# Dynamic Information Design: A Simple Problem on Optimal Sequential Information Disclosure

arXiv: https://arxiv.org/abs/2003.07965 · HTML: https://mlsanity.com/paper/2003.07965/

## Key facts

- Cited by 4 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 17 Mar 2020.

## Built on by (most cited first, 4 of 4)

- [Sequential Information Design: Markov Persuasion Process and Its Efficient Reinforcement Learning](https://mlsanity.com/paper/2202.10678/) (2022)
- [Human-in-the-loop Learning for Dynamic Congestion Games](https://mlsanity.com/paper/2404.15599/) (2024)
- [Off-Policy Evaluation for Sequential Persuasion Process with Unobserved Confounding](https://mlsanity.com/paper/2504.01211/) (2025)
- [Semantic Rate Distortion and Posterior Design: Compute Constraints, Multimodality, and Strategic Inference](https://mlsanity.com/paper/2602.03949/) (2026)

## Abstract (arXiv)

We study a dynamic information design problem in a finite-horizon setting consisting of two strategic and long-term optimizing agents, namely a principal (he) and a detector (she). The principal observes the evolution of a Markov chain that has two states, one "good" and one "bad" absorbing state, and has to decide how to sequentially disclose information to the detector. The detector's only information consists of the messages she receives from the principal. The detector's objective is to detect as accurately as possible the time of the jump from the good to the bad state. The principal's objective is to delay the detector as much as possible from detective the jump to the bad state. For this setting, we determine the optimal strategies of the principal and the detector. The detector's optimal strategy is described by time-varying thresholds on her posterior belief of the good state. We prove that it is optimal for the principal to give no information to the detector before a time threshold, run a mixed strategy to confuse the detector at the threshold time, and reveal the true state afterwards. We present an algorithm that determines both the optimal time threshold and the optimal mixed strategy that could be employed by the principal. We show, through numerical experiments, that this optimal sequential mechanism significantly outperforms any other information disclosure strategy presented in literature.