運用社會網絡分析法探討讀者之閱讀偏好代表性—以aNobii網路書櫃為例

Abstract

並非所有顧客看過、消費過的品項皆能代表其偏好，產品代表偏好的程度也不一。對此，本研究提出「閱讀偏好代表性」的概念，藉由援引認知心理學的「範例觀點」，探究是否有少數書籍作品特別能代表讀者的閱讀偏好，為讀者閱讀範疇中具有「偏好代表性」之作品。 本研究的主要目的為衡量個別書籍代表讀者個人閱讀偏好的程度，將讀者收藏在aNobii網路書櫃中的書籍視為其閱讀興趣檔（reading profile），以任兩本書籍被aNobii全體使用者共同擁有的次數，計算任兩本書籍之間的相似度，建立書籍的相似性網絡。並利用社會網絡分析的中心性/核心指標，計算個別書籍作品之中心性/核心度，再與讀者針對所選作品的自評偏好代表性、作品類型知識與涉入進行統計分析，以反映讀者的閱讀偏好。 研究結果顯示，書籍之中心性或核心度與受試者自評之書籍偏好代表性、作品類型知識與涉入呈現顯著正相關，即受試者所選作品中，中心性/核心度越高的作品，其閱讀偏好代表性、作品類型知識與涉入的程度也就越高，表示以社會網絡分析指標判斷具有讀者閱讀代表性之作品的作法具有可行性。 除了討論以中心性/核心度討論具有閱讀偏好代表性的作品外，本研究的另一目標為探討讀者偏好多樣性的調節作用，也就是偏好多樣性較高的讀者，其書櫃相似性網絡可能呈現多個分群或核心，而難以判斷具有閱讀偏好代表性之作品。因此本研究假設閱讀偏好多樣性會影響以書籍的網絡中心性/核心度判斷讀者偏好代表性的效果。結果顯示，偏好多樣性越高的書櫃，其書籍中心性/核心度與讀者的閱讀偏好代表性的相關性較低，多未達顯著顯準；反之，偏好多樣性越低之書櫃的相關程度較高，且達到顯著水準，表示偏好多樣性對於書籍中心性/核心度與偏好代表性之間的關係確實具有調節作用。

Not every item that customers browse or buy can represent their desired preferences. An additional problem is that the degree of preference that each item represents can be different. By using the “exemplar view” theory of cognitive psychology, this study proposes using the concept of “representativeness” to determine whether a few books can represent one’s reading preference. This study’s main aim is to measure the degree to which a book in a user’s online bookshelf is able to represent their preferences by collecting data from aNobii online bookshelves. This data can be used to build a user’s reading profile. A reading profile can be used to generate an item-to-item network by calculating the similarity of each book-to-book pair in a user’s bookshelf. By applying social network analytical (SNA) metrics such as “centrality” and “coreness” to the network one can determine a book’s “representativeness.” Results show that the degree of centrality or coreness of books were significantly correlated to the user’s self-assessed “representativeness.” There was also significant correlation between the social network analytical metrics and the user’s self-assessed “genre knowledge” and “involvement” for books on their bookshelf. It means that it is feasible to use social network analytical metrics to determine how representative a book is of a user’s reading preference. The other objective of this study is to explore the moderated effect of a user’s preference diversity. It is assumed that a user’s preference diversity would affect the correlations between the centrality of books and the user’s self-assessed representativeness. Results show that the correlation between social network analytical metrics and a book’s representativeness is weaker in those who have highly diverse preferences than those who have less diverse preferences.