This paper characterizes extreme points of the set of incentive-compatible mechanisms for screening problems with linear utility. Extreme points are exhaustive mechanisms, meaning their menus cannot be scaled and translated to make additional feasibility constraints binding. In problems with one-dimensional types, extreme points admit a tractable description with a tight upper bound on their menu size. In problems with multi-dimensional types, every exhaustive mechanism can be transformed into an extreme point by applying an arbitrarily small perturbation. For mechanisms with a finite menu, this perturbation displaces the menu items into general position. Generic exhaustive mechanisms are extreme points with an uncountable menu. Similar results hold in applications to delegation, veto bargaining, and monopoly problems, where we consider mechanisms that are unique maximizers for specific classes of objective functionals. The proofs involve a novel connection between menus of extreme points and indecomposable convex bodies, first studied by Gale (1954).

We consider a multi-sender cheap talk model, where the receiver faces uncertainty over whether senders have aligned or state-independent preferences. This uncertainty generates a trade-off between giving sufficient weight to the most informed aligned senders and minimizing the influence of the unaligned. We show that preference uncertainty diminishes the benefits from specialization, i.e., senders receiving signals with more dispersed accuracy. When preference uncertainty becomes large, it negates them entirely, causing qualified majority voting to become the optimal form of communication. Our results demonstrate how political polarization endangers the ability of society to reap the benefits of specialization in knowledge.