選擇性暴露悖論：既存態度、眼動與態度極化

Abstract

選擇性暴露在政治傳播中已有許多討論，但過去研究結果並非一致。本研究使用眼動追蹤，測量不同黨派支持者於正反政治意見並陳頁面之注意力分配模式，以及測量政治文章選擇性暴露與態度極化之關連。

本實驗在2016年5月至6月間，測量美國人民於美國2016總統大選前，對於特定爭議政策之態度以及測量選擇性暴露行為是否造成態度極化。實驗分成兩階段，第一階段使用問卷，針對特定議題，測量既存態度、受測者黨派偏好，及其餘態度控制變項，如政治參與興趣、媒體使用行為；階段二使用眼動追蹤測量選擇性暴露行為，包括受測者於正反態度並陳頁面之注意力分配、文章頁面閱讀時間，並且再次測量受測者對於特定議題之態度。研究結果顯示，受測者在文章選擇及閱讀階段，仍有明顯選擇性暴露，偏好與既存態度一致的政治資訊；然而在正反意見並陳的搜尋結果頁，眼動追蹤顯示注意力並非僅集中於與既存態度一致之政治資訊，在最低薪資議題中，受測者注意力甚至顯著集中於與既存態度相反之標題引言。此外，使用中介調節模式分析既存態度、黨派、及眼動行為對於態度極化之結果顯示，於槍枝管制議題中，黨派與既存態度，可同時顯著預測第一次暴露正反意見並陳頁面之注意力，且眼動行為可顯著預測態度極化。中介調節模式結果顯示，即便受測者於在文章選擇與閱讀時偏好立場一致資訊，此行為並無法對於態度極化造成影響。

In the present article we used eye-tracking technology to assess the different attention allocation patterns among partisans and their effect on attitude polarization. The concept of selective exposure was derived from cognitive dissonance, which was first proposed by Festinger (1962). It is known that in order to decrease the unpleasant feeling caused by dissonance, people would seek congenial information instead of uncongenial ones and therefore cause the selective exposure behavior. Prior selective studies mainly focused on the confirmation bias and used reading time or the number of messages accessed by voters to measure selective exposure.

A 2-session study examined the relations among prior attitude, partisanship, selective exposure, and attitude polarization. In the session 1 pre-survey, participants were asked to report their attitudes toward some controversial policies and policy importance, attitude certainty, news using habits, political interest, and demographics. In the session 2 eye tracking experiment, participants were invited to the lab. Participants' eye movements were recorded unobtrusively during the second session, the in-lab study.

Results showed that attitude-consistent articles about abortion and minimum wage were read for a significantly longer time. However, when participants were simultaneously presented with attitude-consistent and attitude-discrepant messages confirmation bias decreased, which corresponded to prior studies about presentation styles. We further found the effect differed by issues. Interaction effect was found among partisanship, prior attitude, and selective exposure on the issues of gun control: Republicans and Independents who opposed gun control performed longer selective exposure to attitude-consistent messages which resulted in attitude polarization. Those who supported gun control had shorter selective exposure to attitude-consistent messages, resulting in attitude attenuation. The results of our eye movement analysis also indicate that there was no significant polarization for the Democrats in our sample.