Browse Teach Curate

Reinforcement Learning for Adaptive Respondent-Driven Sampling

Casts respondent-driven sampling of a hidden population as a sequential decision problem and uses reinforcement learning to tailor recruitment incentives (coupon number, value, framing) over time, maximizing cumulative utility instead of holding a fixed a-priori incentive structure. Because the induced sampling is adaptive and subjects are dependent through a latent social network, the paper also supplies valid post-study inference and proves asymptotic regret bounds for the design.

Demerzel Text

Reinforcement Learning for Respondent-Driven Sampling Justin Weltz ∗ Angela Yoon † Yichi Zhang ‡

Respondent-driven sampling (RDS) is a chain-referral method for surveying hidden or hard-to-reach populations: seed participants receive a fixed number of coupons and are incentivized to recruit thei…