Responsible AI deployment review rubric

// Moderate the review of a bounded AI deployment, its human oversight and its monitoring with distinct evidence-based descriptors; this illustrative rubric is not a validated instrument.
// Synthetic worked example: all entities, schedules and measurements are illustrative.
// Every string below is EXAMPLE text from a fictional model: replace it.
// Replace assumptions and inputs before applying this model to actual work.

rubric assessment {
  title: "Illustrative Responsible AI deployment review rubric"
  course: "Advanced professional practice"
  assignment: "Evaluate a bounded deployment with human oversight and monitoring"
  author: "Flowss teaching example"
}

criterion criterion_0 {
  name: "Use context"
  weight: .25
  description: "Affected people and purpose are explicit"
}

level level_0_0 {
  criterion: criterion_0
  label: "Emerging"
  points: 1
  descriptor: "Defines only model capability"
}

level level_0_1 {
  criterion: criterion_0
  label: "Proficient"
  points: 2
  descriptor: "Describes users, harms and benefits"
}

level level_0_2 {
  criterion: criterion_0
  label: "Advanced"
  points: 3
  descriptor: "Identifies excluded uses and affected non-users"
}

criterion criterion_1 {
  name: "Evaluation"
  weight: .25
  description: "Evidence matches deployment conditions"
}

level level_1_0 {
  criterion: criterion_1
  label: "Emerging"
  points: 1
  descriptor: "Reports one aggregate score"
}

level level_1_1 {
  criterion: criterion_1
  label: "Proficient"
  points: 2
  descriptor: "Evaluates relevant slices and failure cases"
}

level level_1_2 {
  criterion: criterion_1
  label: "Advanced"
  points: 3
  descriptor: "Uses independent data and uncertainty analysis"
}

criterion criterion_2 {
  name: "Human oversight"
  weight: .25
  description: "Intervention and appeal are actionable"
}

level level_2_0 {
  criterion: criterion_2
  label: "Emerging"
  points: 1
  descriptor: "Mentions a human in the loop"
}

level level_2_1 {
  criterion: criterion_2
  label: "Proficient"
  points: 2
  descriptor: "Defines authority and escalation routes"
}

level level_2_2 {
  criterion: criterion_2
  label: "Advanced"
  points: 3
  descriptor: "Tests workload, override and appeal effectiveness"
}

criterion criterion_3 {
  name: "Monitoring"
  weight: .25
  description: "Post-release accountability is sustained"
}

level level_3_0 {
  criterion: criterion_3
  label: "Emerging"
  points: 1
  descriptor: "Assumes launch evaluation is enough"
}

level level_3_1 {
  criterion: criterion_3
  label: "Proficient"
  points: 2
  descriptor: "Assigns monitoring and incident ownership"
}

level level_3_2 {
  criterion: criterion_3
  label: "Advanced"
  points: 3
  descriptor: "Defines drift triggers, rollback and review cadence"
}

view criteria: rubric_grid