맞춤화 정보 환경에서 뉴스 추천 알고리즘에 대한 이용자 이해도와 인식의 중요성 : 관점 일치 뉴스 노출, 뉴스 신뢰, 뉴스 추구 행위와의 관계를 중심으로 = User Understanding and Perceptions of News Recommendation Algorithms : Relationships with Attitude-Consistent News Exposure, News Trust, and News-Seeking Behavior

다국어 초록 (Multilingual Abstract)

The use of news recommendation algorithms is becoming more widespread, raising concerns that their implementation could lead to a “filter bubble”, a biased and closed information environment tailored to individual preferences. Previous literature presents mixed findings about whether or not recommendation algorithms contribute to constructing filter bubbles. While a body of literature has confirmed that an algorithmic environment can encourage attitude polarization and a homogenous information environment, some research has shown contrasting evidence that the algorithmic environment can increase incidental exposure to diverse information or cross-cutting viewpoints. Each body of literature offers valid points about recommendation algorithms and their filter bubble possibilities, however, little work has considered users’ agency in using algorithms, and how individuals’ use of algorithms can result in making differences in the information diversity level. Drawing on the possibility that users’ understanding and perceptions of algorithms can influence the use of algorithms, this study explored whether users’ information environment is distinctively formed by users’ understanding and perceptions of the algorithms. We first examined the relationship between the level of algorithm understanding and three algorithm perceptions that are frequently discussed, namely, convenience, privacy risk, and bias perceptions. Convenience perception is defined as individuals’ belief that the utilization of a recommendation algorithm service would enable users to effortlessly locate the most recent information aligning with their desires and needs, ultimately contributing to the enhancement of their goals. Privacy risk perception concerns users‘ recognition that the recommendation algorithm may compromise personal information during the collection and analysis of users‘ private data. Bias perception pertains to the awareness that the employment of news recommendation algorithms may hinder users’ exposure to a diverse array of social issues, potentially leading to the consumption of information that predominantly emphasizes one-sided perspectives or political viewpoints. We also examined how these algorithm perceptions are respectively associated with the following three predictors of a closed information environment: (1) exposure to attitude-consistent news, (2) news trust (trust in algorithm-recommended news), and (3) news-seeking behavior. The analysis of 1,169 online survey responses revealed that first, individuals with a higher understanding of news recommendation algorithms perceived a greater level of convenience and bias in algorithms, but algorithm understanding had no significant relationship with privacy risk perception. Second, convenience perception had a positive relationship with recommendation algorithm-accepting behavior, and this, in turn, was associated with an increase in exposure to attitude-consistent news. Convenience perception also showed positive relationships with news trust and news-seeking behavior respectively. Meanwhile, privacy risk and bias perceptions - while having no significant relationships with news trust - had negative relationships with attitude-consistent news exposure through a lowered algorithm-accepting behavior. Bias perception showed a positive relationship with news-seeking behavior while privacy risk perception was negatively linked to news-seeking behavior. We confirmed the importance of users’ algorithm understanding and perceptions in achieving information diversity in the customized environment. The findings present meaningful insights into the necessary enhancements of users’ perceptions concerning news recommendation algorithms, and the corresponding corrections needed to foster a more diverse information environment. Implications for algorithm literacy education were also discussed.

국문 초록 (Abstract)

뉴스 추천 알고리즘이 보편화됨에 따라 추천 알고리즘의 이용이 이용자의 개별적 선호사항과 관심사에 치우친, 폐쇄적 정보 환경을 초래할 수 있다는 우려가 제기되었다. 본 연구는 이용자의 정보 환경이 알고리즘 맞춤화의 이용 여부에 따라 획일적으로 결정되는 것이 아닌, 알고리즘에 대한 이용자의 이해 정도와 인식에 따라 다르게 형성될 가능성을 탐색하였다. 이를 위해 (1) 추천 알고리즘에 대한 인지된 이해도와 알고리즘 인식(편리성, 프라이버시 침해, 편향성)간의 관계를 탐색하고, (2) 알고리즘 인식들과 이용자 정보 환경의 폐쇄성을 예측할 수 있는 세 가지 변인—관점일치 뉴스 노출, 뉴스 신뢰, 뉴스 추구 행위—간의 관계를 분석하였다. 1,169명의 온라인 설문 응답을 분석한 결과, 첫째, 뉴스 추천 알고리즘에 대한 이해도가 높은 개인일수록 편리성과 편향성을 높게 인식하였으나, 프라이버시 침해 인식과는 유의미한 관계가 없었다. 둘째, 편리성 인식은 추천 알고리즘 수용 행위와 정적인 관계를 보였고, 이를 통해 관점 일치 뉴스에 대한 우연적 노출에 정적 관계를 보였다. 또한 편리성 인식은 알고리즘 추천 뉴스를 신뢰하고, 뉴스를 능동적으로 추구하는 행위와도 정적인 관계를 보였다. 한편 프라이버시 침해 인식과 편향성 인식은 알고리즘 추천 뉴스 신뢰와는 유의미한 관계가 없었으나 알고리즘 수용 행위와의 부적 관계를 통해 관점 일치 뉴스 노출의 감소와 유의미한 관계를 보였다. 또한 편향성 인식이 뉴스 추구 행위와 정적 관계에 있었던 것에 반해 프라이버시 침해 인식은 뉴스 추구 행위와 부적인 관계를 보였다. 맞춤화 환경에서 의 정보 다양성을 위해 알고리즘 이해와 인식이 중요함을 확인하였으며, 리터러시 교육과 관련된 함의를 논의하였다.