# A11 proposed instruments · v1.0

These are new research instruments, not validated scales. Pilot forms and scoring must be frozen before the main study. All case quantities and organizations below are fictional teaching/assessment inputs. No participant work or source questionnaire is reproduced.

## Transfer task (0–24)

Provide a fictional community-monitoring decision, two evidence sources with different sampling periods, a short account of uncertainty, a budget, and a concern raised by an affected community. Learner recommends an action, explains supporting/contradictory evidence and states an uncertainty. Introduce a changed constraint (e.g. sensor failures plus a stricter confidentiality requirement); learner revises and justifies the decision. A parallel form uses a different substantive setting, equivalent reading demands and a different changed constraint. Disclose form order, baseline equivalence evidence and accommodations. Do not reward memorized wording or factual knowledge not supplied in the packet.

Two masked human raters independently score six dimensions0–4. Common anchors:0 absent/contradicts task;1 asserts an answer without adequate reason;2 partially warranted reasoning with a consequential omission;3 coherent relevant reasoning addressing main constraints;4 integrates counterevidence/uncertainty and explains limits robustly. Dimension-specific requirements:

1. Problem framing: distinguish decision, affected audience and success criteria;4 identifies a consequential tension rather than only restating the prompt.
2. Evidence selection: connect claims to the appropriate supplied source;4 addresses sampling/time mismatch and evidential scope.
3. Uncertainty: identify what is not established;4 links uncertainty to a feasible decision or information need.
4. Alternative account: treat a credible competing explanation;4 explains which evidence would discriminate it.
5. Revision under changed constraint: change a substantive inference/action where warranted;4 explains both what changed and what remained defensible. Cosmetic rewriting is at most1.
6. Accountable decision: justify action to affected people;4 identifies tradeoffs, proportionate safeguards and a review trigger.

Every task produces initial response, new-constraint response and explanation. Raters see required form context, not cluster/condition/mentor identity. Mean of two total ratings is primary; third-rater discrepancy audit does not selectively overwrite it. Pilot must examine readability, translation/accessibility, cognitive interviews, form comparability, ceiling/floor behavior and dimension-level disagreement. A trained rater has not thereby validated the instrument.

## Provenance mastery case task (0–16)

Four unfamiliar cases scored0–4 using case-specific anchors.0 materially wrong;1 identifies a concern without explanation;2 partly correct but misses a consequential distinction;3 accurate distinction plus feasible action;4 accurate, justified action with limits/permission boundaries.

A. Contribution: fictional learner combines a published chart, a new analysis and mentor editing. Explain who contributed what, what needs citation, and why editorial/platform help does not automatically make every helper an author. No single universal authorship policy is assumed; apply the stated venue policy and contribution evidence.
B. Rights: metadata isCC0 but linked resource statesCC BY-NC-ND. Explain why metadata rights do not authorize unrestricted adaptation of the resource. A source text is not provided or republished; case uses license labels only.
C. Versions: a DOI record has metadataVersion7,resource version1.2 and an IsVersionOf link. Explain these distinctions and choose which artifact was actually examined. A later registry update is not automatically a new intellectual work.
D. Identity/audience: a record contains a checksum-valid ORCID string and a learner declines public release. Explain why string validation is not account authentication and propose a private assessed route with no grade penalty. Identify permissions needed before any public association or external-account write.

## Technical portability audit (0–8; process only)

Use synthetic or separately permission-cleared package. Export from platformA, import into independent validatorB, export again. For each item award1 only if semantic value and its access boundary survive both transitions; otherwise0 with a reason: artifact key; human contribution assertion; source identifiers; resource version; typed version relation; resource-rights assertion; access state; distinct assessment/editorial status. Save source/export/reimport hashes and validator version. Do not register a real DOI or write to a real ORCID account for this test. Perfect portability is not proof of understanding.

## Time and agency instruments

Designated diary windows at baseline,week10,week20. Learner/mentor record planned paid course/support minutes, additional paid minutes, uncompensated minutes, access/rework cause and publication/private-path task category. Missing diary≠zero time. Use brief respondent-selected workload burden categories and optional narrative; narratives restricted.

Agency interview prompts: What audience options were available? Could you decline public release without losing credit? Who could change the contribution statement? What would happen after rejection? Which decisions did you actually control? Interview explanations are qualitative mechanism evidence, not the primary score. No published-work count is converted into an agency score.

Safeguarding log: date, event category, exposure/access risk, action, resolution and reporting decision. No publicly identifying narratives. Study leaders may pause release independently of statistical consequences. Rules, responsible roles and limits of removal from external copies must be explained before consent.