臺大管理論叢 NTU Management Review VOL.28 NO.3

95 NTU Management Review Vol. 28 No. 3 Dec. 2018 Table 3 Measures of Recommendation Quality and System Interface Quality Construct Item Recommendation Quality (1). Does the system recommend the precise papers you are interested? (2). Do the recommended papers meet your interest? (3). Does the system provide sufficient recommendation information? (4). Is the recommendation service provided by this system useful? System Interface Quality (1). Is the operating procedure unambiguous? (2). Is the information clear? (3). Is this system user friendly? (4). Is this system easy to use? 4.2 Sample and Experimental Procedures A total of 100 volunteers were recruited via e-mail from two universities in Taiwan to participate in our experiment. Though all subjects had to select at least one time slot to be physically present for the lab experiment, they were allowed to use an online system to provide their demographic data and to answer the items related to computer self-efficacy. For the lab experiment, each subject was randomly assigned to one of the four treatments and was offered a gift as an incentive to complete the experimental task. All subjects followed the same procedures for each treatment. Since the users’ interests had to be identified before papers could be recommended, the experimental task was made up of three stages. In the initial stage (Stage 0), subjects were required to log into the system with their personal accounts in order to record a personal interest profile. Based on the login accounts, subjects were grouped into one of the four treatments. Next, all subjects rated the initial ten abstracts, using a Likert scale ranging from 1 to 7 (from dislike to like), and answered a review test to evaluate their English ability. In order to separate the experimental stages and to allow the subjects to relax, the subjects were asked to log out of the system and watch a five minute film. The purpose of watching the film was to keep the subjects in a positive mood, since mood could influence their performance during the experiment (Oaksford, Morris, Grainger, and Williams, 1996). Each subject had to rate ten papers at each stage, so it was necessary to avoid generating a negative mood and inducing user inertia. Thus, each subject was to watch one entertaining film at the end of stage 0 and stage 1. In the next stage (Stage 1), subjects were required to log into the system again. Each subject rated the abstracts of the ten papers recommended by each treatment, and indicated the perceived degree of recommendation quality, ranging from 1 (low quality) to 5 (high quality). After that, they logged out of the system and watched another five minute film. In