Social Products and Overambitious Seeding
This article talks about product developers and how they often use the ideas we’ve talked about in class such as network effects and network diffusion to get people to adopt their products. Product developers count on people influencing each other to adopt the product and aim to find the right number and right group to introduce the topic to in order for long-term adoption to be maximized. This article introduces this concept in the real world in which it costs money to “seed,” or to introduce the product to people via giving it to them or marketing to them. The argument here is that it is actually not optimal to introduce the product to everyone, if it’s a social product, because people will experience it in suboptimal contexts, and people will either increase or decrease usage based on their experiences. Therefore, gradual seeding can actually be better for long-term adoption when it comes to influence maximization.
This study’s simplified model of social-product usage incorporates people’s experiences in their decision to use the product or not, and therefore takes into account the need for social support in decision making, the change in usage based on experience, and the possibility that people will stop using the product after a negative experience. The model was simulated on real-world networks and on synthetic networks, to test how long-term activity was maximized – through giving everyone the product immediately or through the gradual seeding process. The results showed that a balance needed to be struck between the cost of giving someone access to the product and risking a negative experience because of the initial low usage rates of people’s connections and the cost of not giving someone access early, as the social support of an additional user may be lost. The study, in short, demonstrated that the people product developers needed to target for initial product adoption are the ones who can sustain long-term activity of the product by themselves, without the social support of all of their friends also using the product.
This directly relates to what we’ve been looking at in class with network effects and network diffusion. Another connection is, in the real-world networks with which this model was tested, clustering was present and taken into account. There is also a specific end goal with this diffusion. The SIR model was also referenced, in that another study (Sela et al.) showed a similar phenomenon while investigating seeding budgets. One expansion of our class topics was the dynamics in synthetic networks (single clusters) in which initial usage initially dropped and then rebounded into expansion of usage, indicating two processes happening simultaneously – unsatisfying usage and resulting decrease and satisfying experiences and resulting increasing usage. Another extension, one not actually tested but mentioned in this study, is the complexity in how people adopt. This model suggests a “hard-coded gradual access expansion rule, triggered by activity oft their friends in the social network” (11). Another way people adopt, more seen in real-world scenarios, is the access expanding through invitation to friends via social network. All of these are ideas we’ve either modeled in class, or things we could extend our knowledge to, using the basics of diffusion, adoption, clustering, etc.
Link: https://research.fb.com/wp-content/uploads/2018/11/The-Costs-of-Overambitious-Seeding-of-Social-Products.pdf?