Skip to content

Screening Menu or Self-Selection

Mechanism-design method — instantiates Private Information Asymmetry Governance

Offers a deliberately shaped menu whose best choice differs by hidden type, so a party reveals a materially private fact simply by which option it picks — no interrogation required.

A Screening Menu or Self-Selection governs an asymmetry by getting the informed party to reveal a private fact through a choice instead of a statement. Rather than asking a question that invites a self-serving answer, it designs a set of options whose terms — price, delay, effort, restriction — are arranged so that a party's best option differs by type. Each party, picking what serves it best, reveals which type it is. Its distinctive move among its siblings is that the informed party classifies itself: no fact is demanded, disclosed, or attested; the private trait is inferred from behavior under an incentive structure built precisely to make honest choice the profitable one. The design task, and the whole risk, is keeping that structure incentive-compatible so imitation never pays.

Example

An airline cannot ask "are you a deadline-driven business traveler who will pay almost anything, or a flexible leisure traveler hunting a bargain?" and expect a useful answer — yet that private fact governs how much each will pay. So it builds fare fences instead of asking. A cheap fare requires booking weeks ahead and staying over a Saturday night; a costly fare carries neither restriction and allows free changes. Business travelers, who value flexibility and can't predict their trips, self-select the expensive fare; leisure travelers accept the restrictions for the low price. The choice reveals the type the airline could never observe. The airline must keep the fence hard to game — business travelers dodging the Saturday-stay — and accept that a budget-strapped business flyer forced onto the costly fare, or a leisure traveler who games the cheap one, is a misclassification it chose to tolerate.

How it works

  • Name the hidden type the choice will reveal. Decide which private distinction the menu must expose — willingness to pay, patience, risk, commitment — and design toward that separation.
  • Build options the types rank differently. Each option pairs a benefit with a cost the two types value unequally, so their preferences diverge.
  • Enforce the incentive-compatibility gap. Space the terms so each type strictly prefers its intended option and gains nothing by imitating the other — this is the entire game.
  • Read the choice, and price the error. The option selected is the classification; set in advance how much misclassification is tolerable and which way to lean.

Tuning parameters

  • Separation gap — how far apart the options' terms sit. Wider guarantees a clean sort but pushes worse terms onto everyone; too narrow and the types pool.
  • Fence dimension — whether the trade-off runs on time, flexibility, effort, or bundling. It must be a dimension the types genuinely rank differently, or the menu sorts on nothing.
  • Number of tiers — two options or a graded ladder; more tiers extract finer information but add choice friction and design fragility.
  • Error tolerance — how much false-positive and false-negative misclassification is acceptable, and which error is worse to make.
  • Default option — what a non-chooser receives, since a careless default silently pools the passive majority.

When it helps, and when it misleads

Its strength is that it elicits a private fact at almost no measurement cost and feels fair because participation is opt-in: it works exactly where asking is useless, intrusive, or illegal, turning private information into a public choice. It is the natural governance move when you may not interrogate but may design the offer.

Its central failure mode is a menu that is not incentive-compatible: if a type can do better by choosing the "wrong" option, the separation silently collapses into pooling — everyone chooses alike — and the operator keeps trusting a sort that no longer exists.[^ic] It is also frequently regressive, sorting on ability to bear a cost rather than the intended trait, so the option that "reveals low risk" is really the one only the well-resourced can shoulder. The classic misuse is shaping the menu to funnel everyone toward the house-favorable option and then calling the result "self-selection." The discipline that guards against this is to verify each type's best response really is its intended option, monitor that choices still predict the fact, and hold an explicit misclassification policy rather than pretending the sort is clean.

How it implements the components

  • asymmetry_type_classifier — the option a party chooses classifies its hidden type; the menu is a classifier driven by self-selection rather than by inspection.
  • screening_signal_integrity_monitor — it watches that choices still separate the types and have not been gamed into a pooling collapse that reveals nothing.
  • false_positive_false_negative_policy — it sets, in advance, how much misclassification is tolerable and which error — trapping the wrong type or freeing it — to favor.

It elicits the fact by choice, not by compelled disclosure (Structured Disclosure Requirement) or third-party proof (Trusted Third-Party Attestation); acting on the revealed types by splitting or repricing a pool is Adverse-Selection Pool Segmentation.

Notes

Self-selection elicits the type; acting on the revealed types — splitting them into pools, repricing, excluding — is a separate mechanism (Adverse-Selection Pool Segmentation). Keeping the two apart matters because a menu that isn't incentive-compatible reveals nothing while still looking like it sorts, and any downstream segmentation built on it would then be acting on noise.