Selling Experiments
summary
Model frames information sales as Blackwell experiments, seller maximizes revenue subject to incentive and participation constraints by offering a menu of experiments priced upfront; each experiment needs at most as many signals as actions and every optimal menu includes a fully informative option while other items are locally non-dispersed (some signal rows are pure). With binary states and continuum priors the revenue-optimal menu is extremely coarse — at most two experiments — and screening exploits how different priors value specific error tradeoffs.
AI-written summary · about