Designing Personalization Technologies to Form Rather Than to Inform: Similarity and Identification in Recommendation Agents
Speaker: Dr. Sherrie Komiak, Faculty of Business Administration
Abstract: In e-commerce, consumers are faced with a choice of the recommendation agents. A recommendation agent (RA) recommends products to a customer based on the customers personal preferences. Existing literature on RAs seems to focus on designing RAs as a persuasion tool, which likely assumes that an RA knows better about products than a customer does thus the RA can and should persuade a customer with its better knowledge. As an alternative, this paper assumes that an RA may know products better but the customer knows himself/herself better, and that a customer may want to dominate the shopping decision making rather than being persuaded by the RA. Therefore we propose a theory-grounded model of designing an RA as an extension of a customer himself/herself. Our point is to design an RA to form a customer rather than just to inform the customer. To be specific, we propose that an RA can model the customer himself/herself in addition to just model the customers product requirements. When an RA models a customer, it aims at achieving deep similarity with the customer. In contrast, the currently-widely-used RA type tries to match a customers expressed product attributes which only target on achieving surface similarity with the customer.
This paper uses a novel construct, users identification with an RA, as a construct indicating the human-computer connectivity (i.e. customer-RA connectivity). We examine the effects of different similarities (i.e. different RA types) on a customers cognitive and emotional identifications with the RA, which in turn affect customers adoption of the RAs.
This paper conducts an experiment and a survey to test our proposed research model. The results show that RAs with surface similarity design will increase a customers cognitive identifications which then increase the customers intention to adopt the RA, while RAs with deep similarity design will increase a customers both cognitive and emotional identifications which then increase his/her intention to adopt the RA. This paper contributes by exploring theoretical and practical considerations of user-centered design.