Split delivery
Use class time to teach the disciplinary mechanism. Move the consequential decision into a separate bounded task rather than mixing explanation and assessment into one activity.
A FACULTY-ADOPTABLE PEDAGOGICAL FRAMEWORK
Teach knowledge. Transfer judgment.
ISCARB is a split-delivery framework for university teaching in the GenAI era. It separates disciplinary instruction from bounded decision practice, requires inspectable evidence, and keeps final professional responsibility with the human learner.
ISCARB = Interactive System for Case-based Analysis, Reasoning & Boundaries in the AI Era
PEDAGOGICAL MANIFESTO
Why ISCARB now?
Generative AI can produce plausible explanations, code and polished prose. The educational problem therefore shifts from “Can a student produce an answer?” to “Can a student own, bound and defend a consequential decision?”
ISCARB treats professional judgment as an observable learning outcome. A student may use tools to prepare evidence, but the student must still state what fits, where it stops fitting, what action follows, what evidence supports that action, and what new condition would force a change. The final sign-off remains human.
Evidence before confidence
Students do not earn credit merely for sounding certain. They expose assumptions, measurable boundaries, counter-evidence and inspectable artifacts.
Human sign-off
AI may draft, test, summarize or probe. It does not own certification, approval, risk acceptance, assurance or the final professional verdict.
THE REASONING GRAMMAR
FBR + STRESS
FITSelect the mechanism or option that is justified now.
BOUNDName the observable condition beyond which that choice stops being fit.
ACTTurn the judgment into a concrete professional action or artifact.
EVIDENCEProvide evidence another professional can independently inspect.
REFITAfter a changed condition, retain, revise or replace the decision.
STRESS is sequential by design. The learner commits the initial reasoning before seeing the changed condition. This makes assumptions and boundaries visible before hindsight can rewrite them.
INSTRUCTOR IMPLEMENTATION GUIDE
Apply ISCARB in any technical course
The framework is intentionally discipline-agnostic. The course supplies its own subject matter; ISCARB supplies the decision architecture.
1
Teach the mechanism In-Class
Identify what students must understand before judgment is possible: model, process, architecture, metric, hazard, algorithm, policy or design principle. Build the lecture around source figures, worked examples and explicit trade-offs.
2
Convert it into an FBR decision
Write a realistic case with incomplete but sufficient information. Require FIT, BOUND, ACT and EVIDENCE. Avoid questions that can be completed by paraphrasing the lecture or prompting an AI for a generic answer.
3
Verify ownership
Introduce STRESS, then use a short rubric and selective oral verification. Ask the learner to explain what changed, whether the boundary was crossed, and why the final verdict remains defensible.
LARGE COHORTS
Viva without exhausting the instructor
Rotating spot-checkOrally verify a changing sample of students each week rather than interviewing the whole cohort every time.
Risk-triggered vivaPrioritize unusual, internally inconsistent, highly polished or evidence-poor submissions for human verification.
2–3 minute micro-vivaUse the same three prompts: claim, boundary, evidence. Short fixed prompts improve consistency and throughput.
Calibrated panelsFor large classes, trained TAs or faculty panels can use the same rubric with periodic moderation and audit samples.
These are scalable implementation patterns, not claims that one sampling rate fits every course. Each instructor should set sampling intensity according to class size, assessment stakes and institutional policy.
TURNKEY ADOPTION KIT
What another instructor needs
The public site now exposes the methodology. The reusable faculty kit is the next layer for course transfer and institutional pilots.
In-Class Presenter templateReusable lecture shell with source figures, trade-offs, timing and decision prompts.
FBR assignment builderTemplate for FIT · BOUND · ACT · EVIDENCE · STRESS · REFIT.
Rubric + micro-viva sheetFast analytic scoring plus standardized ownership-verification prompts.
LMS setup checklistBlackboard / Canvas-neutral structure for release, submission, declarations and feedback.
Instructor-only guideDesign notes, common failure modes, hidden STRESS logic and exemplar boundaries.
Institutional pilot packCLO mapping, implementation log, student feedback instrument and evidence-capture plan.
Public methodology · available nowLive CPIT-455 implementation · available nowPackaged cross-course adoption kit · early access / on request
EMPIRICAL EVIDENCE & LEARNING ANALYTICS
Evidence before institutional claims
A framework intended for broader adoption must show more than an attractive implementation. It needs reproducible learning evidence.
Evidence program
- Learning-outcome performance before/after or with appropriate comparison designs.
- Quality of boundary-setting and evidence use across repeated FBR tasks.
- Student confidence in independent engineering judgment.
- Micro-viva agreement between written work and oral ownership.
- Instructor workload and scalability in large cohorts.
Current evidence status
ISCARB is implemented as a live student-facing course system in CPIT-455. This page does not claim cross-institution causal effectiveness or universal superiority. Such claims should follow a defined evaluation protocol, transparent measures and replicable data.
Adoption principle: publish the framework early enough to be tested, but reserve strong effectiveness claims for evidence that can survive independent scrutiny.
INSTITUTIONAL ADOPTION LADDER
From one course to a university model
Course pilot
One instructor, one course, explicit CLO mapping, a small number of FBR tasks, and documented workload.
Department pilot
Multiple instructors across different technical subjects use a shared rubric and common implementation log.
Multi-institution replication
Independent teams implement the same reasoning grammar, report deviations, and compare learning and workload evidence.
Current course implementation · one shared learning artifact
The nine CPIT-455 chapters use a 20-slide classroom route, five local objectives, short explanation points and complete source review. Three timed, coached think–pair–share stations build one decision card. The assessed case remains separate and reveals its changed condition only after commitment.
The canonical 20-rule map distinguishes architecture (07), portfolio launch (16), constraint mutation (17), evidence policy (18), four-level capability rubric (19), and bounded assurance/professional decision (20). Old unit IDs are not reused as rule numbers.
Inspect all 20 rules and their actual implementation · Teaching workflow. This implementation is not evidence of causal effectiveness or official accreditation.