短视频算法“破茧”报告：近八成用户认可推荐内容多样性

Core Insights - The report indicates that nearly 80% of short video users perceive algorithm recommendations as diverse, contrasting with common concerns about "information cocoons" [2][5] - Users exhibit strong agency, with over 75% believing their actions can "tame" algorithms, and more than 60% showing willingness to actively "break free" from information silos [5][8] User Perception of Algorithms - 77.8% of users find algorithm recommendations to include "unexpected but interesting content" [2] - 73.9% of users feel that algorithms help them discover "new areas of interest" [2] User Agency and Interaction - Over 75% of respondents believe they can influence algorithm recommendations through their actions [5] - More than 65% of users are adept at using features like "not interested" to adjust content [5] Experimental Findings - The report conducted experiments on platforms like Douyin, Bilibili, Xiaohongshu, revealing that Douyin maintains high diversity in recommendations while responding positively to user exploration [5][7] - Bilibili's recommendation diversity is highly driven by user behavior, while Xiaohongshu is characterized by efficient convergence and precision [5] Academic Discussions - Experts acknowledged the complexity of algorithms and their societal impacts, emphasizing the need for further academic exploration [7][8] - Discussions highlighted the importance of user autonomy in algorithm design and the potential for algorithms to inadvertently reinforce narrow interests [8][9] Recommendations for Algorithm Development - Scholars suggest enhancing algorithm transparency and user education to foster a healthier human-machine collaboration ecosystem [10][11] - The need for a balanced approach in algorithm design that respects user autonomy while promoting information diversity was emphasized [9][10]