An insurance upsell at the end of a booking flow. A coupon nobody hunted for. A subscription nobody remembered to cancel. Those are the examples Patrick Collison reaches for in a set of notes on agents in the economy that Tyler Cowen posted late Saturday, and each one is a small tax on inattention. Collison predicts personal agents will stop paying it: they read the fine print, find the coupon, cancel the trial, and in doing so act as "a kind of structural subsidy for product quality." When every buyer effectively spends ten hours researching a purchase, the better mousetrap finally beats the better-distributed one.
The mechanism is right. Collison also names the one assumption it rests on, and leaves it open:
Much of what I'm writing hinges on the assumption that personal agents will be on the consumer's side. I suspect that that will be the winning strategy, but whether it's true, and what it means to be true, involves a lot of other questions.Marginal Revolution
What it means for an agent to be on your side has an old, unglamorous answer. You're the one paying for it. A placement fee buys the seller a vote in the answer.
We have run this experiment once. Search was the first machine that did the consumer's research, and in 1998 its two most famous builders said where it would drift. Appendix A of the paper that introduced Google is titled "Advertising and Mixed Motives":
We expect that advertising funded search engines will be inherently biased towards the advertisers and away from the needs of the consumers.Brin and Page, Stanford (1998)
Then they built the largest advertising-funded search engine in history. Nobody had to turn venal for that to happen. The incentive did the work, slowly, and the result is the demand-routing economy Collison now expects agents to unwind: ranking algorithms to placate, placement fees to pay, a first click worth fighting over.
The next sentence of the appendix matters more for agents: "Since it is very difficult even for experts to evaluate search engines, search engine bias is particularly insidious." A results page at least showed ten links, with the ad labeled at the top, and you could scroll past it. An agent hands back one recommendation, or skips the recommendation and books the flight or renews the policy. Brin and Page worried about a search engine adding "a small factor" for friendly companies. An agent can add that factor and leave nothing on screen to compare it against.
Collison's own list holds the irony. People who forget to cancel subscriptions, he notes, subsidize those who don't. The agent will be a subscription, or it will be free.
The strongest objection is competition. Switching agents is cheap, a tilted agent should lose to an honest one, and you can point a second agent at the first to check its work. All of that holds today: the serious agents are sold by subscription, the market is young, and Collison's quality subsidy is real for as long as that lasts. But the audit takes the one thing the premise says consumers lack. The essay starts from a buyer who is "inattentive, underinformed, irrational" in front of an insurance upsell. That buyer is not going to compare loyalty functions across agents. Delegation doesn't cure inattention. It moves it up a level, into one decision that's harder to see than the hundred it replaced. Cowen's addendum makes the point from the seller's side: the law may have no simple way to stop agents colluding on price, because the collusion can be illegible even to the merchants running them. Illegibility doesn't care whose agent is doing the tilting.
Delegation doesn't cure inattention. It moves it up a level, into one decision that's harder to see than the hundred it replaced.
For anyone underwriting an agent company, the revenue line is the alignment spec. The insurance upsell was priced for a buyer who wasn't paying attention. An agent the seller pays for is the same upsell, sold to a buyer who has stopped paying attention on purpose.