網路書評中的確認偏誤：論初始印象對於不同涉入程度讀者信任書評之影響

Abstract

近年來，網路評論大量成長，吸引了業界與學術界的注意，進而探討消費者在網路評論環境中的決策行為。相關研究指出人會選擇相信與自己原先立場一致的資訊，此行為傾向被稱為「確認偏誤」。此外，亦有研究發現消費者對自己喜好產品的涉入，將會影響其後續的購買決策。故本研究招募對神祕小說有所涉入的讀者，以探討神祕小說的讀者閱讀網路書評時，是否存在確認偏誤，且此確認偏誤的形成，是否會受到讀者對神祕小說的涉入程度所調節。 本研究採實驗法，受試者經由實驗網站，先觀看7本神秘小說之書籍資訊，對該小說形成一初始印象，並根據閱讀書籍的興趣進行排序，就其最喜歡的2本與最不喜歡的2本小說，繼續閱讀相應之正、負面書評，最終蒐集受試者兩階段評分與排序、對書評的歸因及其認為書評可信度等資料。完整的實驗歷時約50分鐘。 將62位受試者的填答資料進行分析，結果顯示讀者在閱讀神祕小說的網路書評時，確實存在確認偏誤，換言之，當讀者閱讀與其初始印象立場相同的書評時，其容易認為該書評能夠反映書的品質，認同評論者是為了告知他人書的品質好壞，而撰寫該書評，同時，也認為該書評可信度較高。進一步比較確認偏誤的強度，結果顯示相較於形成負面初始印象，接續閱讀負面書評的情境，若讀者一開始對小說形成正面印象，接續閱讀正面的書評時，其所形成的確認偏誤強度較強。此外，涉入程度為加強讀者確認偏誤傾向的調節變項，表示讀者對神祕小說的涉入程度越高，其閱讀書評後形成的確認偏誤強度越強。 綜觀前述研究結果，可知讀者在閱讀神秘小說網路書評時，不但存在確認偏誤，且這樣的行為傾向，會隨著受試者對神祕小說的涉入程度而被加強。未來的推薦系統應將讀者的確認偏誤納入考量，嘗試推薦與讀者原先立場不一致的書評，以改進讀者先入為主的傾向，為讀者帶來更為全面的閱讀經驗。

In recent years, the prevalence of online consumer reviews and their influence on consumers’ judgment have attracted great interests from both marketers and academia. Previous research indicated that consumers tends to search for evidences which confirm an existing belief, rather than for those which disconfirm it, a widely observed phenomenon called “confirmation bias”. In addition, it was also found that consumers’ involvement with the product could be an important moderator in the persuasive effect of online reviews. This study set out to examine the confirmation bias in the context of reading online book reviews of mystery fiction, also the moderating role of individual’s involvement with mystery fiction. A total of 62 participants took part in a 2 (initial impression: positive vs. negative) × 2 (review valence: positive vs. negative) within-subjects experiment. Through a mock book review site, participants were presented with the basic bibliographic information about the seven mystery titles, including the cover, author, book series, publisher, language, format and the summary. After the first impression was formed, they were asked to rate each books, and then ranked the seven titles, from the most to the least favored. The top two and the two lowest ranked titles were chosen for the second stage of the experiment. Next, they were asked to read the positive and negative reviews about the titles which were chosen. After finishing the reading, they were asked to answer the questions for each review. At last, they would rate the chosen titles again. It was shown that confirmation bias did exist in the context of reading online book reviews of mystery fiction. Readers were more likely to infer that the comments supporting their existing position were motivated by a desire to inform other buyers about the book, and believed that these comments were more likely to be based on the reviewer’s true experiences. Therefore, readers found these comments much more persuasive as compared to the ones disconfirming their existing position. With regard to the intensity of confirmation bias, when the positive impression about the book was formed, reading the positive reviews subsequently, the confirmation bias would be much more stronger as compared to situations in which negative impression has been formed, and then reading the negative ones. Furthermore, readers' involvement with mystery fiction moderated the effect of the interaction between initial impression and review valence on the intensity of confirmation bias. As the degree of involvement increased, the intensity of confirmation bias toward the mystery fiction were also enhanced. Overall, the existence of confirmation bias was found in the context of reading online book reviews of mystery fiction. Moreover, the intensity of confirmation bias were enhanced due to the moderation effect of involvement. Accordingly, readers’ confirmation bias should be considered in the future design of online decision aids.