COHORT INTEGRITY - 0.2-E1

Count each choice.
Count it once.

A precise survival curve can still be wrong if the same decision-maker appears twice, inconvenient runs disappear, or sponsor-controlled homes masquerade as independent environments. Profile 0.2-E1 makes recruitment flow and independence auditable before W can rank.

Assess a cohort ->Download template Open contract

1:1CHOOSER / ENVIRONMENT

0POST-ACTIVATION EXCLUSIONS

0DUPLICATE IDS

7HARD GATES

01 / ANTI-CHERRY-PICKING

Follow the cohort
from screen to score.

The analysis set is fixed by activation. Replacement may add exposure, but it never erases the original run or its terminal disposition.

E1FROZEN ELIGIBILITY

Freeze selection rules before screening.

REQUIRED

E2SAMPLING FRAME

Disclose where participants came from.

REQUIRED

E3VOLUNTARY CHOICE

Never reward keeping the robot.

REQUIRED

E4PROSPECTIVE FLOW

Keep every activated run in analysis.

REQUIRED

E5DECISION UNITS

One residence, one primary chooser.

REQUIRED

E6TARGET + CARRYOVER

Meet N and reset reused hardware.

REQUIRED

E7INDEPENDENT AUDIT

Bind flow and hashed linkage evidence.

REQUIRED

02 / FLOW IDENTITY

No disappearing
households.

Every activated residence enters the analysis set. The public pack reports aggregate counts; independent auditors reconcile them against a controlled hashed linkage register.

INTENTION TO OBSERVEN_analysis = N_activated

Post-activation exclusions = 0. Replaced runs remain in the data. Unrelated exits are censored or reported as terminal causes under frozen endpoint rules.

03 / METHOD BASIS

Borrow the discipline.
State the boundary.

WANTED adapts established transparency principles to longitudinal human-robot residence studies. These references inform the profile; they do not make WANTED a clinical trial or claim conformity.

CONSORT 2025Participant flow

Report assigned, exposed, analysed, lost, and excluded units with reasons.

PRIMARY SOURCESTROBE COHORTSelection transparency

Report eligibility, participant sources, selection methods, stage counts, and follow-up.

PRIMARY SOURCEICH E9(R1)Estimand discipline

Separate the target question, estimator, estimate, and handling of intercurrent events.

PRIMARY SOURCE

04 / LOCAL EVALUATOR

Audit the count.
Protect the choice.

The checker validates arithmetic, chronology, decision-unit uniqueness, target thresholds, evidence bindings, and attestation topology without uploading participant data.

LOCAL COHORT-INTEGRITY EVALUATORPROFILE / 0.2-E1

AGGREGATE SELECTION + LINKAGE MANIFEST

{
  "profile_version": "0.2-E1",
  "target_certification": "WANTED_WILD",
  "study": {
    "study_id": "SYNTHETIC-STUDY-001",
    "preregistration_sha256": "b10123456789abcdef0123456789abcdef0123456789abcdef0123456789abcd",
    "eligibility_frozen_at": "2025-12-15T00:00:00Z",
    "screening_opened_at": "2026-01-01T00:00:00Z"
  },
  "sampling": {
    "frame_description": "Synthetic probability-informed community recruitment frame across declared operating geographies.",
    "recruitment_channels": [
      "community_registry",
      "public_call"
    ],
    "geographic_scope": [
      "US-CA"
    ],
    "inclusion_rule_count": 4,
    "exclusion_rule_count": 3,
    "base_compensation_independent_of_retention": true,
    "no_removal_penalty": true,
    "no_persuasion_after_removal_request": true
  },
  "flow": {
    "screened": 70,
    "eligible": 56,
    "consented": 50,
    "activated": 24,
    "analysis_set": 24,
    "pre_activation_exclusions": 26,
    "post_activation_exclusions": 0,
    "replacements_after_activation": 2,
    "original_runs_retained": true
  },
  "independence": {
    "independent_environments": 24,
    "unique_environment_ids": 24,
    "unique_primary_decision_makers": 24,
    "duplicate_environment_ids": 0,
    "duplicate_primary_decision_makers": 0,
    "maximum_environments_per_primary_decision_maker": 1,
    "sponsor_controlled_environments": 0,
    "developer_employee_environments": 0,
    "related_environment_clusters": 0,
    "hardware_units": 12,
    "reused_hardware_units": 12,
    "reset_between_assignments_attested": true
  },
  "evidence": {
    "participant_flow_uri": "https://example.org/wanted-participant-flow.json",
    "participant_flow_sha256": "c20123456789abcdef0123456789abcdef0123456789abcdef0123456789abcd",
    "hashed_linkage_register_uri": "https://example.org/wanted-hashed-linkage-register.json",
    "hashed_linkage_register_sha256": "d30123456789abcdef0123456789abcdef0123456789abcdef0123456789abcd",
    "public_aggregate_only": true
  },
  "assessor": {
    "name": "Synthetic Cohort Auditor",
    "organization": "Independent Example Assurance",
    "independent_of_sponsor": true,
    "attested": true,
    "signed_at": "2026-08-28T18:00:00Z"
  }
}

COHORT INTEGRITYPASSEDAll seven cohort-integrity gates pass.

SCREENED70

ANALYSIS SET24

SELECTION RATE34.3%

POST-ACTIVATION EXCLUSIONS0

FROZEN ELIGIBILITY

Eligibility and recruitment rules were frozen before screening opened.

PASS

DISCLOSED SAMPLING FRAME

Recruitment channels, geography, and inclusion and exclusion rules are disclosed.

PASS

VOLUNTARY CHOICE

Compensation and removal rights do not reward retention; developer employees are excluded.

PASS

PROSPECTIVE FLOW

Every activated environment remains in the analysis set, including replaced runs.

PASS

INDEPENDENT DECISION UNITS

Every analysis environment has one unique ID and one unique primary decision-maker.

PASS

TARGET + CARRYOVER

WANTED_WILD meets its environment threshold and physical carryover rule.

PASS

BOUND INDEPENDENT AUDIT

Participant flow and hashed linkage evidence are bound and independently attested.

PASS

INDEPENDENT != REPRESENTATIVE

WANTED reports the sampling frame, recruitment flow, geography, and selection rate. A certification never licenses generalization beyond the observed cohort.

OPEN PROTOCOL KIT ->

Count each choice.Count it once.

Follow the cohortfrom screen to score.

No disappearinghouseholds.

Borrow the discipline.State the boundary.

Audit the count.Protect the choice.

Count each choice.
Count it once.

Follow the cohort
from screen to score.

No disappearing
households.

Borrow the discipline.
State the boundary.

Audit the count.
Protect the choice.