Week 8
In the eighth week, we began exploring the relationship between algorithms and platform content.
Through reading, I learnt that Sumpter(2018) used principal component analysis(PCA) and k-means clustering to categorize his friends into three distinct groups based on their social media activity: those focus on their private life, those centered on their work and work-related lifestyle and those engaged with broader societal events, demonstrating how algorithms can reveal patterns in social behavior.
He found that the most significant differences among his friends were based on two dimensions: public VS. personal and culture VS. workplace. This classification information is the foundation for algorithms applied in various commercial activities on social media platforms. Sumpter points out that while algorithms can aid in disseminating and understanding diverse social behaviors, they may also oversimplify complex interpersonal relationships, leading to misperceptions of the understanding provided by such classifications. Furthermore, Sumpter cites the 'black box' concept to explain the opacity in algorithmic processes, which refers to the internal operation of the algorithm being invisible or incomprehensible to users. Users are unaware of how their data is processed and utilized within the system, nor do they understand how various decisions are made. This lack of transparency often causes users to feel uneasy and powerless, which can influence their usage behavior.
Sumpter's discovery made me reflect on how algorithmic categorization can influence individual behavior on social media. For instance, if users are aware of how algorithms categorize content, they might embellish or alter their posts to align with the audience they are facing and to gain better data, thus impacting the authenticity of the content.
This reminded me of one of China's largest online sharing communities, Xiaohongshu, which uses powerful algorithms to categorize content creators into different groups and tag their posts. The platform then pushes this content to target users and provides creators with visual data analysis of each post's performance. What I find unique about this platform's algorithmic mechanism is that creators who consistently post within the same category tend to gain more traffic and visibility, while diversifying their content reduces the likelihood of being recommended to others. The system classifies creators into clear categories based on their content, such as fashion, cooking or painting. Once a creator reaches 1,000 followers, their category will be displayed on their account profile (this information is not displayed if the number of followers is less than 1,000).
This is actually an algorithm and incentive mechanism that requires authors to try to post only the content of the categorised area to which the account belongs, in order to facilitate the platform's content management and push. Such an algorithmic mechanism can, on the one hand, improve the efficiency and speed of content acquisition, ensuring that users receive a steady creative sources of content in areas they are interested in. but on the other hand, can lead to content homogenization and the creation of 'filter bubbles', that is, the algorithm has mastered the user's contact with the media content of the active filtering rights. The content outside of users' interests is excluded from recommendations, limiting their exposure to diverse information and, to some extent, hindering the spread of diverse content.