Eligibility Screening¶

An institutional review method — instantiates Search Space Pruning

Applies formal, published eligibility criteria to applicants, cases, or bids — with an owner, an audit trail, and an appeals path — so exclusions are accountable and reversible, not just efficient.

Eligibility Screening is what constraint-based pruning becomes when the candidates are people or their claims and the exclusion has consequences someone can be held answerable for. It applies formal, usually published criteria to applicants, benefit claimants, vendors, or proposals — but its distinguishing weight is not the filter itself; it is the governance wrapped around the filter. Every exclusion names the criterion it failed and is logged (the audit trail); a designated owner is accountable for the criteria and for handling disputes; and a rejected candidate has a defined appeals or exception path back into consideration. Where Constraint Filtering is the bare screening logic, Eligibility Screening is that logic made accountable and reversible because the cost of a wrong exclusion falls on a real person.

Example¶

A public agency runs a heating-assistance benefit. Eligibility Screening applies the formal criteria — household income below a published threshold, residency in the service area, no active benefit elsewhere — to each application. An applicant just over the income line is denied, and the denial is recorded with the specific failed criterion, the income figure used, and the caseworker who reviewed it.

That record is the point. When the applicant contests it — a recent job loss the snapshot missed — the appeals path lets a supervisor reopen the case against a documented reason, not a shrug. The named program owner reviews patterns in the denial log and notices the income snapshot systematically misses seasonal workers, triggering a criteria fix. The screen still cheaply removes clearly-ineligible applications; what makes it an eligibility screen rather than a raw filter is that every removal is attributable, contestable, and monitored.

How it works¶

Publish the criteria — the eligibility conditions are formal, documented, and knowable in advance, not discretionary.
Apply and record — each candidate is tested against the criteria, and every exclusion is logged with the failed condition, the evidence, and the reviewer.
Assign ownership — a named role owns the criteria, their updates, and the response to disputes and audits.
Provide reentry — a rejected candidate can appeal or seek an exception, reopening the case against a documented rationale.

The distinctive machinery is everything after the pass/fail: attribution, contestability, and accountable ownership.

Tuning parameters¶

Criteria formality — how rigidly rule-bound versus discretion-allowing the screen is. Rigid criteria are fair and auditable but brittle at the edges; discretion is flexible but harder to hold accountable.
Evidence standard — what proof a candidate must supply to establish eligibility; higher standards cut fraud but exclude legitimate claimants who cannot document their case.
Appeals threshold and window — how easy and how time-bounded reentry is; generous appeals catch wrong exclusions but add load and delay.
Audit-log granularity — how much of each decision is recorded; finer logs enable oversight and fairness review at more overhead.
Owner authority — how much power the owner has to grant exceptions or amend criteria without a formal process.

When it helps, and when it misleads¶

Its strength is legitimacy: it narrows a large caseload cheaply and makes each exclusion attributable, contestable, and correctable — indispensable where a wrong "no" harms a real person and the decision must withstand scrutiny.

It misleads when the criteria encode bias or drift out of step with reality and the surrounding governance is treated as paperwork. A screen can look impeccably procedural while its documentation requirements systematically exclude a group who cannot easily produce the paperwork — a disparate impact that a clean audit log will reveal only if someone actually reviews it.^[1] The appeals path can also become theatrical, formally present but so slow or opaque that reentry never really happens. The discipline is to treat the audit trail and appeals as live oversight instruments — reviewed by the owner and used to trigger criteria updates — not as compliance ornaments.

How it implements the components¶

constraint_filter — the formal, published eligibility criteria are the hard filter applied to every candidate.
reentry_or_exception_path — the appeals/exception process is a defined route for a rejected candidate to return to consideration.
pruning_audit_trail — each exclusion is logged with its failed criterion, evidence, and reviewer, making the screen inspectable.
pruning_owner — a named role owns the criteria, their updates, and the handling of disputes and audits.

It does not, in itself, test comparative quality or dominance (Dominated-Option Removal), and while it logs exclusions, systematically re-examining a sample of them for false negatives is Sample Audit of Exclusions's dedicated job.

Instantiates: Search Space Pruning — Eligibility Screening is the accountable, people-facing form of constraint-based pruning.
Consumes: Constraint Filtering — it wraps that screening logic in ownership, audit, and appeals.
Sibling mechanisms: Constraint Filtering · Sample Audit of Exclusions · Branch and Bound · Beam Search · Decision Tree Pruning · Dominated-Option Removal · Negative Keyword Filter · Red-Flag Screen · Safety or Compliance Exclusion · Shortlisting · Triage Filter

References¶

[1] Disparate impact is a facially neutral rule that nonetheless falls more heavily on a protected group. It is a real and regulated concern in hiring and benefits screening; the audit trail exists partly so such patterns can be detected and the criteria corrected. ↩