Shortlisting¶

A decision-process artifact — instantiates Search Space Pruning

Reduces a broad field to a small, deliberately varied working set that a team can evaluate in depth — a soft, reversible narrowing that keeps the finalists distinct rather than clustered.

Shortlisting produces an artifact — the shortlist: a small, named set of finalists carved from a field too large to evaluate deeply. Its defining character among the siblings is that the cut is soft and reversible, and that it does not merely keep "the top N by score." A good shortlist is chosen to stay diverse — to span genuinely different candidates so the deep evaluation compares real alternatives, not five variations of one favorite. Where a hard filter removes what fails a bar and dominance removal cuts only the provably-pointless, Shortlisting deliberately sets aside many perfectly viable options to make detailed comparison feasible — and takes on the specific risk that it drops a good one, which is why it guards diversity and keeps the cut reopenable.

Example¶

A firm receives 22 responses to an RFP for a large office redesign and cannot run full evaluations — reference calls, site visits, cost deep-dives — on all of them. Shortlisting produces a set of 4 finalists. A naive "top 4 by score" would have surfaced four large, similar corporate firms, because they scored alike on the scoring rubric. Instead the shortlist is built to span the space of approaches: one large full-service firm, one boutique known for sustainability, one strong on cost, one with a bold unconventional concept.

That diversity is the point. The deep evaluation now compares meaningfully different bets rather than near-clones, and the two rejected large firms — viable, just redundant with the one that made it — are kept on a reserve list in case a finalist drops out. The shortlist is the working artifact the whole selection now runs on, chosen to keep the real range of options alive at a size the team can actually study.

How it works¶

Bound the field — take the full defined set of candidates as the starting space to narrow.
Set the target size and the success criterion — how many finalists the deep evaluation can afford, and what the search is ultimately trying to achieve.
Select for coverage, not just rank — choose finalists that score well and span distinct families, so the shortlist represents the space rather than one corner of it.
Keep it soft — treat non-finalists as set-aside (a reserve), not deleted, so the cut can be reopened if a finalist fails.

The distinctive move is optimizing the set for diversity and manageability, not picking items independently by score.

Tuning parameters¶

Shortlist size — how many finalists. Larger preserves more options and diversity at higher evaluation cost; smaller is cheaper but more likely to drop a good candidate.
Diversity weight — how hard the selection pushes for coverage versus raw score. High diversity guarantees varied finalists but may admit a lower-scoring outlier; low diversity risks a homogeneous shortlist.
Score-vs-judgment balance — how much the cut leans on a rubric versus expert read; rubrics are consistent, judgment catches what the rubric misses.
Reserve depth — how many near-miss candidates are held ready in case a finalist withdraws.
Reopening rule — the conditions under which a set-aside candidate can re-enter the shortlist.

When it helps, and when it misleads¶

Its strength is making deep evaluation feasible while keeping the real range of options alive: it cuts a field to a studiable size, and by protecting diversity it prevents the classic failure of finalists who are all the same safe bet. As a named artifact, it also makes the narrowing transparent and revisable.

It misleads when it collapses to "top N by a single score," quietly producing a homogeneous shortlist that looks thorough but explored one region — and when the softness is forgotten, so a set-aside candidate is treated as permanently gone. There is also a consideration-set trap: candidates outside the shortlist effectively vanish from the decision, so a good option cut early never gets a fair hearing.^[1] The disciplines are to select the shortlist for coverage as well as merit, keep a reserve, and make the cut genuinely reopenable rather than final.

How it implements the components¶

search_space_definition — it starts from an explicit full field and defines the reduced working set drawn from it.
objective_or_success_criterion — the shortlist is built toward what the selection is ultimately trying to achieve, not an arbitrary trim.
diversity_preservation_rule — its signature safeguard: choosing finalists that span distinct families so the surviving set doesn't collapse into near-duplicates.

It applies no hard constraint (Constraint Filtering) and issues no dominance proof (Dominated-Option Removal); it is a soft, reversible narrowing, and auditing whether it dropped a good candidate belongs to Sample Audit of Exclusions.

Instantiates: Search Space Pruning — Shortlisting is the soft, diversity-preserving reduction to a working candidate set.
Consumes: Dominated-Option Removal — an uncontroversial dominance cut is the natural step before shortlisting the surviving frontier.
Sibling mechanisms: Dominated-Option Removal · Triage Filter · Branch and Bound · Beam Search · Constraint Filtering · Decision Tree Pruning · Eligibility Screening · Negative Keyword Filter · Red-Flag Screen · Safety or Compliance Exclusion · Sample Audit of Exclusions

References¶

[1] In decision research a consideration set is the small subset of options a chooser actually evaluates; options outside it are almost never chosen regardless of merit. A shortlist is an engineered consideration set, which is why what it leaves out matters as much as what it keeps. ↩