공유행위는 일종의 커뮤니케이션 행위이며, 정보원의 사회적 영향력 및 여론 발생과 밀접하게 관련 있다. 본 연구는 트위터의 메시지가 어떻게 공유되고 확산하는지를 이용자의 공유행위를 중심으로 살펴보았다. 구체적으로, 언어 규범이 이용자의 공유행위에 영향을 미치는지 살펴보고자 언어학적 강도 변인과 힘없는 언어 변인을 도입하여 메시지의 공유된 횟수를 분석하였다. 또한, 선행연구 및 정보원과 이용자의 팔로워 관계 여부를 토대로 공유행위를 공유 여부와 공유 빈도로 구분하여 이들의 차이를 분석하였다. 이를 위해 제1차 남북정상회담 동안 작성되거나 공유된 트위터 메시지를 수집하였으며, 적극적으로 공론장에 참여한 정보원과 그의 메시지를 분석하였다. 그리고 분석자료의 다층적 특성과 0과잉 현상을 고려하여 다층 허들모형을 분석모형으로 선정하였다. 분석 결과, 공유 여부와 공유 빈도는 서로 다른 기제가 있음을 확인하였다. 언어학적 강도는 공유 빈도에 부정적으로 작용하였으며, 힘없는 언어는 정보원 수준 변인과 상호작용하여 공유 빈도에 긍정적인 영향을 미쳤다. 본 연구는 언어 규범이 이용자의 공유행위에 영향을 미치는 요인임을 발견하였으며 이를 언어기대이론(Language Expectancy Theory)을 통해 논의하였다.

Sharing in social media is an act of communication, reflecting the social influence of the sources being shared and contributing to the formation of public opinion. This study explored the underlying mechanism of sharing on Twitter, focusing on the characteristics of languages used in Tweets. Linguistic characteristics have been rarely examined in the previous research as a potential factor that may influence sharing. According to the Language Expectancy Theory (LET), based on socio-cultural standards, people have certain expectations for norms of languages used by sources and the violation of these expectations affects the sources’ credibility, which in turn can exert significant influence on sharing. Thus, we investigated the effects of linguistic intensity and linguistic powerlessness and, their interactions with the source-level variables on users’ sharing behavior. Furthermore, we conceptually and analytically distinguished between the possibility (whether or not to be shared) and the frequency (to what extent shared) of a Tweet being shared. For these analyses, Tweets posted or shared during the first inter-Korean summit were collected, and among them, Twitter accounts that actively participated in the discussions of the summit and their Tweets were included in our multilevel hurdle model (which considers both the multilevel and the zero-inflation characteristics of the data and allows us to analyze binomial (possibility of being shared) and zero-truncated negative binomial (frequency of being shared) models together. Our findings suggest that the mechanisms between the possibility and the frequency of being shared are different. The former was significantly affected by the message-level variables only, whereas the latter was significantly influenced by both the source-level and message-level variables. This indicates that users mainly considered message content when they decided to share a Tweet that had never been shared previously; but considered not only the content of messages but also who the sources were when they decided to share a Tweet that had been shared already. We also found that linguistic variables were statistically significant predictors of the frequency of being shared. Linguistic intensity had a negative effect on the frequency of being shared, and the interaction between linguistic powerlessness and verified account (a proxy measure for source credibility) had a positive effect on the frequency of being shared. Considering the context that the data used in this study was collected from Twitter and addressed the first inter-Korea summit, Tweets that convey linguistically intense expressions might have prevented users from sharing them. However, when Twitter accounts were regarded as having high source credibility (being verified by Twitter), linguistically powerless language appearing in their Tweets might have caused a positive expectancy violation and encouraged users’ sharing. This indicates that, contrary to previous findings, powerless language might have been perceived as positive by users, showing consistency with language norms identified in the Tweets frequently shared about the inter-Korean summit. Overall, these findings suggest that language norms are important factors influencing users’ sharing behavior and have implications with the LET.

1. 서론
2. 이론적 배경
3. 연구 방법
4. 연구 결과
5. 논의 및 결론
References
국문초록