互动最大化

We investigate the management of information provision to maximize user engagement. A principal sequentially reveals signals to an agent who has a limited amount of information processing capacity and can choose to exit at any time. We identify a “dilution” strategy—sending rare but highly informative signals—that maximizes user engagement. The platform's engagement metric shapes the direction and magnitude of biases in provided information relative to a user‐optimal benchmark. Even without intertemporal commitment, the platform replicates full‐commitment revenue by inducing the user's belief to remain “as uncertain as” the prior until the rare, decisive signal arrives and induces stopping. We apply our results to two contexts: an ad‐supported internet media platform and a teacher attempting to engage test‐motivated students.