Spiekermann, Sarah: Online Information Search with Electronic Agents: Drivers, Impediments, and Privacy Issues


Chapter 1. Introduction

The dramatic growth of the Internet as an electronic shopping and commerce environment for both B2C and B2B transactions has led to a strong research interest into the effects of this new electronic medium on economic relationships.

Current revenues from European B2C retail markets, which form the research frame for this thesis, are estimated to account for around euro 19 billion. And this number is expected to grow quickly to euro 174 billion by 2005 [Nordan, 2000].

One major driver for this expected growth is the increasing deployment of automated tools, including electronic agents, that assist users in the buying process. While today‘s online shopping is mostly a ’user-driven‘ task that offers only limited interaction and confronts consumers with the tedious problem of information overload, electronic consumer agents are promising to deliver a whole new way of purchasing goods and services. Practitioners and academics alike expect this to ring in a ’second-generation‘ of electronic commerce [Pazgal, 1999; Vulcan, 1999]. In this scenario, many of the consumer‘s decision-making tasks are delegated to, or at least assisted by, virtual assistants. These have access to a myriad of information sources and are able to filter them according to user preferences [Alba et al., 1997]. Eventually, these agents may even negotiate purchase conditions on behalf of users [Maes et al., 1998; Preist, 1998].

Against the background of these anticipated developments, this thesis focuses on the deployment of agent technology from a marketing perspective. The goal in doing so is to study potential drivers and impediments for the further acceptance and use of electronic agents in consumer markets.


1.1 Electronic Consumer Agents in Marketing Research

The study of human-agent interaction, from a marketing perspective, was probably initiated by Alba et al. in 1997, who investigated the theoretical implications of agent assisted search for consumers, retailers and manufacturers. This group of academics argued that the informational advantages provided by electronic consumer agents would have the potential to reduce buyer search cost and optimize decision making, but also outlined some main criteria on which the growth of interactive home shopping with agents would depend (like reliability of information sources and access to a vast selection of products).

Based on these theoretical reflections empirical research was conducted to test some of the hypotheses made. Häubl and Trifts [2000], for example, showed how recommendation agents are able to effectively reduce consumers‘ search effort for product information, augment the quality of the consideration set as well as of the final purchase decision. Pederson [2000] presented similar work showing how consumer agents are able to optimize the information search part of the buying process and partially enhance consumer choice. Brynjolfsson and Smith [2000] investigated the effect of ’shopbot‘ use on price sensitivity and found that brands and retailer reputation have a significant effect to obtain price advantages.

All of these research projects studied agents with a view to their role as facilitators in information search. Yet, little attention has been paid to the fact that many different agent roles can be distinguished, offering different types of benefits to consumers and reaching beyond the support of information search.

One group of marketing academics who distinguished agent roles were West et al. [2000]. They showed that agents can act as tutors, clerks, advisors and bankers for consumers. While tutor-agents educate clients about the features available in a product category and help them uncover preferences, clerk-agents focus on assisting their clients in complex information search processes and product screening. Advisor agents may be called upon to express expert opinions on products and are able to provide tailored advice. Banker-agents are envisioned to negotiate purchases on consumers‘ behalf and facilitate the purchase of products and services.


Despite this existing distinction of agent roles in electronic commerce, there has been little research on design challenges or economic effects that these different roles entail. In contrast, academics tend to use the general term ’agent‘ or ’shopbot‘ when they actually refer to clerk or advisor agents. And as a result, it seems as if a more systematic exploration of the technology in its different facets has so far been widely ignored.<1>

In order to correspond to this lack of ’role-recognition‘ in consumer-agent marketing research, chapter 2 of this thesis starts out with a detailed analysis of West et al.‘s framework [2000] on agent roles, and proposes an extension for it, relating these roles to different purchase situations. This extension is then used to argue that agent acceptance is particularly challenging when it comes to the deployment of the technology in high-involvement purchase situations.

When agents are used to support high-involvement purchase decisions, one major challenge for the technology is to win consumers‘ trust [West et al., 2000; Urban et al., 1999]. An empirical study that has explicitly investigated this issue is the one presented by Urban et al. [1999]. The group of academics tested the acceptance of a trust-based advisor-agent<2>for the truck market and found that only half of those subjects who indicated to like buying online really preferred an agent-based site for product search. A clear preference was found among all subjects for Web sites that offer not only an agent system, but also manually accessible, “information intensive“ shopping sites.

Thus, there was a need detected among users to manually control at least parts of the information flow. Consumers who then expressed a preference for using an advisor-agent were those who were not very knowledgeable about vans, younger and more


frequent Internet users. They also had visited more van dealers in advance of a real-world shopping trip.

Urban et al.‘s research shows that agent advice is not always the most preferred solution for all consumers. In addition, it suggests that there are drivers and impediments such as product knowledge or demographics that influence the degree to which agents are accepted. Given this first empirical evidence, chapters 4 and 5 of this thesis explore such drivers and impediments in more detail. A number of factors derived from information search literature are investigated that are hypothesized to motivate or impede users‘ reliance on agents. The frame to do so is similar to Urban et al.‘s in that interactions between consumers and an advisor-agent are studied for a high-involvement purchase context. In a next step, one particular impediment for agent interaction and trust in the system is then studied in more detail: online consumers‘ privacy concerns.

Urban et al. [1998 cited in West et al., 2000], in fact, suggested that privacy is one major trust building cue when consumers interact with agents. Many household surveys indirectly support this view, reporting strong privacy concerns of online users [Pew Internet & American Life Project, 2000; Ackerman et al., 1999; Westin, 1996]. In many cases, these concerns even lead to false data provisions [Grimm et al., 2000; Sheehan and Hoy, 1999]. Consequently, a number of agent researchers have pointed at privacy concerns as a major challenge for agent acceptance [Shearin and Maes, 2000; West et al., 2000, Norman, 1994].

Acknowledging the significance of privacy concerns as an impediment to human-agent interaction, chapter 6 of this thesis focuses on the issue in more detail.

1.2 Thesis Structure

To study consumer interactions with agents, chapter 2 starts out with a general introduction outlining what electronic consumer agents actually are. A definition of the agent concept is included to avoid the widespread misconception of the term [Franklin and Graessner, 1996] and make clear what type of technology and application is referred to in the rest of the thesis. Then, potential agent roles in


electronic commerce are discussed and it is argued how these different roles gain relevance for different types of purchase situations. Most importantly, it is shown that agent acceptance by consumers is particularly challenging when these software tools are used to support high-involvement, targeted search.

Targeted search with consumer agents, and especially with dialogue-based systems, requires considerable effort on the part of the user [West et al., 2000]. In order to comprehend what drives and impedes consumers in high-involvement situations to use or avoid agents, an experiment has been conducted with over 200 subjects in the form of a ’real-world‘ online shopping trip. This experiment and the results obtained are presented in chapters 3 to 6 of the thesis. While chapter 3 gives a detailed overview of the experiment, chapters 4 to 6 report on the findings made. Chapter 4 presents a structural equation model. It tests potential drivers and impediments for consumers to rely on agent-based and/or manually controlled search forms in a high-involvement purchase situation. The results obtained in this analysis help to nail down some concrete factors that influence consumers‘ interaction readiness and reliance on agent technology.

Chapter 5 then looks in more detail into how consumer interactions differ when they shop for two different product categories online. Here, again, a focus is being put on the two main search forms available: agent-based versus manually controlled search. Finally, chapter 6 focuses on one particular and potential impediment to consumer interaction with agents: privacy concerns. The chapter contains the elaboration of a model that captures personal consumer information cost, a measure for the negative utility attached to the revelation of personal information to electronic agents. Based on this measure (and other variables), the degree of disclosure practiced by experimental participants during the shopping session is investigated. In a next step, the degree of disclosure is compared with subjects‘ proclaimed attitude towards online privacy.

Chapter 7 closes with major conclusions that can be drawn from the empirical work and suggestions for future research.



For example, no marketing research has been done on the effects of automatic recommender systems (tutor-agents) on decision making, an agent type amongst the widest ones used in electronic commerce today.


The concept of advisor agent will be presented below. It refers to agents that give tailored advice to consumers, mostly in ’high-involvement‘ purchase situations.

© Die inhaltliche Zusammenstellung und Aufmachung dieser Publikation sowie die elektronische Verarbeitung sind urheberrechtlich geschützt. Jede Verwertung, die nicht ausdrücklich vom Urheberrechtsgesetz zugelassen ist, bedarf der vorherigen Zustimmung. Das gilt insbesondere für die Vervielfältigung, die Bearbeitung und Einspeicherung und Verarbeitung in elektronische Systeme.

DiML DTD Version 2.0
Zertifizierter Dokumentenserver
der Humboldt-Universität zu Berlin
HTML - Version erstellt am:
Mon Dec 23 13:23:15 2002