RESEARCH

Social Presence: Influence on Bidders in Abstract We study effects of social processes’ Internet Auctions representation in Internet auctions using laboratory experiments. The inability of participating parties to learn about each SHEIZAF RAFAELI AND AVI NOY others’ bidding behaviour and to extend their comprehension beyond what is pre- sented in auction sites today decreases the social influence compared to traditional face-to-face auctions. This reality is about to change since new interaction and collaboration technologies are emerging. A conceptual model used here associates social factors influencing bidders and the resulting bidding behaviour. We utilized an online auction simulation framework that facilitates transmission of social cues during an auction and conducted empirical study of behaviour in both English and Dutch INTRODUCTION The Internet is perceived to auctions. We found that social presence, be ‘cold’ and impersonal. Many expressed by virtual presence and Internet auctions are software-based e-commerce websites enhance interpersonal information, significantly implementations of a traditional auc- usability and sociability by providing affects both bidding behaviour and market tion format. Online auctions may visual display of sales representatives, outcomes. have some inherent advantages. sales agents and virtual 3D environ- Keywords: auctions, bidding behavior, They allow access to a larger pool of ments. Social presence in traditional computer mediated communication, potential buyers; remove some time face-to-face commerce is replicated internet auctions, social presence, social and space constraints; decrease trans- in online environments. However, cognition online action costs and increase information though online auctioning is tremen- availability. At the same time, in a dously popular and social interaction computer-mediated environment, as is an essential part of traditional opposed to face-to-face interactions, auctions, Internet auction sites were buyers cannot manually touch or not affected by this trend. handle the item that is on sale, and This study tests how social presence Authors both sellers and buyers have more of bidders in online auctions affects Sheizaf Rafaeli difficulties sizing up the other. This bidding behaviour. The conceptual ([email protected]) is a professor of

Downloaded By: [Schmelich, Volker] At: 20:39 23 March 2010 limitation raises issues of perceived model emphasizes the relationship Information Systems and is the Director risk and imposes behavioural limita- between social factors affecting the of the Center for the Study of the tions. Since direct and indirect inter- individual bidder, the influence that is Information Society at the Graduate action among auction bidders is generated by others, and the resulting School of Business Administration, limited, the influence of one on the bidding behaviour. We utilize an University of Haifa, Israel. He has published extensively about computer- others is ostensibly precluded. innovative auction framework that mediated communication, virtual Many scholars have noted the embodies communication technol- communities and the value of influence of technology on the social ogy to extend and express social information. He has been involved in context in shaping the ways in which context during the auction. We report studying and building Internet-based the medium is used. There is even a on the results of a study conducted activities, and is interested in their social

popular slogan about the medium with this framework. and organizational impacts and 2005 Electronic Markets being the message. In this study the Two auctions models: English and potentials. ß focus is on the reverse influence – Dutch auctions were used in this Avi Noy from social context and layout to study. An English auction is an ([email protected]), Ph.D in Information Systems, is a researcher of technology. Social factors that affect ascending price auction and is the Copyright Volume 15 (2): 158–175. www.electronicmarkets.org DOI: 10.1080/10196780500083886 Internet auctions, e-commerce and bidding strategies in brick and mor- most popular auction mechanism in Computer Mediated Communication tar auctions may have different traditional and online environments. (CMC). He is a teaching fellow at the effects when using computerized A Dutch auction differs from the Graduate School of Business interaction mechanisms that enable English auction mechanism in its Administration, University of Haifa, transmission of social cues. descending price mechanism. This Israel. Electronic Markets Vol. 15 No 2 159

type of auction became famous as the mechanism used in 2. Environmental factors – these include the auction flowers market in Netherlands. mechanism and rules (Pinker et al. 2003) such as: The rest of this paper is organized as follows. First we initial price, reserved price, bid increment and review previous literature on auctions, social processes in auction ending rule. These factors affect the price, groups and emerging Internet communication and time of bidding and the number of bids. presentation technologies. Next we present our research 3. Social factors – these factors include people who are questions and describe the conceptual and research physically or virtually present. These factors also models. Hypotheses regarding the variables in the model affect the price, the time of bidding and the number are presented and the research method is described. of bids. Social factors are composed of three sub- Following a report of the results of our online groups: simulations we conclude by identifying the experiments’ limitations. Finally, we present the contribution and N Other bidders – the history of other bidders’ propose future research directions. behaviour may provide valuable information. The behaviour may also be perceived as having greater credibility than seller-originating content. For THEORETICAL BACKGROUND example: it was found that herding bias is negatively related to the bid price (Dholakia and Research on auctions Soltysinski 2001). The number of bidders at the auction may also affect bidding behaviour and Auctions are of research interest in the fields of have a negative effect. The ‘winner’s curse’, a term economics, marketing and consumer behaviour. which was coined while analysing loss of profits of Various laboratory experiments and empirical field winners at auctions, is explained by the rules of an studies have been conducted in traditional markets (see auction – the prize goes to whoever has the most Kagel 1995; Kagel and Levin 2002 for a thorough optimistic view of the value of the object being bid review). Internet auctions have been studied following upon. In some cases, this implies that the winner the increase in trade values. Many empirical field studies is the person who has overestimated the most. A collect data available in auction sites to analyse bidding larger number of bidders implies higher winner’s behaviour: In what auctions do bidders participate? How curse (Andreoni and Miller 1995; Kagel and Levin do various auction rules affect the bidding? (for example: 1986). Bapna 2003; Bapna et al. 2001, 2004; Dholakia and N The seller – a large body of empirical studies on Soltysinski 2001; Dholakia et al. 2002; Lucking-Reiley effects of seller’s reputation have been undertaken et al. 2000; Wilcox 2000). Most of these data were in the last few years (for overviews see: Dellarocas collected at eBay, the largest Internet auction site. These 2003; Resnick et al. 2003). It was found that studies relate to various aspects of online auctioning. seller’s past history and feedback rating affect the The following paragraph relates to studies of various probability of the sale as well as the price paid. factors affecting the individual bidder. History and feedback ratings have a measurable

Downloaded By: [Schmelich, Volker] At: 20:39 23 March 2010 effect on the auction price (Lucking-Reiley et al. 2000). Determinants of bidding behaviour N Other people – these are people who are not involved directly in the auction but may be the A range of studies identify the determinants of bidding cause for participating in the auction, such as in Internet auctions: price, time of bidding and number family members and other consumers. of bids. These determinants can be classified into three groups according to their source: While this classification identifies the influencing factors, 1. Personal factors – these factors include perceived risk a range of variables measures bidding behaviour: (Ha 2002; Resnick et al. 2003), internal indepen- dent estimates (Chakravarti et al. 2002; Dholakia 1. The decision to participate in a specific auction and Soltysinski 2001), pre-purchase information (Dholakia and Soltysinski 2001; Dholakia et al. search (Ha 2002; Hoyer and Brown 1990), level of 2002; Lucking-Reiley et al. 2000) and the number experience in previous auctions (Dholakia et al. of bidders in a specific auction (Wilcox 2000). 2002; Hayes et al. 1995; Shogren et al. 2000; 2. Times of entry and exit (Bapna 2003; Bapna et al. Wilcox 2000), information and learning (Hoffman 2004; Dholakia et al. 2002). et al. 1995; Jeitschko 1998) and enjoyment and 3. The values of last and winning bids (Bapna 2003; thrill derived from the participation (Herschlag and Bapna et al. 2001, 2004; Dholakia et al. 2002; Zwick 2000). These factors affect the price a bidder Lucking-Reiley et al. 2000; Resnick et al. 2003; is willing to pay. Wilcox 2000). 160 Sheizaf Rafaeli and Avi Noy & Influence on Bidders in Internet Auctions

4. Number of bids (Bapna et al. 2004; Dholakia et al. Sources and process of social influence are explained 2002; Lucking-Reiley et al. 2000; Resnick et al. by two general theories: normative and informational 2003). influence (Bearden and Etzel 1982; Kaplan 1987; Latane 1981). Normative and informational influence Contemporary consumers, including auction partici- models differ in the underlying mechanism of persuasion pants, are exposed to a great amount of information that they propose and in assumptions made about through both offline and online advertising. human nature. Normative influence presupposes an Interpersonal information in the form of Word-Of- emphasis on the group and on individual’s position in Mouth (WOM) is one of the oldest mechanisms in the it. The motivation for normative influence is based on history of human society. WOM has been identified as social rewards and is centred on interpersonal relations. an influencing factor on consumers both in traditional Informational influence emphasizes the task concerns and online environments (Dellarocas 2003; Senecal and and is driven by an individual’s desire to arrive at Nantel 2001). WOM is used to eliminate risk involved in accurate or correct decisions. If a change or a shift in the e-commerce transactions (Ha 2002) and it forms a basis decision occurs, it is caused by the incorporation of new for various online reputation mechanisms. data that was provided to the individual. When both In the next paragraph we explore the type of models are introduced, studies show that informational influence that is induced by other bidders and discuss influence is more common than normative influence and differences in the social influence between traditional causes stronger shifts (Kaplan and Miller 1983). (brick and mortar) and Computer-Mediated Social Impact Theory (Latane 1981) proposes that Communication (CMC) environments. social influence is a function of the strength (S), the immediacy (I), and number (N) of the people (or sources) present. Despite the lack of direct physical Social influence in traditional and CMC groups contact or immediacy, even mediated groups can be very real to their members (Postmes et al. 1998, 1999; To classify the type of influence imposed by other Sudweeks et al. 1998). Social influence in CMC groups bidders we need to explore the context in which is at the focus of many studies. These investigate how this influence arises and examine the conditions favour- limitations of social interaction, when using CMC, ing each type. We begin with a discussion of social affects group communication and decisions (Kiesler influence theory in traditional groups followed by social and Sproull 1992; Rafaeli 1988; Spears and Lea 1992; influence in CMC groups. Then, we examine relevant Sproull and Kiesler 1991). It was found that content and studies of social influence in traditional and online form of communication within a CMC group is auctions. normative and that conformity to group norms increases Groups are of interest in the fields of marketing, over time (Postmes et al. 2000). While early theories consumer studies and other applied research because an suggested that face-to-face interaction strengthen the individual consumer belongs to several groups and is interpersonal bonds that transmit social influence, influenced by others. Behaviour in groups is often more whereas isolation and anonymity weaken them, more readily predictable than that of individuals (Foxall et al. recent theories including the Social Identity model of Downloaded By: [Schmelich, Volker] At: 20:39 23 March 2010 1998). The interdependence of group members is made Deindividuation Effects (SIDE) (Reicher et al. 1995) enduring by the evolution of group identity that predicts that relative anonymity in CMC group can integrates beliefs, values and norms of the group. A enhance the influence of group norms. reference group, in its social context, is any individual or Bidders aggregated or congregated with other bidders collection of people whom an individual uses as a source present at an auction share some characteristics of a of attitudes, beliefs, values or behaviour (Wilkie 1994). group, though of course this group is not a community. The group is used by the individual as a source of A collection of individuals who congregate at an auction comparison for behaviour, personal attitudes and has a limited life span. The individuals are present at the performance. There are extensive implications of refer- same place, have a similar goal, and are influenced by one ence groups for consumer behaviour since a reference another. Earliest research insights in the literature about group may influence the individual. One example is of auctions claimed that a bidder should study the past purchases that are subject to group pressure. behaviour of competitors (Friedman 1956). Other Membership in a group involves accepting a degree of auction literature emphasized the importance of learning conformity. Conformity also occurs through the referent and being aware of the influence of other bidders (Kagel power of groups. An individual who conforms to a 1995; McAfee and McMillan 1987; Milgrom 1989; group gives the group the power to influence her Smith 1989; Wilson 1992). The influencing process in actions. The group acts as a standard of social auctions is strengthened as the vast majority of auctions comparison or as a point of reference. The individual contain a CV (Common Value) component. Items sold compares her behaviour to that of the group and adjusts in CV auctions do not have a definite market value. It is a to match what is observed. situation where the individual may need to refer to other Electronic Markets Vol. 15 No 2 161

bidders’ estimations. The individual may also change personalization (Daft and Lengel 1986). A medium valuation due to the influencing power of others. affects the process of communication and collaboration Research on auctions in the digital market reported between people in different ways due to which similar observations. Despite the anonymity, and only modalities it supports. It has been argued that media weak sense of a group, bidders are influenced by one differ in their capacity to carry data that are rich in another (Chakravarti et al. 2002; Herschlag and Zwick information (Daft and Lengel 1986; Rice 1993; Short 1999; Dholakia and Soltysinski 2001; Dholakia et al. et al. 1976). Social presence ranking depends on an 2002; Jeitschko 1998; Senecal and Nantel 2001; Varian interaction between the medium and the task at hand, 2000; Wilcox 2000). This influence occurs despite the and is based on the subjective judgment of the user limited ability to receive social cues in online auctions. (Lombard and Ditton 1997). Social cues, which consist of verbal and non-verbal In the years since the formulation of the medium signs, provide signals to behaviour in face-to-face richness approach, technology’s evolution has con- environments and are replicated in some online envir- tended with the challenge of limited social cues of onments (as will be discussed later). computers. In contrast, recent times were marked by the Though online auctioning is tremendously popular emergence of numerous communication and visualiza- and both direct and indirect social interactions are part tion technologies that enhance person-to-person inter- of traditional auctions, Internet auction sites do not action. These new arrangements include but are not provide virtual presence and real-time interaction limited to Instant Messaging (IM), interpersonal inter- capabilities. One reason for limiting communication faces, peer-to-peer, streaming audio, Voice over IP, and interaction among bidders during the auction is a video and much more. concern about an increase in fraud. With the incorporation of human or ‘human-like’ Absence of social cues in online auctions may limit faces into mediated environments the participants gain a bidders’ ability to analyse behaviour of other bidders and stronger sense of their community (Donath 2001). make improved decisions. In the next paragraph we People react differently to a ‘human-like’ interface. They discuss theories about social cues in different media and are more aroused and present themselves in a more describe several online mechanisms used today to positive light (Sproull et al. 1996). Mediated entities, as enhance social cues in a CMC environment. Later we social actors, have been introduced in computers just examine effects of incorporating similar mechanisms in after television (Lombard and Ditton 1997). Microsoft, an online auction framework. for example, presented ‘Bob’ which featured several characters and other products, tools and components with social interfaces and human like appearance such as Enhancing social cues in CMC ‘Office Assistant’ and ‘Microsoft Agent’. However, we should not overlook the distinction between human-like Social cues can be enigmatic in CMC, especially in the impersonations of real people, and those of imaginary absence of supporting technologies that facilitate inter- entities. action. Two theories should be addressed in this context: Numerous social and personalizing technologies were (1) Social Presence theory and (2) Media Richness adopted by educational and e-commerce sites. To name Downloaded By: [Schmelich, Volker] At: 20:39 23 March 2010 theory. a few: BuddySpace1 provides enhanced presence man- Social presence is a variable in mediated communica- agement in IM and other contexts by presenting tion (Short et al. 1976). It is defined as the sense of location maps and visual representations of the parties intimacy and immediacy, leading to increased satisfac- in conversation. MSN messenger,2 a popular IM tool tion, involvement, task performance and social interac- enables users to add their picture and even conduct tion (Lombard and Ditton 1997). It affects the way video chats. LivePerson3 offers online salesperson individuals perceive a medium and people whom they assistant tools each with intuitive ability to chat with a receive messages and communication from. The amount customer. OddCast4 creates broadband interfaces. of social presence varies among each type of medium. Included are realistic, computer-generated ‘spokespeo- While face-to-face yields a high level of social presence, ple’ who guide users through a variety of tasks (Berman computers have been found to have less social presence 2003). OddCast found that talking characters enhance than other mediums due to an absence of nonverbal cues users’ experience to positively improve most key business (Papacharissi and Rubin 2000). Social presence has two metrics, such as: ‘click through rate’, ‘visitors converted dimensions related to intimacy (interpersonal versus into registrants’, ‘conversions’ and ‘traffic’. Streaming mediated) and immediacy (asynchronous versus syn- animation of a face and other characters is also used in chronous). Social presence theory predicts that several mobile technologies. types of CMC can create in users a sense of intimacy and Although high bandwidth connections have elimi- immediacy. nated many technical barriers to making CMC features A medium’s richness is measured by its capacity for real world appearance, a large body of research (includ- immediate feedback, multiple cues, language variety and ing several research groups5) has focused on the 162 Sheizaf Rafaeli and Avi Noy & Influence on Bidders in Internet Auctions

presentation of participants and social interaction in a signals is Interpersonal Information which may be minimalist way. Interaction maps (Baker 2002), generated by consumer’s WOM. visualization of crowds (Erickson et al. 2002; Minar In this study we use these two sources as independent and Donath 1999), conversations (Donath et al. 1999; variables that may generate normative and informational Erickson et al. 1999) and other online activities (Donath influence. We are interested in finding the effects on 1995; Erickson et al. 2002) are a few examples. three dependent variables that comprise the ‘strategic bidding behaviour’ construct: (1) Purchasing decision – the purchasing decision variable gauges bidders’ eager- RESEARCH QUESTIONS, MODELS AND HYPOTHESES ness to succeed in the auction; (2) High bid – the price a bidder is willing to pay for the item; (3) Number of bids Earlier, we described evidence of social influence in – submitted by a bidder throughout the auction. This traditional and in Internet auctions. Of course, tradi- variable reflects the level of involvement in the auction tional and Internet auctions are very different. In (Bapna et al. 2004). Internet auctions, due to the lack of interaction among These three variables have been identified as measures bidders, a participant is limited to relevant information of bidding behaviour in an earlier section. We inten- that is transmitted by the explicit behaviour of other tionally held constant two other (real life) variables: (1) bidders. While interaction technologies have been Time of entry; and (2) Time of exit from the auction, as adopted in various e-commerce domains, online auc- we wanted to control the exposure time to the tions have been left out. Our research turns attention to manipulation of the independent variables. the following question: How does social presence of The research model presented in Figure 2 utilizes bidders in online auctions affects strategic bidding Virtual Presence and Interpersonal Information as the behaviour? The term ‘strategic bidding behaviour’ refers independent variables and Purchasing Decision, High to decisions and activities made during an auction, such Bid and Number of Bids as the dependent variables. as the decision to participate in a specific auction, time of Social Influence Theory forms a relationship between entry and time of exit from the auction, the highest bid the independent and dependent variables, while ‘Social and number of bids a bidder submits. Presence theory’, ‘Media Richness theory’ and the SIDE Studying this question will provide better under- model are used as part of the explanations for the standing of the impact of social influence in CMC hypotheses. environment and will test the SIDE model in a We assume that in settings with higher levels of competitive environment as online auctioning. Virtual Presence or Interpersonal Information a bidder Additionally, by analysing the influence of the under- may demonstrate: lying components of social presence and by utilizing the ‘Media Richness theory’ and ‘Social Presence theory’, it N A stronger willingness to make a purchase. The desire will provide recommendation to constructing richer to win the auction and make a purchase may increase online environments. due to the following reasons. First, in setups with The conceptual model presented in Figure 1 correlates higher levels of Virtual Presence or Interpersonal between three groups of influencing factors that were Information additional data becomes available and Downloaded By: [Schmelich, Volker] At: 20:39 23 March 2010 identified earlier: (1) Personal factors that characterize some uncertainties about the good and its qualities, the individual bidder as: perceived risk, internal inde- about the price and the trust of other bidders and the pendent estimate and level of enjoyment from the seller, are removed. Second, a higher level of auction; (2) Environmental factors which include information is related to a lower perceived risk (Ha auction’s mechanism and rules; and (3) Social factors: 2002; Resnick et al. 2003). Third, higher level of relating to other people who are physically or virtually Virtual Presence of other bidders may be perceived by present during the auction such as other bidders, the the individual as having a bigger crowd at the auction seller and other influencing people. These factors, and and may drive competition, which sometimes leads to especially social factors, generate social influence on the a winner’s curse (Andreoni and Miller 1995; Kagel individual bidder, which may have both normative and and Levin 1986). And finally, higher level of Social informational constructs and some other types of Presence increases enjoyment, involvement and thrill influence. As a result of the influence, strategic from the auction, which may increase the willingness behaviour of the bidder, which is demonstrated by: to win. Hence we hypothesize: selection of the auction site and the specific auction, time of entry and exit from the auction, her bids and H1(a): Purchasing Decision is more likely to occur when number of bids, may change. participating in an auction with a higher level of Virtual Social cues can be generated by multiple sources. In Presence of other bidders. this study we focus on Virtual Presence because of its H2(a): Purchasing Decision is more likely to occur when role in the ‘Social Presence theory’, the ‘Media Richness participating in an auction with higher level of Interpersonal theory’ and the SIDE model. A second source of social Information of and from other bidders. Electronic Markets Vol. 15 No 2 163

Figure 1. Conceptual model

N Paying lower prices. Two explanations lead to this inputs and thereby enhance the salience of the hypothesis. First, the ability to analyse social cues group and of its norms. The SIDE model predicts transmitted by other bidders and the addition of that in CMC settings with a high level of social information provided by interpersonal sources can presence, the social influence will be weaker com- improve decision-making process during the bidding pared to setups with relatively low level of phase. In this situation a bidder may examine other’s social presence, When the level of social presence is strategic behaviour from a different point of view and higher a bidder will feel confidence in her evaluation can arrive at an improved strategic behaviour in order and be less dependent in others’ valuation. In this to win the auction but avoid a winner’s curse. A situation a bidder will ignore other’s valuation when second explanation, which is derived from the SIDE it is higher then her own valuation. The formal model, is that anonymity may obscure individual hypotheses are: Downloaded By: [Schmelich, Volker] At: 20:39 23 March 2010

Figure 2. Research model 164 Sheizaf Rafaeli and Avi Noy & Influence on Bidders in Internet Auctions

H1(b): Winning at lower prices is more likely to occur when H3(b): Interpersonal Information will have a higher effect on participating in an auction with higher level of Virtual Presence bidders in an English auction compared to the effect on bidders of other bidders. of a Dutch auction. H2(b): Winning at lower prices is more likely to occur when participating in an auction with higher level of Interpersonal During the auction several output variables can be Information of and from other bidders. measured. Table 1 correlates between these variables and the dependent variables in each of the two auction N A tendency to bid less often. This outcome is explained models. by two main justifications. First, even if other partici- pants submit a high number of bids, the bidder may ignore it and rely on her own considerations. When the METHOD level of social presence is high, the bidder has a lower need to follow other’s strategic behaviour due to a We test the hypotheses by conducting two empirical removal of uncertainties and the integration of addi- experiments: one for each format. Each experiment tional information provided by interpersonal sources. included several sessions where human participants In this situation the bidder will submit fewer bids competed with simulated bidders in a simulated online compared to a situation where she has a higher need to auction. Data pertaining to the hypotheses were follow other’s behaviour. Second, at setups with a high collected during a series of eight meetings. To improve level of social presence a bidder will try to hide her reliability of the measures, a pre-test was carried out to valuation of the item by submitting less bids, as much as detect any necessary changes in the wording of the possible, since in this setup she may perceive other instruction pages, in the display and in the operation of bidders more realistically. The hypotheses derived from the auction software. these arguments are:

H1(c): When participating in an auction with higher level of Participants Virtual Presence of other bidders a participant will likely post fewer Number of Bids. Participants were undergraduate business and MBA H2(c): When participating in an auction with higher level of students in Israeli universities. Participation was volun- Interpersonal Information of and from other bidders a participant tary, but class credit points were offered as inducement will likely post fewer Number of Bids. to the best performers in each experiment. Some of the participants performed the experiments online via a home connection, while others performed the experi- Auction theorists identify four basic auction models: ment in a classroom computing laboratory. No differ- First Price Sealed Bid, English, Second Price and Dutch ences were found between location groups. A total of (McAfee and McMillan 1987; Milgrom 1989). Recent 188 students participated in the English auction experi- Internet auction research has focused on eBay partici- ment and 123 students participated in the Dutch pants. eBay uses a hybrid of the English and Second auction experiment. Downloaded By: [Schmelich, Volker] At: 20:39 23 March 2010 Price sealed bid formats. Other auction formats are often overlooked in recent studies but are common in traditional auction research. We chose to test our Experimental procedure hypotheses in the two contexts of English and Dutch formats. A262 design was used in each experiment. Each English and Dutch auctions differ in their underlying independent variable: Virtual Presence and Interpersonal bidding rules. An English auction is defined by an Information, was tested in two levels: high and low. ascending price mechanism and an ability to adjust bids While at the highest level presence or information during the auction. A Dutch auction has a descending components were included in the setup, at the low level price mechanism with only one bid allowed. However they were absent. This research design generated four both are characterized by their dynamic format: the conditions (Condition 1–4) as presented in Table 2. ability of participants to react during the auction by Figure 3 and Figure 4 present a screen shots of part of posting a bid. We assume that bidding behaviour in the conditions. these two auction formats will slightly vary due to higher The two experiments were identical in their design. levels of involvement and ability to adjust bids in the Participants were randomly assigned to one of four English auction. The formal hypotheses are: Presence and Information conditions and each trial was composed of four or more phases: (1) Instructions; (2) H3(a): Virtual Presence will have a higher effect on bidders in an Trial session; (3) First auction session; (4) Second English auction compared to the effect on bidders of a Dutch auction session; and (optional) (5) Additional auction auction. sessions. The total number of auction sessions a Electronic Markets Vol. 15 No 2 165

Table 1. Relationship between variables

English Variable Meaning Measure dependent variable auction Dutch auction

High Bid Highest bid posted by a participant Purchasing decision + 2 Win Bid Winning bid of a participant Bid ++ Num of bids Number of bids submitted Number of Bids + 2 Num of bids if won Number of bids submitted by a winner Number of Bids + 2 Wins Percentage of winning participants Purchasing decision ++ Bids Zero Percentage of participants that didn’t submit a bid Purchasing decision + 2 Continue Percentage of participants that continued to Purchasing decision ++ additional session

(+) Was measured (2) Was not measured

session. Results included a list of all bids placed during Table 2. Conditions of the experiment the session in descending order. However participants Independent Condition Condition Condition Condition were not informed about the value of the item and did variable 1 2 3 4 not know whether the winning bid is relatively high or low. Final results comparing the winning bids to a Virtual Presence Low High Low High market value (which was the mean of all winning bids) Interpersonal Low Low High High included names of the best performers. These were Information presented following the completion of the assignment.

Auction rules. No real or artificial money was used, no participant would experience was not limited and varied real payment was made and no real item was awarded to between participants. Every session included participa- the winner. Winning determination was based on buying tion in an online auction taking 2–3 minutes and at lower prices relative to an unknown market value. In followed by a period that allowed viewing results. The every session a single item (an antique watch) was ‘sold’. only difference between the two experiments was in the This item has a high CV component as its market price underlying auction mechanism. was unclear and participants bought it for trading Experiments were constructed entirely online to purposes. In the English auction experiment participants reduce experimenter availability or interference. In order joined a seemingly already ongoing auction. The auction to eliminate external distraction participants were ending rule was ‘hard’ (meaning the that the ending instructed to surf only to the auction site and to avoid time was pre-defined) and a bidder with a highest bid at

Downloaded By: [Schmelich, Volker] At: 20:39 23 March 2010 any other Internet-related activities throughout their the end ‘won’ the item and ‘paid’ according to this bid. participation. During the experiment participants As the auction progressed, five simulated bidders joined received online feedback about their results after each the arena and actively participated in the auction. Each

Figure 3. Conditions 1 (left) and 4 of the English auction 166 Sheizaf Rafaeli and Avi Noy & Influence on Bidders in Internet Auctions

Figure 4. Conditions 1 (left) and 4 of the Dutch auction

of the joining simulated bidders was endowed with an simplified the characterizing variables to a minimal set independent predefined bidding strategy and a price which includes two major groups. The first group of limit higher than the initial price. These characteristics of variables controls bidding strategy, while the second the simulated bidders were interpreted by the simulation group controls the interaction of the simulated bidders software into a sequence of bids (in the English auction) via a chat mechanism. The simulated bidders were until they reached their price limit. programmed to have a similar bidding strategy in all In each session of the Dutch auction participants experimental conditions; however their visual appear- joined an auction just before it started and initiated it by ance and interaction abilities were dissimilar. The pressing a button. Five additional simulated bidders also bidding strategy was based on a maximal price limit, participated in the auction, each having an independent which was assigned randomly to each of them, and on price limit. Upon reaching this limit, the simulated predefined parameters, which controlled their response bidders were programmed to post a bid. time and willingness to react to a previous bid and post a higher bid during different phases of the English Auction simulation framework. The experiment tool auction. was an auction simulation framework program devel- oped for research and teaching purposes (Rafaeli et al. Social and informational cues. Four conditions were 2003). This framework is composed of software created as two levels of Virtual Presence and two levels of components operating on both client and server Interpersonal Informational. Cues were implemented by computers and simulating an Internet auction site, inducing different components to the auction site. auction mechanism and some of the auction bidders. Virtual Presence components included: Downloaded By: [Schmelich, Volker] At: 20:39 23 March 2010 The client side of the framework was implemented in Java, HTML pages and JavaScripts, operating in any Java 1. Pictures of other bidders’ faces – As other bidders enabled Internet browser. The server’s components joined the arena, their pictures were shown on the include CGI/Perl scripts and operated on an Apache screen. Web sever. 2. Instant messaging – simulation of an online chat room. Since other bidders were simulated, this chat Simulation of other bidders. All simulations are based process was simulated too. Messages ostensibly on a simplified but accurate representation of some posted by simulated bidders were displayed and aspects of the real world (Maidment and Bronstein the participant could send messages of their own. A 1973). A use of a simulation program in our context very simple feedback mechanism was implemented enabled us to control the underlying behavioural to respond to these sentences (Rafaeli and Noy characteristics of the simulated bidders. Nevertheless, 2002), preserving the illusion of a chat room. these were not independent variables and were not 3. Auction proxy – An auction proxy was constructed manipulated. based on Erickson (2001) who visualized a socially Developing a social simulation that models behaviour translucent traditional bidding room by minimal in a social context is harder than a development of a graphical representation. A bounding circle repre- physical simulation. There is no accurate description of sents the bidding room and bidders are represented the behaviour and behaviour of people and groups can’t by colour dots, which changed according to their be predicted because of the numerous variables with position in the auction and according to others’ which to contend. For the purpose of this study we bids. Electronic Markets Vol. 15 No 2 167

4. Additional presence cues – displaying the number of purchasing decision’s component. Participants who offer the previous interested bidders, displaying names of higher bids and win are more likely to desire the item. In the bidders near their bids (instead of presenting a contrast, if a participant is involved in an auction without general data such as: ‘Bidder No. 4’). placing a bid, she may exhibit lower buying intentions. The ‘Continue’ variable provides supplementary indica- tion about a willingness to continue to an additional The Interpersonal Information components were: auction session, and hence, is also attributed to a 1. Names of other bidders – these are rarely presented purchasing decision’s components. in real auction sites. Hypotheses H1(b) and H2(b) were evaluated through 2. Bidders’ information – Auction sites provide data the ‘Win Bid’ variable. Hypotheses H1(c) and H2(c) about the reputation of the seller and sometimes of were evaluated through the ‘Num of bids’ and ‘Num of the buyer. The simulation framework used here bids if won’ variables. provides additional data about the bidders: (1) Type of participation which was classified into: business or private participation, indicating the intentions and English auction results the eagerness to win; (2) Past winning history; (3) Experience level, which was classified to either 188 different bidders participated in 734 different professional or amateur levels; (3) Bidding interest auction sessions. Four additional sessions were disqua- – level of interest in the current auction, which was lified and dropped from the data set due to extreme classified to: high, medium and low levels. outlier, inconsistent and erroneous data. Some of the 3. Recommendations – During auction sessions, inter- participants quit after a single session and some played personal recommendation about the auction and the up to 26 auction sessions. Overall there were 3.9 item on sale were presented to the participants. sessions on average per participant. Participants won in 473 sessions (64.44%) and lost in 230 (31.33%). In 31 auctions (4.22%) no bid was submitted. The range of winning bid prices was $290 to $1000 and market value (mean winning bid) was $517.73. Up to 18 bids were RESULTS posted in each session with a mean number of 3.31 bids. Winning bidders posted up to 16 bids with a mean of Data analysis consisted of classifying auction sessions 3.22 bids. into groups and comparing the independent results of In their first auction session 109 (57.97%) participants the first auction session of each participant. The won and 79 (42.02%) lost. Mean winning bid was additional auction sessions contained a reduced amount $504.30 and mean number of bids of a participant was of data and were not independent. Auction sessions were 4.71 bids per session. classified into one of the following groups according to the condition: Effects of Virtual Presence and Interpersonal N LP (Low Presence – low level of Virtual Presence)– Information. Results strongly support H1(b) and par- Downloaded By: [Schmelich, Volker] At: 20:39 23 March 2010 Conditions 1 and 3; tially support H1(a). Virtual Presence created significant N HP (High Presence – high level of Virtual Presence)– difference in ‘High Bid’ (see Figure 5). HP (High Conditions 2 and 4; Presence) participants produced a ‘High Bid’ lower by N LI (Low Information – low level of Interpersonal 12.34% than LP (Low Presence) participants (F(1, Information) – Conditions 1 and 2; and 186)57.35 p,0.01). A significant difference of N HI (High Information – high level of Interpersonal 20.86% was found in the ‘Win Bid’ variable 5 , Information) – Conditions 3 and 4. (F(1,107) 21.01 p 0.0001). Results strongly support H2(b) and partially support H2(a). Interpersonal Information did not produce any Hypotheses were tested by comparing results in LP and significant difference in the ‘High Bid’ parameter. HP groups, and in LI and HI groups in both English However, a significant difference was found in the and Dutch setups. To test differences in the ‘High Bid’, ‘Win Bid’ parameter. HI (High Information) partici- ‘Win Bid’, ‘Num of bids’ and ‘Num of bids if won’, four pants won at 10.19% lower prices than LI (Low separate one-way analysis of variance (ANOVA) were Information) participants (F(1,107)56.41 p,0.01). performed. Difference in the ‘Wins’, ‘Bids Zero’ and These results are presented in Figure 5. ‘Continue’ variables were measured by performing four HP participants posted 30.58% fewer bids (‘Num of separate Chi-Square tests. bids’) than LP participants (F(1,186)59.78 p,0.005). Hypotheses H1(a) and H2(a) were evaluated through Winning participants showed the same tendencies. the results of the ‘Wins’, ‘Bids Zero’, ‘Continue’ and ‘Num of Bids if won’ in HP was 25.96% lower than in ‘High Bid’ variables. The result of each is attributed to a LP (F(1,107)53.95 p,0.05). HI participants posted 168 Sheizaf Rafaeli and Avi Noy & Influence on Bidders in Internet Auctions

Figure 5. High and Win bid in the English auction

8.10% fewer bids than LI participants (insignificant). Regarding the Virtual Presence effect; results strongly Winning bidders at both HI and LI placed almost the support H1(b) and H1(c), namely that with higher level same number of bids (see Figure 6). of Virtual Presence of other bidders, winning is expected HP participants won 26.80% more times than LP at lower prices and with less bids. Results partially participants (t55.37 p,0.05). LP participants support H1(a) (‘High Bid’ and ‘Wins’ percentage were ‘Continue’ 8.02% more times to the next session than significantly higher in HP but ‘Continue’ and ‘Bid Zero’ HP participants (insignificant). The number of partici- results were not), namely that with higher level of pants that ‘Bid Zero’ in the two conditions was almost Virtual Presence winning may, but will not necessarily be identical (five in HP, four in LP). HI participants won achieved. 9.63% more times than LI participant (see Figure 7). LI Regarding Interpersonal Information effect; results participants ‘Continue’ to the subsequent auction strongly support H2(b), namely that with higher level of session 7% more times than HI participants and the Interpersonal Information of and from bidders, winning number of participants who did not bid at all was almost is expected to occur. On the other hand results provide identical. None of these differences are significant. less support to H2(a), namely that with higher level of Downloaded By: [Schmelich, Volker] At: 20:39 23 March 2010

Figure 6. Number of bids in the English auction Electronic Markets Vol. 15 No 2 169

Figure 7. Wins and continuation percentages in the English auction

Interpersonal Information winning bids may occur in a (insignificant). HI participants won more times lower price. H2(c) was rejected, namely that with higher (57.42%) than LI participants (t510.56, p50.0012). levels of Interpersonal Information a bidder does not They also exhibited a higher tendency (7.49%) to necessarily post fewer bids. participate in an additional auction session than LI participants (see Figure 9). Results regarding Virtual Presence effect provides Dutch auction results strong support for H1(b) and partial support for H1(a) (the ‘Wins’ percentage was significantly higher 123 different bidders participated in 426 auction in HP, but HP participants ‘Continue’ fewer times than sessions. Some of the bidders participated in a single LP participants). Results regarding Interpersonal session and some continued playing up to 37 sessions for Information effect provides some support to H2(a) a single participant. The mean number of auction (‘Wins’ was significantly higher at HI and ‘Continue’

Downloaded By: [Schmelich, Volker] At: 20:39 23 March 2010 sessions per participant was 3.46. Participants won 268 was only higher) but H2(b) was rejected. (62.91%) sessions and lost 158 (37.08%) sessions. In their first auction session 78 (63.41%) participants won and 45 (36.59%) lost. The range of winning bids was English and Dutch auction comparison $268 to $997 and market value was $509.75. Hypotheses H3(a) and H3(b) received partial support. Effects of Virtual Presence and Interpersonal The change in the winning percentages (‘Wins’) due to Information. Results strongly support H1(b) and par- Virtual Presence is higher in the English auction tially support H1(a). Virtual Presence produced signifi- compared to the Dutch auction. The change in cant difference in the ‘Win Bid’ (see Figure 8). HP ‘Win Bid’ due to Virtual Presence is higher in the participants won at 19.40% lower bids than LP Dutch auction compared to the English auction. The participants (F(1, 76)57.1 p,0.01). Interpersonal change in the winning percentages (‘Wins’) due to Information registered an opposite effect. Results Interpersonal Information is higher in the Dutch auction partially support H2(a) but did not support H2(b). HI compared to the English auction. The change in ‘Win participants won at 4.24% higher prices than Bid’ due to Interpersonal Information in the Dutch LI participants (this difference is not statistically auction is in a reverse direction compared to the English significant). auction. HP participants won significantly more times Table 3 summarizes the hypotheses and whether they (t54.15, p,0.05) than LP participants (32.07%). They received any support (significant or partial) or were ‘Continue’ 9.32% fewer times to the next auction session rejected. 170 Sheizaf Rafaeli and Avi Noy & Influence on Bidders in Internet Auctions

Figure 8. Win Bid in the Dutch auction

DISCUSSION indirect communication is part of traditional auctions and may influence bidding behaviour. Though online auctions are popular and have operated In this study we found evidence that the introduction successfully without enabling social interaction among of direct and indirect communication among bidders in bidders, it has been observed that both direct and Internet auctions as expressed in Virtual Presence and Downloaded By: [Schmelich, Volker] At: 20:39 23 March 2010

Figure 9. Wins and continuation percentages in the Dutch auction Electronic Markets Vol. 15 No 2 171

Interpersonal Information produced social influence that experiment provide full support for H1(b) and partial affected bidding behaviour in both English and Dutch support for H1(a). auctions. When introducing Interpersonal Information cues, Results strongly support hypotheses H1(b) and H1(c) winning bid was reduced in the English setting, thus and provide strong support to some of the constructs of H2(b) received strong support. Bidders did not submit H1(a) in the English auction. ‘High bid’ and ‘Num of fewer bids, thus H2(c) was not supported. H2(a) was bids’ of both participants and winners were significantly partially supported. Percentage of ‘Wins’ was increased reduced due to Virtual Presence effect. and the percentage of ‘Bid Zero’ and ‘Continue’ were Participants’ eagerness to succeed (as measured by decreased. These findings allude to a presence effect, but ‘Wins’, ‘Bids Zero’ and ‘Continue’) was strengthened were not significant. In the Dutch auction settings due to an increase in the winning percentage, but the ‘Wins’ percentage was significantly affected by number of participants who did not place a bid was not Interpersonal Information providing some support to reduced, and ‘Continue’ percentage was not increased. H2(a), but H2(b) was not supported, namely ‘Win Bid’ Taking into account that only a negligible number of was not reduced. bidders did not place a bid (4–5 participants), and that Results provide some support for H3(a) and H3(b). continuation is not an implicit determinant of purchas- The difference in the percentage of ‘Wins’ due to ing decision, since satisfaction with the current settings Virtual Presence was higher in the English auction than may lead to a decision not to leave the state; we can in the Dutch auction. However, difference in ‘Win Bid’ conclude that Purchasing Decision was also affected by was higher in the Dutch auction. The difference in the the virtual presence. The results of the Dutch auction percentage of ‘Wins’ due to Interpersonal Information

Table 3. Summary of results

Hypothesis English Auction Dutch Auction

H1(a) – Stronger purchasing Partial support Partial support decision as a result of virtual % Wins: (t55.37 p50.02) % Wins: (t54.15 p50.041) presence % Continue: (t52.36 p50.12) % Continue: (t53.21 p50.073) % Bid Zero: (t50.21 p50.64) H1(b) – Lower bids as a result Strong support Strong support of virtual presence High Bid: (F(1, 186)57.35 p,0.01) Win Bid: (F(1,76)57.1 p50.0094) Win Bid: (F(1,107)521.01 p,0.0001) H1(c) – Fewer bids as a result Strong support N.A. of virtual presence Num of Bids:(F(1,186)59.78 p50.002) Num of Bids if won (F(1,107)53.95 p50.049) H2(a) – Stronger purchasing Partial support Partial support Downloaded By: [Schmelich, Volker] At: 20:39 23 March 2010 decision as a result of % Wins: (t50.63 p50.43) % Wins: (t510.56 p50.0012) interpersonal information % Continue: (t51.87 p50.17) % Continue: (t51.81 p50.1781) % Bid Zero: (t50.01 p50.92) H2(b) – Lower bids as a result Strong support Rejected of interpersonal information High Bid: (F(1,186)50.7 p50.40) Win Bid: (F(1,76)50.22 p50.64) Win Bid: (F(1,107)56.41 p,0.01) H2(c) – Fewer bids as a result Rejected N.A of interpersonal information Num of Bids: (F(1,186)50.49 p50.48) Num of Bids is won: (F(1,107)50p51) H3(a) – Higher effect of virtual Partial support presence in an English auction The change in the winning percentages (% Wins) due to Virtual Presence is higher in the English auction compared to the Dutch auction. The change in Win Bid due to Virtual Presence is higher in the Dutch auction compared to the English auction. H3(b) – Higher effect of Partial support interpersonal information in The change in the winning percentages (% Wins) due to Interpersonal Information is higher in the Dutch an English auction auction compared to the English auction. The change in Win Bid due to Interpersonal Information is in a reverse direction in the Dutch auction compared to the English auction. 172 Sheizaf Rafaeli and Avi Noy & Influence on Bidders in Internet Auctions

was higher in the Dutch auction but the difference in the understanding of the dynamics of auctions. They ‘Win Bid’ was higher in the English auction. might contribute to the broader study of social networks The effect of Interpersonal Information was not as and the influence flows in them. Another field that may pronounced as the effect of Virtual Presence. One benefit from this study is the impact of social influence possible explanation is that most of the participants on economic behaviour. were inexperienced bidders. Studies of experience in The study suffers from several limitations. First, we Internet auctions realized differences between experi- use a simulation as the primary research method. enced and novice bidders (Andreoni and Miller 1995; Although participants were not aware of the fact that Dholakia et al. 2002; Wilcox 2000). The firsts are more they competed with simulated bidders, real-life bidders likely to appreciate information and informational cues may act slightly different under human or mechanized than inexperienced bidders who (in general) are more contexts. The major reason for using a simulation was to emotional. This explanation is supported by the control for social presence cues. However, a growth in observation in the literature that informational influence the use of autonomous bidding agents (e.g., Proxy is more cognitive than normative influence and less bidding in eBay) in the past few years (Bapna et al. emotional (Kaplan and Miller 1983). Another explana- 2004) may mitigate this problem. We suggest that field tion lies in the enthusiasm with facial cues and studies will supplement our findings when such cues perception of others’ presence signals. The addition of become more prevalent in real implementations of these may have been deemed more attractive at earliest online auctions. stages of the experiment (first auction session) than A second limitation is that auctions were not about Interpersonal Information which mostly was displayed in real money or real items. While winning was encouraged a textual manner. To verify this explanation we analysed by having credit point incentives, participants had small all auction sessions in each experiment. We found that stakes compared to real-life auctions. Wilson (1992) and Virtual Presence effect, which caused lower winning bids Kagel (1995) addressed the absence of economic risk in and higher tendencies to win, declines as participants experimental settings. Despite the differences between became familiar with the setting and results of the effect field studies and experiments, results are complementary becomes comparable to results of the effect of (Wilson 1992) and enhance understanding. Interpersonal Information. However there is another Another limitation of the method is not controlling explanation for a change in bidders’ outcome overtime. for the experience level of participants. This limitation Because participants experienced the auction in con- was moderated by randomly and evenly assigning secutive trials they might have learned the bidding rules participants to each condition and by exposing them to of the agents. Once these rules were extracted it a new auction setting and rules with which no becomes easier to play against the agents and exploit participant had any prior experience. the rules. Internet auction sites today do not provide means for communication among bidders, probably because of Research contribution concerns about fraud. Bidders and sellers may harness technology and use it to win unethically or unlawfully. There are numerous theoretical and practical implica- Downloaded By: [Schmelich, Volker] At: 20:39 23 March 2010 Another explanation is that some bidders and sellers may tions of this study. The key theoretical implication is in prefer the anonymity enabled by the Internet. However, demonstrating the social influence determinants in the inability to interact may damage bidders’ decision Internet auctions. We found that normative and processes by limiting information provided by the social information influence affects bidding behaviour. Our environment. Previous research shows that bidders rely findings support conclusions of the SIDE model. on past behaviour of sellers and other bidders. In this Relative anonymity of auction bidders enhances the study we look one step ahead. We found that, when influence of the group valuation. The study also provided, Virtual Presence and Interpersonal contributes to ongoing research in marketing, consumer Information, generate social influence which affects behaviour and social psychology. From a practical bidding behaviour. perspective the research demonstrates the possibility to construct social influence even with relatively minimal technological cues. It suggests further directions LIMITATIONS, CONTRIBUTION AND FURTHER to commercial auction sites and other e-commerce RESEARCH initiatives.

Research limitations Further research The implications of our findings are of theoretical interest, and they have importance for the design of Social software and the exploitation of communication online auction systems. These findings are of interest to avenues between anonymous and semi-anonymous Electronic Markets Vol. 15 No 2 173

players on the net are growing in popularity and use. Berman, D. K. (2003) ‘Special Report: Technology, Next Up: This phenomenon is, therefore, of further concern. We A Look at New Broadband Services that are in the Pipeline’, propose further research in several directions. The first is The Wall Street Journal 13 October 2003. to complement the experimental paradigm described Chakravarti, D., Greenleaf, E., Sinha, A., Cheema, A., Cox, J. here with field studies. Can better data regarding C., Friedman, D., Ho, T. H., Isaac, M. R., Mitchell, A. A., informational and normative influence be collected in Rapoport, A., Rothkopf, M. H., Srivastava, J. and Zwick, R. real settings? It should also be interesting to extend the (2002) ‘Auctions: Research Opportunities in Marketing’, scope of this study to other forms of online interaction Marketing Letters 13(3): 281–96. where a group has a power to influence the individual. Daft, R. L. and Lengel, R. H. (1986) ‘Organizational Examples are online games and online gambling Information Requirements, Media Richness and Structural settings. We believe that in these settings, because of Design’, Management Science 32(5): 554–71. the similar influence conditions, influence results will Dellarocas, C. (2003) ‘The Digitization of Word of Mouth: replicate those found here. A third direction is an Promise and Challenges of Online Feedback Mechanisms’, extension of the research to other types of e-commerce Management Science 49(10): 1407–24. where the influencing power of the group is lower, such Dholakia, U. M. and Soltysinski, K. (2001) ‘Coveted or as electronic stores. In this type of setting social presence Overlooked? The Psychology of Bidding for Comparable of the seller may act as a source of influence. Listings in Digital Auctions’, Marketing Letters 12(3): 225–37. Dholakia, U. M., Basuroy, S. and Soltysinski, K. (2002) ‘Auction or Agent (or Both)? A Study of Moderators of the Notes Herding Bias in Digital Auctions’, International Journal of 1. http://kmi.open.ac.uk/projects/buddyspace/ Research in Marketing 19: 115–30. 2 http://messenger.msn.com/ Donath, J. (1995) ‘Visual Who: Animating the Affinities and 3. http://www.liveperson.com/ Activities of an Electronic Community’, Proceedings of 4. http://www.oddcast.com ACM Multimedia ’95 5–9 November 1995, San Francisco, 5. For example, Social media group at MIT Media Lab: CA. http://smg.media.mit.edu/; Social computing group at Donath, J. (2001) ‘Mediated Faces’, Cognitive Technology, IBM: http://www.research.ibm.com/SocialComputing/; 373–90. Social Computing Group at Microsoft: http:// Donath, J., Karahalios, K. and Vie´gas, F. (1999) ‘Visualizing research.microsoft.com/scg/ Conversation’, Proceedings of HICSS-32, Maui, HI, 5–8 January 1999. Erickson, T. (2001) ‘The Auction Social Proxy: Portraying Social Activity in Online Auctions’, in, Proceedings of the References Seventh European Conference on Computer Supported Andreoni, J. and Miller, J. (1995) ‘Auctions with Artificial Cooperative Work, ECSCW 2001 (poster). Adaptive Agents’, Games and Economic Behavior 10(1): Erickson, T., Halverson, C., Kellogg, W. A., Laff, M. and

Downloaded By: [Schmelich, Volker] At: 20:39 23 March 2010 39–64. Wolf, T. (2002) ‘Social Translucence: Designing Social Baker, A. J. (2002) ‘Intimate Bonding Through Online Infrastructures that Make Collective Activity Visible’, Communities: Just Like Real Life?’, Presented at the Communications of the ACM 45(4): 40–4. International Conference Association of Internet Erickson, T., Smith, D. N., Kellogg, W. A., Laff, M. R., Researchers (AoIR), Internet Research 3.0: NET/WORK/ Richards, J. T. and Bradner, E. (1999) ‘Socially THEORY, Maastricht, The Netherlands 13–16 October Translucent Systems: Social Proxies, Persistent 2002. Conversation, and the Design of ‘‘Babble’’’, in, Human Bapna, R. (2003) ‘When Snipers Become Predators: Can Factors in Computing Systems: The Proceedings of CHI ’99. Mechanism Design Save Online Auctions?’, ACM Press. Communications of the ACM 46(12): 152–58. Foxall, G. R., Goldsmith, R. E. and Brown, S. (1998) Bapna, R., Goes, P. and Gupta, A. (2001) ‘Comparative Consumer Psychology for Marketing (2nd edn), UK: Analysis of Multi-item Online Auctions: Evidence From the Thomson Business Press. Laboratory’, Decision Support Systems 32: 135–53. Friedman, L. (1956) ‘A Competitive Bidding Strategy’, Bapna, R., Goes, P., Gupta, A. and Jin, Y. (2004) ‘User Operations Research 4: 104–12. Heterogeneity and its Impact on Electronic Auction Market Ha, H. (2002) ‘The Effects of Consumer Risk Perception on Design: An Empirical Exploration’, MIS Quarterly 28(1): Pre-purchase Information in Online Auctions: Brand, 21–43. Word-of-Mouth, and Customized Information’, Journal of Bearden, W. O. and Etzel, M. J. (1982) ‘Reference Group Computer Mediated Communication 8(1), online at: Influence on Product and Brand Purchase Decisions’, http://www.ascusc.org/jcmc/vol8/issue1/ha.html Journal of Consumer Research 9: 183–94. [accessed 10 May 2004]. 174 Sheizaf Rafaeli and Avi Noy & Influence on Bidders in Internet Auctions

Hayes, D., Shogren, J., Shin, S. and Kliebenstein, J. (1995) Papacharissi, Z. and Rubin, A. M. (2000) ‘Predictors of ‘Valuing Food Safety in Experimental Auction Markets’, Internet Use’, Journal of Broadcasting & Electronic Media American Journal of Agricultural Economics 77(1): 40–53. 4(2): 175–96. Herschlag, M. and Zwick, R. (2000) ‘Internet Auctions – Pinker, E. J., Seidman, A. and Vakrat, Y. (2003) ‘Managing Popular and Professional Literature Review’, Quarterly Online Auctions: Current Business and Research Issues’, Journal of Electronic Commerce 1(2): 161–86. Management Science 49(11): 1457–84. Hoffman, D. L., Novak, T. P. and Chatterjee, P. (1995) Postmes, T., Spears, R. and Lea, M. (1998) ‘Breaching or ‘Commercial Scenarios for the Web: Opportunities and Building Social Boundaries? SIDE-effects of Computer- Challenges’, Journal of Computer Mediated mediated Communication’, Communication Research 25: Communication 1(3), online at: http://www.ascusc.org/ 689–715. jcmc/vol1/issue3/hoffman.html [accessed 10 May 2004]. Postmes, T., Spears, R. and Lea, M. (1999) ‘Social Identity, Hoyer, W. D. and Brown, S. P. (1990) ‘Effects of Brand Group Norms, and Social Influence’, in N. Ellemers, R. Awareness on Choice for a Common, Repeat-purchase Spears and B. Doosje (eds) Social Identity: Context, Product’, Journal of Consumer Research 17: 141–48. Commitment, Content, Oxford: Blackwell, pp. 164–83. Jeitschko, T. D. (1998) ‘Learning in Sequential Auctions’, Postmes, T., Spears, R. and Lea, M. (2000) ‘The Emergence Southern Economic Journal 65(1): 98–112. and Development of Group Norms in Computer-mediated Kagel, J. H. (1995) ‘Auctions: A Survey of Experimental Communication’, Human Communication Research 26: Research’, in John H. Kagel and Alvin E. Roth (eds) 341–71. Handbook of Experimental Economics, Princeton: Princeton Rafaeli, S. (1988) ‘Interactivity: From New Media to University Press, pp. 501–86. Communication’, in R. P. Hawkins, J. M. Wiemann and S. Kagel, J. H. and Levin, D. (1986) ‘The Winner’s Curse and Pingree (eds) Sage Annual Review of Communication Research: Advancing Communication Science Vol. 16, Public Information in Common Value Auctions’, American Beverly Hills, CA: Sage, pp. 110–34. Economics Review 76: 894–920. Rafaeli, S. and Noy, A. (2002) ‘Online Auctions, Messaging, Kagel, J. H. and Levin, D. (eds) (2002) Bidding in Common- Communication and Social Facilitation: A Simulation and Value Auctions: A Survey of Experimental Research, Experimental Evidence’, European Journal of Information Princeton, NJ: Princeton University Press. Systems 11(3): 196–207. Kaplan, M. F. (1987) ‘The Influencing Process in Group Rafaeli, S., Raban, D., Ravid, G. and Noy, A. (2003) ‘Online Decision Making’, in C. Hendrick (ed.) Review of Simulations in Management Education about Information Personality and Social Psychology Vol. 8, Beverly Hills, CA: and its Uses’, in C. Wankel and R. Defillippi (eds) Sage, pp. 189–212. Educating Managers with Tomorrow’s Technologies A Volume Kaplan, M. F. and Miller, C. E. (1983) ‘Group Discussion and in: Research in Management Education and Development, Judgment’, in P. Paulus (ed.) Basic Group Processes, NY: Greenwich, CT: Information Age Publishing, pp. 53–80. Springer-Verlag, pp. 65–94. Reicher, S. D., Spears, R. and Postmes, T. (1995) ‘A Social Kiesler, S. and Sproull, L. (1992) ‘Group Decision Making and Identity Model of Deindividuation Phenomena’, in W. Communication Technology’, Organizational Behavior & Stroebe and M. Hewstone (eds) European Review of Social 52: 96–123. Human Decision Processes Psychology vol. 6, Chichester, UK: Wiley, pp. 161–98. Downloaded By: [Schmelich, Volker] At: 20:39 23 March 2010 Latane, B. (1981) ‘The Psychology of Social Impact’, Resnick, P., Zeckhauser, R., Swanson, J. and Lockwood, K. American Psychologist 36: 343–56. (2003) ‘The Value of Reputation on eBay: A Controlled Lombard, M. and Ditton, T. (1997) ‘At the Heart of it all: Experiment’, working paper, online at: http:// The Concept of Telepresence’, Journal of Computer- www.si.umich.edu/,presnick/papers/postcards/, Mediated Communication 3(2), online at: http:// accessed 14 October 2003. www.ascusc.org/jcmc/vol3/issue2/lombard.html Rice, R. E. (1993) ‘Media Appropriateness: Using Social [accessed 10 May 2004]. Presence Theory to Compare Traditional and New Lucking-Reiley, D., Bryan, D., Prasad, N. and Reevs, D. Organizational Media’, Human Communication Research (2000) ‘Pennies from eBay: The Determinants of Price in 19: 451–84. Online Auctions’, World Conference Econometric Society, Shogren, J. F., List, J. A. and Hayes, D. J. (2000) ‘Preference 2000 Seattle. Learning in Consecutive Experimental Auctions’, American Maidment, R. and Bronstein, R. H. (1973) Simulation Games, Journal of Agricultural Economics 82(4): 1016–21. Design and Implementation, Columbus, OH: Charles E. Short, J., Williams, E. and Christie, B. (1976) The Social Merrill Publishing Company. Psychology of Telecommunications, London: John Wiley. McAfee, R. P. and McMillan, J. (1987) ‘Auctions and Smith, C. W. (1989) Auctions: The Social Construction of Bidding’, Journal of Economic Literature 25(2): 699–738. Value, Berkeley, CA: University of California Press. Milgrom, P. (1989) ‘Auctions and Bidding: A Primer’, The Spears, R. and Lea, M. (1992) ‘Social Influence and the Influence Journal of Economic Perspectives 3(3): 3–22. of the ‘‘Social’’ in Computer-mediated Communication’, in Minar, N. and Donath, J. (1999) ‘Visualizing The Crowds At M. Lea (ed.) Contexts of Computer-mediated Communication, A Web Site’, Proceedings of CHI 99. London: Harvester-Wheat-sheaf, pp. 30–65. Electronic Markets Vol. 15 No 2 175

Sproull, L. and Kiesler, S. (1991) Connections: New Ways Of Varian, H. R. (2000) ‘Online Auctions and Bidding working in the Networked Organization, Cambridge, MA: Strategies’, Northeast Wisconsin’s Internet Monthly MIT Press. Published also in New York Times, 16 November Sproull, L., Subramani, M., Kiesler, S., Walker, J. H. and 2000. Waters, K. (1996) ‘When the Interface is a Face’, Human- Wilcox, R. T. (2000) ‘Experts and Amateurs: The Role of Computer Interaction 11: 97–124. Experience in Internet Auctions’, Marketing Letters 11(4): Sudweeks, F., McLaughlin, M. and Rafaeli, S. (eds) (1998) 363–74. Network and Netplay. Virtual Groups on the Internet, Wilkie, W. L. (1994) Consumer Behavior (3rd edn), New York: Menlo Park, Cambridge: AAAI Press. John Wiley & Sons, Inc. Senecal, S. and Nantel, J. (2001) ‘Online Interpersonal Wilson, R. (1992) ‘Strategic Analysis of Auctions’, in R. J. Influence: A Framework’, working paper, MIT eCommerce Aumann and S. Hart (eds) Handbook of Game Theory With Research Forum, online at: http://e-commerce.mit.edu/ Economic Applications Vol. 1, Amsterdam: Elsevier Science cgi-bin/viewpaper?id5117 [accessed 10 May 2004]. Publishers, pp. 227–80. Downloaded By: [Schmelich, Volker] At: 20:39 23 March 2010