After speaking to Justin Weltz about his life as a graduate student, he spoke about RDS, which stands for respondent-driven sampling. Basically, it’s a form of gathering data from participants by having an initial participant bring in new participants. that creates an array of nodes and connections. Often these connections are very important, and they require inference to complete, as not all true connections can be seen through RDS.
An example of RDS can be seen below.
RDS also has a problem with bias, as similar people often group together. Justin Weltz also spoke about using reinforcement learning to optimize RDS, allowing for better recruiting and slimmer budgets.