How the analysis is evaluated
High achievement requires disciplinary grounding, authoritative evidence, fair comparison of alternatives, a bounded workflow, qualified human review, and a justified recommendation. A fluent generated output is evidence to evaluate, not proof of disciplinary validity.
Criteria and weighting
| Criterion | Points | Focus |
|---|---|---|
| Use case, purpose, scope, and disciplinary context | 15 | Legitimate purpose, outcome, boundaries, standards, authority, information, assumptions, and exclusions |
| Current workflow, stakeholders, and alternatives | 15 | Existing practice, affected users and non-users, needs, AI and non-AI options, criteria, and trade-offs |
| Proposed workflow, AI role, and human responsibility | 15 | Inputs, sources, stages, outputs, review gates, bounded AI role, human-only work, PARTNER, escalation, and stopping |
| Evidence, claims, sources, and validation | 15 | Claim checking, source authority, method, qualified review, acceptance criteria, correction, limitations, and uncertainty |
| Disciplinary framework application | 15 | Relevant selection and accurate application of TRACE, CLAIM, LEARN, RIGOUR, VOICE, WORKS, LIFE, or justified supporting frameworks |
| Responsible practice and safeguards | 10 | Accessibility, inclusion, privacy, security, permission, authorship, attribution, integrity, disclosure, controls, and fallback |
| Integrated recommendation and residual risk | 10 | Comparison, proportionality, conditions, qualified conclusion, unresolved uncertainty, and residual risk |
| Communication, accessibility, citations, and disclosure | 5 | Organization, clarity, source integrity, accessible presentation, student authorship, and accurate disclosure |
| Total | 100 |
Performance levels
| Level | Descriptor |
|---|---|
| Excellent, 85–100% | Discipline-specific, precise, evidence-rich, integrated, operational, and convincingly justified. |
| Proficient, 70–84% | Complete and well reasoned, with effective evidence, workflow, framework application, safeguards, and recommendation; minor gaps remain. |
| Developing, 55–69% | Meets many requirements, but disciplinary evidence, comparison, validation, framework application, or responsibility is uneven. |
| Limited, 40–54% | Primarily descriptive, generic, or incomplete; the workflow, evidence, authority, safeguards, or recommendation is weak. |
| Insufficient, below 40% | Does not demonstrate a responsible discipline-specific analysis, or serious evidence, information-handling, or authorship concerns prevent assessment. |
Assessment guidance
Indicators of excellent work
- The discipline's standards and methods materially shape the analysis.
- Alternatives are compared with criteria defined in advance.
- The AI role is bounded and human-only work is explicit.
- Claims and sources are verified with qualified validation.
- Safeguards have owners, evidence, and failure responses.
- The recommendation is qualified and states residual uncertainty.
Common reasons for deductions
- Generic discussion of AI rather than a bounded application
- Vendor claims or generated citations accepted without verification
- Framework name-dropping or mechanical checklists
- No affected non-users or non-AI alternatives
- Nominal human review without qualifications or authority
- Missing accessibility, permission, correction, fallback, disclosure, or residual risk
Grade calculation
The rubric score is converted to the assignment's 15% course weight.
Example: 80/100 contributes 12 percentage points to the course grade.