【摘要】 为揭示个性化推荐系统的用户决策"黑箱",本文基于决策理论和前人成果提出了研究假设,并基于情境实验的主观调查数据和行为数据,对推荐规模、用户感知和决策之间的关系以及不同决策风格用户在决策中的调节作用进行了实证研究。研究发现:推荐规模扩大,用户感知推荐的吸引力、感知选择难度明显提高,同时对感兴趣商品的回忆效果和搜索深度均显著下降,决策时间显著增加;感知吸引力、感知选择难度会影响用户的选择意向和决策努力,进而对决策质量产生影响;选择意向在用户感知与决策努力之间起部分中介作用;用户决策风格在感知吸引力与决策努力之间起调节作用。更多还原

【Abstract】 The user interface of personalized recommendation system is another major factor that inf luences the decision-making process of its user in additions to the recommendation algorithm. It is true that exposure to an excessive number of commodities which is common in traditional retail industry causes the overload of the customer’s choice. However, it is still unclear whether and how the number of items in the recommendation list affects the users’ perception and behavior.To reveal the black box of users’ decision-making in personalized recommendation system, we empirically studies the relationship among recommendation scale, users’ perception, and their decision behavior; and the moderating effect of users’ decision-making style, based on two scenario experiments of online-shopping. Finally, practical suggestions were put forward for the application and design optimization of the recommendation system.In experiment 1, long and short recommendation lists were randomly presented on E-Prime software. Behavioral data, including response time, exploration depth, etc., were obtained when 69 subjects effectively completed the experimental task and user perception survey. In experiment 2, 320 subjects were divided into two groups and completed the questionnaire survey of user perception and decision-making behavior respectively under short and long recommendation list scenes. The results show that:(1) There are significant differences in user perception under different recommendation scales. User’s perceived attractiveness and perceived choice difficulty for a long recommendation list both increase dramatically.(2) According to the statistics of subjective survey, the large recommendation scale has no significant effect on user decision-making, but the result based on the behavioral data shows that the recognition correct rate, response time and search depth have extremely significant differences under different recommendation scales. Especially, the quality of user’s decision-making under the short list is relatively better because the correct rate of recommend items recognition was 84.56%; and the search depth was deeper, indicating that users in short list are more willing to make efforts for choice;(3) Under different recommendation scales, the influence of perception variables on decision variables and the influences among decision variables are different. In particular, the user intention is negatively affected under the short list by relatively lower perceived difficulty of user choice, and the relationship between user intention and perceived choice difficulty is significantly regulated by the user’s decision-making style. At the same time, perceived choice difficulty affects user efforts through the intermediary role of user intention.更多还原