Social selection refers to the tendency of users with similar characteristics to choose to interact with each other [2] 。 In some literature, it is also called homogeneity. As early as 1928, Bott [3] We have recognized the existence of social choice through the behavior of kindergarten children in games, and there are abundant literatures in classical sociological research on this effect. In 2001, McPherson published Birds of a feather: Homogeneity in social networks [4] In this paper, a comprehensive overview is given.
Social selection of social networks is one of the most significant rules in group life [5] It exists in countless social environments. For example, you choose A as your friend in social networks, probably because you and A both like the same type of movies or music. This same interest preference makes you interact frequently. Social choice also exists in a variety of individual attributes. From the perspective of manifestation, social choice can be divided into observable social choice and potential social choice. As the name implies, observable social choices include users' dominant characteristics, such as the same age, gender, education, geographical location, religious belief, etc. The potential social choice is an unobservable feature, such as the same interests, values, etc.
Social Choice and Social Impact
In addition to social choice, another significant rule in group activities on social networks is social influence [6] 。 In real life, we often rely on other people's evaluations of movies, books and restaurants to determine whether these activities are worth their time or money. Through interaction with other users, we will be affected to varying degrees by other users. Many terms have been used to describe this widely recognized role, such as social interaction, peer influence, social contagion, consistency, and imitation. Different terms may have slight differences, but they all describe the dependency between user behavior and user interaction. Generally, social impact is used to describe the phenomenon that people change their own thoughts, feelings, attitudes and behaviors due to interactions with other people or groups.
It should be noted that user behavior in social networks is usually the result of the joint action of social influences, social choices, and external factors (i.e. the stimulation of information from outside the social network) [7] 。
3、 The distinction between social choice and social impact
It is difficult to distinguish the effects of social choice, social influence and exogenous factors. In 1993, Charles F. Manski wrote "Identification of Endogenous Social Effects: Reflection on Problems" [8] In this paper, it has been proved that it is theoretically infeasible to distinguish these three factors in a static situation. Further research also found that there was an obvious feedback effect between choice and influence mechanism, which made it difficult to completely distinguish social influence from social choice.
Researchers conducted large-scale experiments in real social networks to explore the role of social influence and social choice in the network. For example, the study found that Facebook and Flickr have a stronger role in social selection, Youtube, ScienceNet, Epinics and other networks have a stronger role in social influence, and both influence mechanisms exist in the cooperation network of DBLP papers [9] 。 The research on the connection mechanism of new nodes in the paper cooperation network found that the mechanisms such as discipline direction homogeneity (38%), social impact (36%), and institutional homogeneity (26%) have different explanatory power [10] 。
To solve this problem, Liye Ma [11] And his team had an interesting discussion. The team obtained all the call records of a large city within three months provided by a large Asian communication operator. The data includes more than 3.7 million users and more than 300 million call records. Each call log contains the phone numbers of the dialer and the receiver, the call duration and the call duration. More importantly, the data also includes the purchase of RBTs, that is, when and what kind of RBTs the user purchased. RBT is a kind of cheap personalized RBT that requires monthly payment. If user A purchases RBT, when user B dials user A's phone, user B can hear the RBT purchased by A before A answers the call, and user B may also have purchase intention after hearing it. Therefore, every call to RBT users can be regarded as an interaction between users about RBT products. The unique nature of RBT gives us the opportunity to discover the potential driving factors of users' purchase behavior in social networks, especially its role in measuring social impact. To this end, Ma Liye et al. made the following assumptions:
If the driving factor of RBT purchase behavior is social influence, it will be shown that a user A first buys RBTs, then his friend B calls user A after user A buys RBTs, and end user B also buys similar products;
If the common external stimulus is the driving factor, it will show that two users have purchased the same RBT in the same time period without calling each other.
If social choice is the driving factor, users will buy the same product without calling each other, and the purchase behavior does not need to be carried out at the same time.
Through further experiments, Ma Liye and others found that RBT There is a strong social influence on purchase time and purchase decision, and a strong homogeneity effect is found in the RBT type preference and sensitivity to influence. 
