Retrospective Search: Exploration and Ambition on Uncharted Terrain
summary
Models search outcomes as a realized Brownian path, lets the agent choose costly variance (search speed) and a stopping time, and fully characterizes the optimum: the search speed is U-shaped (fast near breakthroughs and near termination) and the stopping rule is a linear drawdown stopping boundary (stop when current falls a constant D_R below the historical maximum). Closed‑form solutions exist for zero discounting (payoff, duration, speed), D_R falls with patience, and in the principal–agent extension speed is independent of commission while the stopping gap responds to wages and shares.
AI-written summary · about
chapters (8)
- 0:00Introduction to search
- 2:32Search features and mechanics
- 6:24Stopping rules and frameworks
- 9:01Formalizing the search model
- 14:17Stopping boundary analysis
- 20:36Optimal search speed
- 31:22Utility and performance
- 36:53Contracting and conclusion
related talks
32mRevisiting the Exploration-Exploitation Trade-Off in Bandit Models
18mTo Explore or To Exploit: Decoding Human Decision Making with R and Python
29mPure Exploration Problems
39mBandits and Agents: How to Incentivize Exploration?
36mBeating the Curse of Dimensionality in High-Dimensional Optimal Stopping
01h 17mOnline Reinforcement Learning and Regret
27mEconometrics for Learning Agents
30mBuying Data Over Time: Approximately Optimal Strategies for Dynamic Data-Driven Decisions