University:
California State University, Northridge
Course:
ECON 101 | Economics for Everyday Life
Academic year:

2023
Views:

298

Pages:

3
Author:

Jithin J I

BAYESIAN PERSUASION IN SEQUENTIAL TRIALS Introduction  Information Design: Focuses on how informed agents (senders) persuade uninformed agents (receivers) to take specific actions through information disclosure.  Canonical Model: Kamenica-Gentzkow model, where senders manipulate receivers' beliefs through signals.  Contrast with Mechanism Design: In information design, the sender influences actions indirectly through information; in mechanism design, the designer uses transfers or other means. Motivation  Real-world problems with constrained signaling:  Multi-phase trials experiments. where interim outcomes determine subsequent  Some experiments are exogenously determined.  Examples: Drug trials, start-up acquisitions, job interviews. Motivating Example: Acquiring Funds from a Venture Capital Firm   Start-up seeking funds:  Multiple rounds of negotiation and evaluation.  Demonstrations vs. assessments.  Start-up designs demonstrations, VC firm determines assessments. Sender's challenge: Design demonstrations to maximize funding chances. Modeling Constraints  Constraints on signaling space:  Sender's signaling space reduced due to determined experiments.  Imposes restrictions on sender's signaling, different from classical Bayesian persuasion. Contributions 1. First study of sender's signaling schemes with exogenously determined experiments in arbitrary positions within multi-phase Bayesian persuasion. 2. Solution for optimal signaling in two-phase trials. 3. Dynamic programming algorithm for optimal signaling in general multi-phase trials. 4. Analysis of constraints on the sender via determined experiments, comparing with classical Bayesian persuasion. Problem Formulation  Two agents: Sender (Alice) and Receiver (Bob).  Binary states of the world: Θ = {θ1, θ2}, with known prior belief p.  Receiver's action: Φ ∈ {φ1, φ2}, representing a prediction of the true state.  Receiver's utility: ur(φi, θj) = 1{i=j}.  Sender's utility: us(φ1, θi) = 1, us(φ2, θi) = 0. Multi-Phase Trials  A sequence of experiments where outcomes influence subsequent phases.  Experiment outcomes affect interim belief and future experiments.  Constraints on sender's signaling due to determined experiments. Two-Phase Trials  Focus on two-phase trials with binary-outcome experiments.  Sender's challenge: experiments. Design signaling strategy considering determined General Multi-Phase Trials  Extension to multi-phase trials with binary-outcome experiments.  Structural insights from two-phase trials used for dynamic programming.  Analysis of constraints on senders signaling.