Balanced Social Learning

Andrew Koh; Ricky Li

Home/Archive/arXiv

arXiv · 2610.25814

Balanced Social Learning

Andrew Koh·Ricky Li

View original source ↗Read source document ↗Export metadata ↓

Abstract

When is social learning efficient if agents have heterogeneous preferences over a set of actions? When myopic agents generate information only about the action they choose, efficiency requires balance such that all high-value actions are discovered. We characterize the value of policies that control the flow of past information to future agents. Under full transparency of past information, the probability of efficient outcomes is approximately one if and only if the prior and preferences are balanced and the log-likelihood ratio distribution satisfies a set of intuitive moment conditions. Noisy bad news is a canonical example. By contrast, approximate efficiency is obtainable on the set of arbitrary adaptive information policies if and only if the prior alone is balanced. Moreover, it suffices to use adaptive evidence policies which permanently and fully disclose subsets of past information. When the prior is imbalanced and approximate efficiency is unattainable, we characterize the value of adaptive information and show that it is uniformly approximated by the value of adaptive evidence as the impact of each individual signal on beliefs vanishes.

Explore related subjects

econ.TH

Keep this discovery

Explore connections, maps & timelines

BibTeX RIS

Andrew Koh, Ricky Li. 2026-10-19. Balanced Social Learning. https://arxiv.org/abs/2610.25814

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related papers

arXiv2026

Collective decisions under uncertainty: efficiency, ex-ante fairness, and normalization

This paper studies preference aggregation under uncertainty in the multi-profile framework and characterizes a new class of aggregation rules that can address classical concerns about Harsanyi's (1995) utilitarian rules. Our aggregation rules, which we call \textit{relative fair aggregation rules}, are grounded in three key ideas: utilitarianism, egalitarianism, and the 0--1 normalization of individual utilities. These rules are parameterized by a set of weight vectors over individuals and evaluate each ambiguous alternative by taking the minimum weighted sum of 0--1 normalized utility levels over the weight set. For the characterization, we propose two novel axioms, \textit{weak preference for mixing} and \textit{restricted certainty independence}, with normative justifications. Additional results clarify how these axioms capture the utilitarian and egalitarian attitudes of the rules.

econ.TH↗

arXiv2026

Theoretical Foundations of Sequential Dynamic Programming

Many economic problems have several controls per period, whose joint choice is subject to the curse of dimensionality. Solving them in stages is computationally cheaper, but no general result says when staging is exact. I develop a separability theory, discrete choices and non-concave objectives included: staging is exact if utility, transitions, and constraints depend on earlier decisions only through a state passed forward. For each ordering the smallest such state exists and is unique up to relabeling. Any state that keeps staging exact for all preferences over later decisions and the future must carry everything the smallest one carries for the transition. For affine primitives, the smallest state consists of the linear combinations, such as cash-on-hand, on which later outcomes depend. Its dimensions pick the order whose largest stage is smallest. In a six-choice household model, each stage chooses one variable on a three-dimensional state, the joint problem all six.

econ.TH↗

arXiv2026

Random Disclosure Costs: A Unified Theory of Disclosure

We propose a unified framework for a class of evidence games in which a sender, seeking to increase posterior expectations (e.g., a market price), chooses whether to reveal or withhold her information. Our model nests several classic models, including type-dependent disclosure costs or benefits, uncertain information endowment, and the market for lemons. Frictions are captured as random disclosure costs that are privately known to the sender and noisily inferred by the receiver. We derive a simple characterization of the equilibrium and show that it features increasing disclosure for lower frictions or greater fundamental uncertainty. Further, with high uncertainty or low frictions, greater uncertainty about disclosure costs decreases the likelihood of disclosure. Positive costs do not necessarily rule out unraveling to full disclosure, especially when outcomes have heavy tails or when disclosure costs are not constant. Finally, we demonstrate that the model can be estimated with a simple probit, suggesting that cost uncertainty is descriptive of real-world behavior.

econ.TH↗