MISES: Minimal Information Sufficiency for Effective Service

Joss Armstrong's new paper, MISES: Minimal Information Sufficiency for Effective Service, introduces a category-based coordination mechanism that allocates resources using only a declared service category for each user, without observing individual demand types. The paper establishes three key results: a tight two-sided bound on the welfare gap in terms of within-category allocation variance, a bound on misreporting gain that holds without assumptions on agent strategy, and a proof that aggregate outcome metrics dominate per-agent metrics for service-level detection under homogeneity. The mechanism was validated on a synthetic population of 50,000 demand vectors and five weeks of production performance-management data from four anonymized operator networks (28,249 cells).

The research reveals a fundamental trade-off for system designers: welfare improves with finer categories, but detection accuracy worsens. This creates a feasibility band [Kmin, Kmax] within which the designer must choose a number of categories K as a value judgment. Armstrong claims that any protocol achieving a welfare gap ≤ ε* and missed-detection rate ≤ β* requires at least Hlb(ε*, β*) bits of category entropy. The paper's practical implication is that collecting per-agent data beyond the category label adds noise to detection problems without reducing the welfare gap, suggesting that minimal data collection can be optimal for service allocation in networks.