2015 Volume 17 Issue 3 Pages 307-316
Although the number of applications of product recommendation virtual agents (PRVAs) for online shopping is rapidly increasing, few studies on effective PRVAs have been done. Hence, we investigated which types of PRVAs are effective through experiments with participants. First, we focused on variation in the appearance of PRVAs and prepared six PRVAs having various appearances from human-like to text. Then, we prepared six products that customers might want in any situation, and conduct withinparticipant experiments in which the participants reported their impressions on the agents and evaluated their buyer motivation for all combinations of the PRVAs and the products. By seeing the averages and performing statistic tests on the results, we found that the human-like PRVA was the most effective among the various PRVAs including video of a real human, and a robot-like PRVA. Furthermore, by performing a factor analysis on the experimental results, we also found the two important factors of the PRVAs, familiarity and intelligence, to influence the recommendation effects.
