Evaluation Dashboards — Astronaut Candidate Selection Platform

Total evaluated

—

awaiting run

Approved

—

Waitlisted

—

Declined

—

AI–Human agreement

—

Decision funnel live

⏳

Run evaluation to see funnel

Outcome distribution

Routing breakdown

Score distribution approve ≥70 · waitlist 50–69 · decline <50

AI confidence vs score colored by final outcome

Cases reviewed by human

—

Agreement rate

—

Disagreements

—

Avg score adjustment

—

AI score vs human-adjusted score ⬤ agree ⬤ disagree

AI recommendation vs final outcome

Disagreement cases

Auto Decision

—

Parallel Review

—

Escalation

—

Senior Adjudication

—

Cases per routing tier

Avg score & avg composite risk per tier

Outcome distribution per routing tier

Mean score

—

out of 100

Median score

—

50th percentile

Near-cutoff cases

—

within ±5 pts of 50 or 70

Score std dev

—

spread of scores

Score histogram stacked by outcome · approve ≥70 · waitlist 50–69

Avg rubric dimension scores each dimension 0–25

Score by confidence level

Rubric dimension breakdown per outcome

Avg composite risk

—

0 = low · 1 = high

High-risk cases

—

composite risk > 0.6

Near-cutoff cases

—

within ±5 pts of boundary

Evidence mismatches

—

score ≠ narrative signal

Sensitivity vs disagreement score bubble = composite risk · color = routing

↖ Low sensitivity, high disagreement

↗ High sensitivity & disagreement → escalate

↙ Low risk → auto-decide

↘ High sensitivity, low disagreement

Composite risk distribution per routing tier

Rubric variance distribution high variance = uneven dimension scores

Signal flags breakdown

Total human reviews

—

Active reviewers

—

unique reviewers

Drift flags

—

agreement rate < 60%

Largest avg adjustment

—

pts from AI score

Agreement rate per reviewer

Avg score adjustment per reviewer + = reviewer scores higher than AI

Reviewer summary