Sato & Shimizu paper redefines information value in social learning

Economists Hiroto Sato and Konan Shimizu have published a theoretical paper titled 'Value of Information in Social Learning' on arXiv (2503.05015) that extends Blackwell's foundational 1953 comparison of information structures to a sequential social learning framework. In this setting, agents make decisions one after another, each using both a private signal and the observed actions of previous agents. The authors introduce a new binary relation: one information structure is 'more socially valuable' than another if it yields higher expected payoffs for every agent, regardless of their preferences or which equilibrium is realized.

The paper shows that this social value ordering is strictly stronger than the classic Blackwell order—meaning an information structure can be Blackwell-superior but still fail to improve welfare for all agents in a social learning context. The authors derive a necessary and sufficient condition for the relation, and also propose a simpler sufficient condition that is easier to verify in practice. They further analyze comparisons in terms of long-run payoffs (as the number of agents grows large), limit welfare, and canonical binary environments where signals and states are binary.

This work has implications for designing information systems in settings where decisions are sequential and socially observed—such as financial markets, online reviews, or organizational hierarchies. It challenges the assumption that more informative signals always benefit everyone, and provides formal tools to evaluate information structures with distributional welfare consequences.