Countervailing Curation Strategic Disclosure and the Design of Attention

Opposing advocates may be unable to fabricate evidence but can choose which true observations to present. We study how a court, editor, or platform should divide potential exposure between an advocate who prefers a higher decision and one who prefers a lower decision. Each advocate controls a separate evidence pool, exposure is committed before the state and evidence are known, and a selection-naive receiver averages displayed observations. Under baseline linear preferences and common knowledge of the realized pools, all Nash equilibria in every finite pool induce the same action and common disclosure bar: the upward advocate reveals observations above it and the downward advocate those below it. Attention therefore changes selection as well as weight. Under a common evidence law, more exposure makes an advocate speak less often and more extremely while moving the decision in its preferred direction. In large pools, the unique state-contingent exposure share that reproduces the complete-evidence decision balances the advocates' directional tail moments, not their observed speech. Because this share generally depends on the unknown state, we characterize the optimal ex ante compromise and solve a primitive two-state economy with a unique second-best policy. For finite pools, we derive the exact risk-minimizing adjustment, separating selection-induced cutoff bias, the receiver's fixed benchmark, and sampling variance. The attention-to-cutoff feedback survives partial weighting of empty slots and vanishes at full imputation.