AI-Augmented Engineering Judgment (AAEJ) v1.0¶
Version: v1.0
Status: Active (Spring 2026 Implementation)
Section 1 - Purpose¶
Engineering education in the AI era requires structured integration of AI tools that strengthen, rather than replace, student reasoning.
AAEJ defines a controlled framework for AI use that preserves independent thinking while leveraging AI for concept clarification and reasoning development.
Section 2 - Core Principle¶
Core Principle
AI is used to sharpen reasoning, never to substitute for it.
Section 3 - System Architecture¶
Layer 1 - No AI (Foundation)¶
Goal: build raw problem-solving ability
Activities: hand calculations, setup, derivations
Layer 2 - Guided AI (Gerbert)¶
Gerbert is the course's guided AI assistant mode used only for reasoning support, not answer production.
Goal: concept clarification and reasoning development
Constraints:
- no final answers
- prompts only
Layer 3 - Independent Output¶
Goal: demonstrate ownership of reasoning
Requirement: student must explain without AI
Section 4 - Allowed vs Not Allowed¶
| Allowed | Not Allowed |
|---|---|
| clarifying concepts | generating final answers |
| interpreting system behavior | copying AI explanations |
| checking reasoning | submitting unverifiable reasoning |
Section 5 - Verification Protocol¶
Submission Completeness Rule
Every submission must include all three elements:
- Observation - what happened
- Mechanism - why it happened (physics-based)
- Failure Condition - when it would not hold
If any are missing -> incomplete.
Section 6 - Enforcement Model¶
- Students must be able to orally or in writing defend their work.
- Responses lacking mechanism or interpretation are downgraded.
- AI use is permitted only in designated phases.
Understanding must be independently demonstrable.
Section 7 - Example (Rankine Cycle)¶
Net work was positive because turbine work is bigger than pump work.
Aspen showed turbine work magnitude significantly exceeded pump work, giving positive net cycle work. This aligns with phase-dependent enthalpy behavior: compressing liquid requires much less specific work than expanding superheated vapor over a similar pressure range. This conclusion would fail if turbine efficiency dropped substantially (for example at off-design operation), reducing turbine work enough to collapse the net-work margin.
- Weak response states an outcome but provides no mechanism and no failure condition.
- Strong response includes observation, mechanism, and failure condition, so the reasoning is verifiable.
Section 8 - Implementation Notes¶
- Used in CBE 251 and CBE 253.
- Integrated with Aspen workflows.
- Paired with Warrant-style reflection exercises.
Reference link: Faculty Adoption Guide