Design probe · network nomination instrument

Nomination Bench

Social network studies ask respondents to name peers, then spend months resolving those names to people. This bench resolves at collection time instead, while the one person who actually knows the answer is still sitting there, and it abstains rather than guessing. The panel on the right shows what that choice does to the peer change agents an intervention would select.

Synthetic roster and synthetic prior responses. No real person appears on this page.

Item 4 · Health talk network

v1.2.0
site: drop-in-center-a roster: 22 respondent: RS-0147

In the last 30 days, who did you talk with about staying healthy?

Up to five people. First names, nicknames or street names are all fine. Leave a row blank if there is no one else.

Resolution console

site to date
Nomination ledger
WroteStateResolved to

Abstain policy

An ambiguous nomination counts toward no one until a human resolves it.

    Force-match policy

    An ambiguous nomination is assigned to the most-nominated candidate, the usual prior-based default.

      What the bench is arguing

      The respondent is the disambiguating oracle

      Resolving nominated names after fieldwork is a modelling problem. Resolving them during the interview is a question. Asking "which Mike?" costs four seconds and beats any posterior estimate, because the person who made the nomination is the only one who knows.

      cf. Peer Disambiguation in Self-Reported Surveys using Graph Attention Networks, 2025

      Abstention has to be a recorded state

      Matched, ambiguous and unmatched are three outcomes, not two. Collapsing ambiguity into a best guess makes an unresolved nomination indistinguishable from a confirmed one downstream, and nothing later in the pipeline can recover the difference.

      three states, never two

      Resolution policy selects the intervention

      Peer change agent selection runs on degree and position. Force-matching sends ambiguous nominations to whoever is already most named, which inflates the already-central and quietly changes the roster of youth an intervention recruits.

      cf. HEALER / DOSIM peer change agent selection

      The unmatched row is a finding, not an error

      A name no one on the roster carries usually means a real person outside the frame: a peer at another site, someone who stopped coming, an adult. Discarding those rows silently shrinks the network to the convenience sample and hides the boundary.

      absent is not the same as missing