Constrained Signaling in Auction Design

We consider the problem of an auctioneer who faces the task of selling a good (drawn from a known distribution) to a set of buyers, when the auctioneer does not have the capacity to describe to the buyers the exact identity of the good that he is selling. Instead, he must come up with a constrained signalling scheme: a (non injective) mapping from goods to signals, that satisfies the constraints of his setting. For example, the auctioneer may be able to communicate only a bounded length message for each good, or he might be legally constrained in how he can advertise the item being sold. Each candidate signaling scheme induces an incomplete-information game among the buyers, and the goal of the auctioneer is to choose the signaling scheme and accompanying auction format that optimizes welfare. In this paper, we use techniques from submodular function maximization and no-regret learning to give algorithms for computing constrained signaling schemes for a variety of constrained signaling problems.

💡 Research Summary

The paper “Constrained Signaling in Auction Design” tackles a fundamental problem that arises when a seller cannot fully disclose the identity of the item being sold. Instead of the classical setting—where the seller announces the exact item and runs a second‑price auction to achieve optimal social welfare—the seller must choose a signaling scheme: a (possibly randomized) mapping from items to a limited set of signals. The signals are constrained by practical considerations such as bounded communication bandwidth, legal advertising restrictions, or certification requirements. The seller’s goal is to select both a signaling scheme and an auction format (the authors show that, without loss of generality, the auction can be taken to be a second‑price auction) so as to maximize expected welfare (or, via a known relationship, expected revenue).

Model and Definitions

• A (possibly infinite) item space Ω with a known prior distribution p.
• n bidders, each with a private valuation function v_i: Ω → ℝ_+. The valuation profile (v_1,…,v_n) is drawn from a common prior D; the paper distinguishes between known valuations (deterministic profile) and unknown valuations (drawn from D).
• A finite signal set S and a family F ⊆ S^Ω of admissible signaling maps, encoding the external constraints (e.g., a bound on the number of distinct signals, or a bipartite feasibility graph).
• A signaling scheme is a distribution x over F; a deterministic scheme corresponds to a single map f ∈ F.

Given a scheme x, each signal s induces a conditional distribution over items, and bidders’ optimal strategy in a second‑price auction is to bid their expected value conditioned on s. The expected welfare of scheme x for a fixed valuation profile v is

 welfare(x, v) = Σ_{s∈S} x(s)·max_i E_{ω∼p}