|Spiekermann, Sarah: Online Information Search with Electronic Agents: Drivers, Impediments, and Privacy Issues |
This thesis has investigated how consumers search for high-involvement product information online. The conclusions of this investigation are based on a significantly rich data set, generated by a large-scale, real-world experiment. Over 270 subjects were observed in their online behavior and their dealings with an electronic agent while shopping for winter jackets and compact cameras.
A number of valuable insights have been gained regarding what drives consumers to interact with agents, and what impedes them, in their search for online product information during high-involvement purchase interactions.
One major finding is that agents do not play the same role in, and are not equally important for, online information search in different product categories. This has become clear from a theoretical perspective, regarding agent roles in different purchase tasks, as well as upon the empirical investigations made. Communication with agent Luci was comparatively less important for jacket shoppers than it was for camera shoppers, because jacket shoppers displayed a high need for product visualization and wished to have an overview of the available product spectrum.
The search for cameras, in contrast, was more ’fact-driven‘ and, as a result, the relative importance of the agent was higher. As cameras and jackets were perceived as search and experience goods, respectively, by experimental participants, it would be interesting to investigate to what extend this product classification cannot be used to explain consumer-agent interaction more systematically. To what extend are electronic agents capable to at all transmit experience qualities of goods? More insights into how product classes call for or impede agent use would certainly be invaluable for the online retail industry, especially during investment deliberations regarding front-end technology.
127A major aspect of this thesis was the separation of online information search into two constructs, manually controlled and agent-assisted search. This separation allowed to observe that, at a significant level, consumers prefer to manually control the search process the more risk they perceive. Consequently, electronic advisor agents are relatively less relied upon in the information search process. A similar direction of behavior was found for product knowledge. In line with older studies [Urban et al., 1999] our model suggests that the more product knowledge a consumer perceives the less he interacts with an agent for information search purposes. However, this result must be regarded with caution as the system employed in the current experiment did not offer a high-level expert-exchange for more knowledgeable customers.
Finally, a potentially major impediment for agent interaction has been investigated in detail: privacy concerns. The results obtained are interesting in that, against expectations, privacy concerns to not impede disclosure. In contrast, if systems offer appropriate returns in the form of personalized recommendations online users seem to be ready to reveal even highly personal information. And there is no incentive for them to lie as this behavior would adversely affect the benefit of search (the recommendation quality). The finding suggests that there is a lot of room for online marketers to communicate with their clients through dialogue-based electronic agents. If marketers used the spectrum of legitimate personal questions that are related to the product selection process more systematically, they could gain valuable insight into their customers‘ decision making process as well as on decisive product attributes. However, unfavorable privacy settings do seem to induce a feeling of discomfort among some users which then leads to less interaction time. Marketers therefore have to provide for a comforting privacy environment in order to make their customers feel good about the interaction.
Summing up, evidence has been generated in this thesis that users have a strong desire to control the information search process. The only significant driver of agent assisted search that could be supported by the structural equation model was purchase involvement. Thus, the more people had an immediate need for the product, the more they performed a search using the agent. However, this behavior was true in the same way for manual search. Thus, the vision of ’agents that buy and sell‘ for consumers [Maes et al., 1999] or that take over the entire purchase process for
128consumers without recurring back to them [Borking et al., 1999; Pazgal, 1999] must clearly be questioned against the background of this thesis‘ findings. At the same time, the often cited challenge of agents to overcome privacy concerns appears to be of rather marginal importance as consumers enjoy ’talking about themselves‘ online and benefit from personalized recommendations. These findings, which are in many respects surprising, suggest that it is easy to have misconceptions about how consumers deal with electronic advisor agents. And, given this, a whole new field of research opens up: management, reliance and trust upon relationships between humans and artificial entities.
© 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||
der Humboldt-Universität zu Berlin
|HTML - Version erstellt am:|
Mon Dec 23 13:23:15 2002