The Shift From Social Graphs to Socio-interest Graphs within Social Media Algorithms
With the rise of TikTok in 2020, a gradual shift from social graphs to socio-interest graphs has emerged within the AI algorithms used by mainstream social media companies to engage users.
In the past, social media platforms such as Facebook and Instagram have used social graphs to digitally connect their users to the network of people they know and their mutual friends. For example, AI algorithms that used social graphs would make suggestions for profiles for users to follow and would even influence what type of advertisements people saw on their page. These were people to people connections, and would rely on the triadic closure principle: If person a had a strong connection with person b and with person c, then person b and c were likely to also have some type of a connection. However, the fact that people may know and follow each other on social media does not indicate that they have similar interests in what they want to see online.
In order to engage more users, apps such as TikTok have heavily relied on the use of socio-interest graphs. When TikTok was first launched, users were immediately addicted to the tiny little videos that would keep them entertained for hours. This is because TikTok’s algorithms determine a user’s potential interest based on their current interest by using a social-interest graph, where interest in content is equally as important as people to people connections. If a user has a strong interest in a particular type of content, then they are likely to be interested in watching videos by the same content creator (interest to person connection), or are likely to be interested in watching videos on closely related topics (interest to interest connection). As users scroll and continue to have interesting content pop up, they are incentivized to stay on the app. Therefore, companies such as Meta are integrating socio-interest graphs in their algorithms with the use of new features that require a shorter attention span from the user. (i.e. introducing the concept of reels to Instagram).
As we learned in class, both types of graphs rely on the use of the triadic closure principle to predict what types of interests are connected and which people are connected to each other. Strong and weak ties are also used to determine a user’s changing interests. For example, if a user’s interest in a subject is beginning to die down, a weak connection can be made between their interests and the AI algorithm can tweak the type of content shown online.
Source: https://www.archetype.co/apac/blog/changing-algorithms-from-social-graph-to-socio-interest-graph/