companion’s mouse move-
effects: ment and navigation process as well as examine the
product or the Web page that his companion shows
Hypothesis 2 (H2). There is an interaction effect bet-
him, provokes their awareness of the common situa-
ween navigation support and communication support on
tion (Kraut et al. 2003) and is comparable to an in-
the number of uncoupling incidents per product discussed,
store social shopping experience where two shoppers
i.e., voice chat leads to fewer uncoupling incidents per prod-
jointly examine the same product (Jarvenpaa and Todd
uct discussed than text chat in the separate navigation con-
1996–1997), thereby leading both shoppers to feel that
dition, but not in the shared navigation condition.
they are together.
Hypothesis 3 (H3). There is an interaction effect bet-
ween navigation support and communication support on Hypothesis 4 (H4). Shared navigation generates high-
the number of communication exchanges used to resolve er social presence than separate navigation in collaborative
each uncoupling incident, i.e., voice chat leads to fewer com- online shopping.
munication exchanges to resolve each uncoupling incident Prior studies have found that media differing in
than text chat in the separate navigation condition, but not richness affect the amount of social presence that com-
in the shared navigation condition. municators perceive (Burke and Chidambaram 1999,
4.2.3. Dependent Variable: Social Presence. So- Chidambaram and Jones 1993, Yoo and Alavi 2001). In
cial presence refers to the degree to which a medium general, it is suggested that face-to-face interaction is
allows a user to establish personal connection with ideal because it conveys not only verbal information,Zhu et al.: Let’s Shop Online Together: An Empirical Investigation of Collaborative Online Shopping Support
Information Systems Research 21(4), pp. 872–891, © 2010 INFORMS 879
but also nonverbal information such as facial expres- 5. Research Method
sion, tone, and gesture, which are, at times, important
5.1. Experimental Design
and even indispensable to revealing a communica-
A laboratory experiment with a mixed 2 × 2 design
tion stance (Chidambaram and Jones 1993). Similarly,
(Gravetter and Wallnau 2000, Sternthal and Craig
because voice can deliver many nonverbal cues that
1982) was used to test the proposed hypotheses.4 Nav-
cannot be communicated via text (Short et al. 1976),
igation support was chosen as the between-subject
such as inflection, pitch, tone, and pauses, voice chat
factor (separate navigation versus shared navigation),
helps shoppers retain their habitual linguistic style
and communication support as a within-group factor
and behavior, and hence is more natural and makes
(text chat versus voice chat).
shopping companions feel socially closer to each other
Two types of products—school bags and watches—
than text chat.
were used to increase the generalizability and applica-
Hypothesis 5 (H5). Voice chat generates higher social bility of the potential findings. The two products were
presence than text chat in collaborative online shopping. selected for this study for several reasons: (1) both
We also predict that the effect of navigation sup- products are social products, inasmuch as they are
port on social presence may depend on the particular used in public settings and therefore serve to exhibit
communication support technique used. In general, their owners’ tastes and values; (2) both contain a vari-
as discussed earlier, shared navigation is expected to ety of attributes (e.g., functionality, look, and size) that
lead to higher social presence than separate navigation can provoke discussion between two shopping part-
because shoppers under shared navigation may sense ners; and (3) both products are gender-neutral prod-
each other’s mouse movement and navigation inten- ucts. Amazon.com was chosen as the experimental
tion. However, these perceptions are relatively indirect website because it provides a rich collection of school
because shoppers do not build substantial and direct bags and watches (over 1,000 types of each product).
interaction with each other; instead, they interact Four types of collaborative online shopping sup-
through manipulating the Web interface. On the other port were implemented using a Web collaboration
hand, voice chat, as compared to text chat, can signif- tool, Microsoft MSN 8, which provides instant
icantly boost the feelings of social presence, because it text/voice chat support, and shared/separate naviga-
provides a direct and substantial interaction between tion support.
the two shoppers. Overall, the effect of communica-
5.2. Experimental Procedures
tion support (voice versus text) is stronger, i.e., more
Participants in this experiment were students from a
salient, than the effect of navigation support (shared
public university. To ensure sufficient power of 0.8
navigation versus separate navigation). Prior research
with a medium effect size for a two-by-two mixed
has suggested that when people are presented with
design, 128 participants (64 pairs) were recruited to
multiple stimulation cues, more-salient information
participate in the final experiment.
cues play a disproportionately more important role
Each person who volunteered was asked to invite
than less-salient cues (Hutchinson and Alba 1991,
a friend to participate, to emulate a real shopping sit-
McGill and Anand 1989). Therefore, although when
uation. The pair was then randomly assigned to one
text chat is used, shared navigation can lead to higher
of two experimental groups (separate versus shared
social presence than separate navigation, this increase
navigation). Each participant was paid $15 for par-
may be less prominent when voice chat is used
because the relatively less direct and less influential ticipation. In addition, participants were told that
effect of shared navigation on social presence is over- they would have a one-in-four chance of receiving a
shadowed by the much stronger effect of voice. $60 bonus toward the purchase of the products they
chose in the study.
Hypothesis 6 (H6). There is an interaction effect bet-
ween navigation support and communication support on 4
To economize on the number of participants, a mixed two-by-two
social presence, i.e., navigation support has a stronger effect factorial design instead of a two-by-two between-factorial design
in the text chat condition than in the voice chat condition. was chosen (Gravetter and Wallnau 2000).Zhu et al.: Let’s Shop Online Together: An Empirical Investigation of Collaborative Online Shopping Support
880 Information Systems Research 21(4), pp. 872–891, © 2010 INFORMS
Participants were randomly assigned the role of collaborative online shopping, it was decided to make
main buyer or opinion giver. They were placed in two the contextual cues more salient in the experimental
separately located rooms equipped with computers and questionnaire design.
and monitors of the same type. Each pair was then In fact, Hufnagel and Conca (1994) also noted
asked to perform two shopping tasks together with the importance of contextual clarity in collecting
the goals of purchasing a school bag and a watch, user response data. They argued that “the likeli-
respectively. Because communication support served hood of context-related errors and biases can be
as a within-subject factor, each pair would experience significantly reduced” by “specifying the popula-
two different types of communication support, i.e., tion to which comparisons should be made” (p. 56).
voice chat and text chat. The order of the two treat- Accordingly, changes were made in the study’s design
ment conditions was counterbalanced across differ- to provide subjects with a common reference frame-
ent groups. Similarly, the order in which participants work or context. Specifically, in the main experiment,
shopped for the two products was controlled; i.e., half before subjects were exposed to their formal tasks,
of the pairs purchased a watch for their first task and they were asked to perform a common task with the
a school bag for their second task, and the product goal of purchasing T-shirts, under a base condition
order was reversed for the other half of the partic- that used separate navigation support with text chat.
ipants. Upon completing each of the two shopping Also, the questionnaire was adjusted to ask the sub-
tasks, participants were asked to write down the prod- jects to compare the treatment condition they were
uct that they intended to purchase,5 then to complete assigned to with the base condition (see the appendix).
a questionnaire. The same design was used by Jiang and Benbasat
A pilot test with 32 subjects (16 pairs) was con- (2005) and Kim and Benbasat (2006).
ducted prior to the main experiment to identify any Two research assistants conducted the experiment,
problems that might occur. Subjects reported that it one with each subject in a separate room to provide
was difficult for them to answer survey questions, assistance if needed. The assistants were also asked to
e.g., to judge the social presence derived from the unobtrusively monitor whether the participants used
shopping experience using a Likert scale. For exam- the tool properly. With the permission of participants,
ple, some subjects tended to rate social presence rather we recorded the entire experimental sessions, includ-
highly because of the use of the normal individual ing the screen action and conversations between shop-
shopping experience as a benchmark; whereas others pers and their shopping partners, using Camtasia,
evaluated social presence lower when they compared a screen-capture software application. These screen
their experimental environment to physical collabora- files were viewed after each experiment. The review
tive shopping. results as well as the research assistants’ observations
The problem identified suggested that subjects had indicated that the experimental manipulations were
not as yet accumulated a uniform experience with successful across all four conditions, i.e., all subjects
collaborative online shopping needed for them to used the collaborative support technologies that were
form a mental reference benchmark to make their assigned to their groups.
judgments. This observation is also consistent with
Helson’s adaptation-level theory (Helson 1964), which 6. Data Analysis
suggests that people’s judgments are based on their
past experiences, a context or background, and a stim- 6.1. Subject Demographics and
ulus. Because the objective of this study is to evaluate Background Analysis
the effectiveness of different designs in the context of Among the 128 participants, 60 were females. The
ages of the participants ranged from 17 to 33. They
5
came from diverse academic backgrounds, such as
Subjects did not perform actual purchase immediately. However,
they were promised that one-fourth of them would be selected to
science, arts, engineering, and business. Almost one-
buy the particular product they chose and be reimbursed $60 on third (31%) had known their shopping partners for
showing us the transaction receipt. more than four years, 22% between two and fourZhu et al.: Let’s Shop Online Together: An Empirical Investigation of Collaborative Online Shopping Support
Information Systems Research 21(4), pp. 872–891, © 2010 INFORMS 881
years, 20% between one and two years, and 27% less Hence, shopping dyads’ conversation protocols
than one year. were collected. Voice chat protocols were transcribed
No significant differences were found between sub- into text format and analyzed later, together with text
jects randomly assigned to each of the four experi- chat protocols. Twenty-four thousand two hundered
mental conditions with respect to age, gender, past eighty-five communication exchanges were thus col-
Internet experience, the length of time shopping part- lected based on subjects’ conversations, in both voice
ners had known each other, sociability,6 and social and text.
intimacy.7 This evidence indicates that participants’ Two graduate research assistants, who were not
demographics were quite homogeneous across differ- aware of the study’s purpose, were asked to go
ent conditions. through all communication protocols and identify
those incidents that evidenced the occurrence of
6.2. Measurement uncoupling as well as subsequent communication
Seven items to measure social presence were adapted exchanges dedicated to resolving these uncoupling
from Short et al. (1976) (see the appendix). Because incidents. When faced with difficulties in coding, the
social presence was reported by both participants in two judges were allowed to refer to the correspond-
each shopping pair, the data were averaged as an indi- ing screen-capture files so as to better understand the
cator of social presence for this dyad and used in later context. To assess the reliability of coding and ensure
analysis. the validity of the data analysis, Cohen’s Kappa was
The evaluation of navigation coordination perfor- calculated to measure intercoder agreement (Todd and
mance encompasses the identification of uncoupling Benbasat 1987). The Kappa coefficient is 0.75, indi-
cating substantial agreement between the two coders
incidents. We noted that it was obtrusive to request
(Landis and Koch 1977). The differences were further
participants to report the occurrence of uncoupling
resolved when compromise was reached between the
during their shopping experience because that would
two judges based on their follow-up discussion.
have distorted shoppers’ natural shopping behavior.
Below are a few conversational examples of uncou-
We also noted that it was also impossible to accu-
pling incidents:
rately identify uncoupling incidents only based on
reviewing the screen-capture files of subjects’ behav- Example 1 (Separate Navigation and Voice Chat
ior because observers could not accurately gauge Condition, in Collaborative Search for Bags):
A: Yeah. Oh, we have another CalPack 19 inches.
shoppers’ browsing and navigation intentions and B: 19". Oh, okay. Where is it?
therefore were unable to determine whether shoppers A: Multipockets. And it’s only $30. $30. And it has dual
experienced any uncoupling. Hence, it was decided compartments.
B: Where ah? Where is it? Where is it?
to judge the occurrence of uncoupling by reviewing A: It’s the next page. Second in the middle from the top.
shoppers’ conversations as transcripts of conversa- B: Yeah?
tions can clearly show when people experience diffi- A: It’s pretty good actually.
culties as well as how they coordinate. B: Oh, okay. This one. Yeah.
Example 2 (Shared Navigation and Voice Chat Con-
6
Sociability represents the extent to which a person likes to do dition; in Collaborative Search for Bags):
things with other people. It was measured by four items based on A: Ya, ok Oh my goodness, do you see this Crumpler “wonder
Eid et al. (2003): “I usually prefer to do things alone,” “I really weenie” messenger bag? Oh, that’s horrible.
enjoy talking to people,” “I like to have a lot of people around me,” B: Which one? I don’t see it.
A: Wait the page you’re on, the second row
and “I would rather go my own way than be leader of others.”
B: The blue one?
7
Social intimacy represents the closeness of the social relationship A: No.
between shopping partners. It was measured by 17 items based
B: Ok, I see it. It’s only $25.
on Miller and Lefcourt (1982), such as “When you have leisure
time how often do you choose to spend it with him/her alone?,” Example 3 (Shared Navigation and Text Chat Con-
“How often do you keep very personal information to yourself dition; in Collaborative Search for Watches):
and do not share it with him/her?,” and “How often do you know A: This one? It looks good.
his/her affections?” B: Mm?Zhu et al.: Let’s Shop Online Together: An Empirical Investigation of Collaborative Online Shopping Support
882 Information Systems Research 21(4), pp. 872–891, © 2010 INFORMS
A: Where are we now? I am lost. Table 1 ANOVA Summary: The Number of Uncoupling Incidents per
B: Hey, I clicked on the wrong spot and went into that company. Product Discussed
I’m sorry. I should click the top one.
Source df Mean square F Sig.
A: That’s ok. Let’s go back.
Between-subjects
6.3. Preliminary Data Analysis Navigation support 1 192 2316 000
The two judges coded the products that were dis- Within-subjects
Communication support 1 029 243 015
cussed during collaborative online shopping. Con- Navigation support × 1 003 022 058
sistent with our prior expectation, voice chat leads Communication support
to significantly more products being discussed than
text chat does. Specifically, shopping dyads discussed
15.1 products per task on average when communicat- particular navigation support mode used, thus failing
ing via voice, as compared to 6.3 products per task to support H2.
when communicating via text (p < 001). In contrast, Table 3 indicates that navigation support has a sig-
navigation support did not make a difference in terms nificant main effect on the number of communication
of the number of products on which collaborative exchanges used to resolve each uncoupling incident,
shoppers exchanged ideas (p > 005). meaning that compared to separate navigation, shared
Because communication support is the within- navigation facilitates the resolution of uncoupling.
subject factor, there is a potential task order effect Hence, H1B is supported. The absence of interaction
(i.e., the order of text and voice chat tasks). Another effect suggests that the effect of communication sup-
concern pertaining to the internal validity of the port on the number of communication exchanges to
experiment is the possible confounding effects of resolve each uncoupling incident is not moderated by
product type (i.e., watches versus bags) and product the type of navigation support. Thus, H3 is not sup-
order (i.e., watches first and bags second versus bags ported. Furthermore, it is imperative to appropriately
first and watches second). A number of analyses of interpret the main effect of communication support.
variance (ANOVA) were performed on the collected As Table 4 shows, both text and voice lead to a similar
data by having these factors as covariates. Results number of communication exchanges to resolve each
show that none of these factors (i.e., task order, prod- uncoupling incident (5.47 versus 5.24, p > 005). How-
uct type, or product order) affects any of the depen- ever, given that it is much easier and faster to speak
dent variables (p > 005). than to type (Kinney and Watson 1992, Walther 1992,
Williams 1977), voice is likely to resolve uncoupling
6.4. ANOVA Results more efficiently than text.
ANOVA was conducted to examine the effects of Shared navigation and voice generate significantly
navigation support and communication support on higher social presence than separate navigation and
coordination performance and social presence. Cor- text, respectively (see Table 5 and 6). Therefore, H4
responding results are shown in Tables 1–6 and and H5 are supported. In line with our prediction,
Figures 2–4. the effect of navigation support is more prominent in
In particular, Table 1 shows that the effect of navi- the presence of text chat than in the presence of voice
gation support on the number of uncoupling incidents chat. In particular, when text chat is used, navigation
per product discussed is significant, suggesting that
shared navigation effectively reduces the occurrence
Table 2 Descriptive Statistics: The Number of
of uncoupling per product discussed as compared to
Uncoupling Incidents per Product Discussed
separate navigation. Therefore, H1A is supported. The
main effect of communication support and the inter- Text Voice Mean
action effect are not significant, indicating that voice Separate navigation 041 053 047
is not different from text in reducing the occurrence Shared navigation 019 025 022
Mean 030 039
of uncoupling per product discussed, regardless of theZhu et al.: Let’s Shop Online Together: An Empirical Investigation of Collaborative Online Shopping Support
Information Systems Research 21(4), pp. 872–891, © 2010 INFORMS 883
Table 3 ANOVA Summary: The Number of Communication Table 5 ANOVA Summary: Social Presence
Exchanges Used to Resolve Each Uncoupling Incident
Source df Mean square F Sig.
Source df Mean square F Sig.
Between-subjects
Between-subjects Navigation support 1 3946 3482 0000
Navigation support 1 4664 392 005 Within-subjects
Within-subjects Communication support 1 16474 27197 0000
Communication support 1 081 010 080 Navigation support × 1 616 1016 0002
Navigation support × 1 523 064 047 Communication support
Communication support
time (i.e., with visual common ground) fail to properly
support significantly boosts social presence (i.e., 0.17 coordinate their search for focal products. For exam-
for separate navigation versus 1.7 for shared naviga- ple, shopper A is inspecting product P while shop-
tion); when voice chat is used, navigation support can per B is inspecting product Q, although both are on
also increase social presence, but to a smaller extent the same screen. Hence, shopper A has no idea about
(i.e., 2.9 for separate navigation versus 3.5 for shared the product shopper B is referring to, and thus feels
navigation). Therefore, H6 is supported (p < 001). the loss of coordination (see Example 2 in §6.2).
Intrascreen navigational uncoupling occurs when
6.5. Supplementary Analysis
a shopper’s action affects his companion’s product
Recall that when proposing H1A and H1B, we
examination despite both looking at the same Web
described how and why uncoupling may occur in both
screen (i.e., with visual common ground). This typi-
separate and shared navigation conditions. In this sec-
cally happens in a shared navigation condition, where
tion, we explore in greater detail the formation of
the navigation of both shoppers are strictly tied
uncoupling. There are conceptually three major types
together, For example, shopper A may get confused
of uncoupling: interscreen uncoupling, intrascreen focal
by a sudden and unannounced navigation initiated
uncoupling, and intrascreen navigational uncoupling.
by shopper B (see Example 3 in §6.2). Consequently,
Interscreen uncoupling occurs when both collabora-
shopper B’s navigational action interrupts shopper A’s
tive shoppers are not exposed to the same Web screen
natural cognitive flow in product examination and
at the same time, and therefore, cannot accurately
increases the chances of loss of coordination.
understand what product the other party is referring
Based on this categorization, intrascreen focal
to (i.e., the absence of visual common ground). For
uncoupling may happen in both shared navigation
example, if shopper A is looking at screen X and shop-
and separate navigation. On the other hand, inter-
per B is looking at screen Y, interscreen uncoupling
screen uncoupling will happen in separate navigation
occurs when shopper A gets confused by shopper B’s
but not in shared navigation, where both shoppers
comments on a product on screen Y (see Example 1
always look at the same screen. In contrast, intrascreen
in §6.2).
navigational uncoupling will occur only in shared
Intrascreen focal uncoupling occurs when shoppers
navigation but not in separate navigation, where both
who are exposed to the same Web screen at the same
shoppers act freely on their own without interfer-
ing with each other. Specifically, our analysis of the
Table 4 Descriptive Statistics: The Number of
Communication Exchanges Used to Resolve Each
Uncoupling Incident Table 6 Descriptive Statistics: Social Presence
Text Voice Mean Text Voice Mean
Separate navigation 604 638 621 Separate navigation 017 287 152
Shared navigation 490 411 450 Shared navigation 171 354 263
Mean 547 524 Mean 094 321Zhu et al.: Let’s Shop Online Together: An Empirical Investigation of Collaborative Online Shopping Support
884 Information Systems Research 21(4), pp. 872–891, © 2010 INFORMS
Figure 2 Results on the Number of Uncoupling Incidents per Figure 3 Results on the Number of Communication Exchanges
Product Discussed Used to Resolve Each Uncoupling Incident
0.6 7
Number of communication exchange
to resolve each uncoupling incident
Separate-navigation
Number of uncoupling incidents per
Shared-navigation 6
0.5
5
product discussed
0.4
4
0.3
3
0.2 2
Separate-navigation
0.1 1
Shared-navigation
0
0 Text Voice
Text Voice
Communication support
Communication support
conversation transcripts shows that separate naviga- might initially have expected. In particular, the overall
tion leads to 3.2 interscreen uncoupling incidents per beneficial effect of shared navigation derives mainly
shopping task as compared to, by definition, zero for from its effect on eliminating interscreen uncoupling;
shared navigation (p < 001). Shared navigation leads in contrast and somewhat surprisingly, separate nav-
to 0.8 intrascreen navigational uncoupling incidents igation is better to suppress intrascreen navigational
per shopping task as compared to zero for separate uncoupling, and is not significantly different from
navigation (p < 001). Results also show that naviga- shared navigation in terms of intrascreen focal uncou-
tion support does not impact intrascreen focal uncou- pling. Hence, although common ground theory plays
pling, with 1.2 for separate navigation and 1.6 for a primary role in predicting overall uncoupling occur-
shared navigation (p > 01).8 rences, coordination performance, such as the two
types of intrascreen uncoupling, are also affected by
7. Discussion and Concluding the way in which collaborative shoppers manage and
Remarks coordinate their product search and navigation inten-
tions (Chu-Carroll and Carberry 2000). Therefore, an
7.1. Discussion of Results important research question is how to reduce the
The results show that shared navigation in general instances of intrascreen uncoupling to further enhance
is superior to separate navigation in reducing the shared navigation. In the next section, we will provide
occurrence of uncoupling and facilitating the resolu- several design suggestions that have the potential to
tion of uncoupling. Although the overall results are do so and need to be assessed in future studies.
consistent with common ground theory, the supple- Prior to the experiment, we expected an interac-
mentary analysis reveals deeper insights about the tion effect on coordination performance based on
formation of uncoupling and the specific applicabil- the media-task fit tenet suggested by Media Rich-
ity of the theory. It is observed that the difference
ness Theory (Daft and Lengel 1986, McGrath and
between shared navigation and separate navigation in
Hollingshead 1993). In particular, we proposed that for
reducing uncoupling is more complex than what one
separate navigation where coordination tasks become
highly equivocal, voice would perform better than
8
In fact, shared navigation has the potential to reduce intrascreen
text; but for shared navigation, communication sup-
focal uncoupling, if shoppers would always use mouse pointing
to show their companions the item they are looking at. However,
port would not make a difference. Indeed, our results
our data analysis indicates that users did not always do so in their have confirmed that separate navigation tasks are
product examination. more equivocal than shared navigation tasks becauseYou can also read
