Communication under Ambiguity

We study communication in ambiguous environments. Agents communicate individual decisions sequentially. Based on the signal that an agent receives, she revises her private information. When signals are ambiguous, we show that agents may agree to disagree forever. Although the learning process converges, leading to common knowledge of the individual decisions, the decisions differ no matter how long the communication lasts. Such persistent disagreements are precluded in Bayesian frameworks endowed with a common prior. When communication is private and nobody is excluded from it, we show that ambiguity-free communication is a necessary and sufficient condition for a consensus to emerge. However, when communication is public, ambiguity-free communication is only sufficient for reaching a consensus. We apply our model to economic forecasting, providing an alternative account for persistently disagreeing forecasters. Finally, we outline an "advisor's dilemma" and explain why deliberation in democracy may never end.