使用者判斷開放性知識分享平台資訊品質之研究─以Yahoo!奇摩知識+為例

Abstract

開放性知識分享平台是現今網路使用者倚重的網路資源，有大量使用者在此類平台上搜尋與分享資訊。然而，此種平台缺乏資訊品質控管的機制，任何網路使用者都可以在平台中分享資訊，使資訊品質良莠不齊，因此，探討使用者如何判斷資訊品質成為重要議題。Yahoo!奇摩知識+是我國極受歡迎的開放性知識分享平台，本研究因此選擇以該平台為研究對象，探究網路使用者判斷資訊品質之情形。 本研究以12名國立臺灣大學的學生為受試者，生農、理工、人文社會背景各有4位，博士生1名、碩士生3名、大學生8名。受試者必須執行四種類型的模擬任務，四類任務分別為「解決特定問題之交談型問題(類型I)」、「解決特定問題之資訊型問題(類型II)」、「非解決特定問題之資訊型問題(類型III)」、「非解決特定問題之交談型問題(類型IV)」，每類任務都選出一項任務執行，12位受試者皆必須執行四項任務。受試者先閱讀模擬任務之內容，並根據任務內容從研究者預先選出的相關回答中選擇三至六筆感到滿意的回答。任務執行過程以Morae軟體全程錄製，研究者則透過電腦連線以另一台電腦觀察受試者執行任務之過程，任務結束後立即以訪談法釐清受試者的行為與想法。 研究結果發現，受試者在判斷資訊品質時會採取預測性判斷和評估性判斷兩種策略，且此兩種判斷策略會交替性進行(交替性模式)、階段性進行(階段性模式)，亦有時而交替、時而階段進行之情形(混合性模式)。受試者判斷資訊品質時感知到的線索可分為五大類：資訊形式類、社會因素類、詮釋資料類、資訊內容類、時間因素類。線索詮釋結果分為正面詮釋和負面詮釋，正面詮釋導致受試者點擊或選擇資訊，負面詮釋則使得受試者不願點擊或選擇資訊。此外，相同線索可能得到相反的詮釋結果，線索詮釋極具主觀性。資訊品質判準除了過去研究已提出的內容類、認知類、效用類、資訊來源類、外部因素類與社會情感類，本研究另發現社會評價類和時間因素類，判準共計分為八大類型。最後，任務類型會影響線索感知次數、資訊品質判準使用次數、判斷策略使用以及資訊品質判斷共識，對於線索詮釋的影響則極小，32種線索中只有1種線索因為任務類型不同而產生詮釋差異。研究最後試提出資訊品質判斷模型。

Nowadays, open knowledge sharing system is on of the Internet resources which users rely on. Lots of users search and share information in this kind of system. However, the state that without information quality control mechanism in such system allows any user to share information, leads to information quality discrepancy. Therefore, investigating how users judging information quality turns into an issue. Yahoo! Knowledge is one kind of such system, which is very popular in Taiwan. Thus, this study selected Yahoo! Knowledge to be the study object. This study recruited 12 subjects from National Taiwan University. Subjects were from 3 disciplines: 4 from biology and agriculture, 4 from science and engineering, and 4 from liberal arts and social science. 1 subject was PhD student, 3 subjects were graduate students, and 8 subjects were undergraduate students. All subjects were asked to perform four types of simulated tasks, inclusive of “problem solving and conversational question (type I)”, “problem solving and informational question (type II)”, “non-problem solving and informational question (type III)”, and “non-problem solving and conversational question (type IV)”. For four types, subjects had to choose one task which he/she was more interested in. After choosing task, subjects had to read the text of task and select 3 to 6 answers which he/she was most satisfied with according to the content of task. All the relevant answers were selected previously by researcher. The process of task performing was recorded by Morae software. Via PC connecting, researcher observed the subject by another computer. After the subject finished 4 tasks, researcher interviewed with him/her immediately to clarify his/her thoughts and actions. Results show that users make predictive judgments and evaluative judgments when making judgments of information quality. Besides, subjects make predictive judgments and evaluative judgments by iterative way, two-stage way, or mixed way (combining the iterative way with two stage way). The study also finds that the clues noticed by subjects when making judgments can be classified into 5 categories, including “information form”, “social aspect”, “metadata”, “information content”, and “time aspect”. The results of clue interpretation are divided into “positive interpretation” and “negative interpretation”; the former leads to the willingness to click or select information and the later leads to rejection to do so. Furthermore, the same clue may be interpreted into the opposite results, which means that clue interpretation is highly subjective. Previous study has discovered 6 classes of information quality criteria, including “content”, “cognitive”, “utility”, “information source”, “extrinsic”, “socio-emotional”. This study not only finds these classes again, but also discovers new classes: “social assessment” and “time aspect”. In sum, information quality criteria have 8 classes. Another significant finding is that task type affects on the frequency of clue noticing, the frequency of criterion using, dependence on predictive judgment or evaluative judgment, and consensus about information quality. However, task type has little impacts on clue interpretation; in 32 kinds of clues, there is only one kind which has interpretation divergence due to task type. To summarize the research results, this study presents the model of judgments of information quality.