How we align with MIT's 2026 recommendations
In August 2026 the MIT Ad Hoc Committee on AI Use in Teaching, Learning, and Research Training published recommendations for teaching in an AI-aware way. MIT says its recommendations are "not a checklist." This page is our own mapping, offered so you can judge the fit yourself. MIT has not reviewed or endorsed this product.
Our 10 principles, and the MIT recommendations they align with
| Principle | What a rebuilt syllabus must include | Related MIT section | How close |
|---|---|---|---|
| Skills build in order | Units follow prerequisite order and AI-off checks come after the skill is taught. | §2.5 backward design; §3.1.1 revisit course goals | Partial |
| Support that fades | AI support steps down across the term: Full → Shared → Light → Off. | §2.6 no one size fits all; §2.7 'cognitive surrender'; §3.1.10 'AI Light' / 'AI Heavy' | Partial (fading within a course is our extension) |
| Clear limits on AI | Every assessment says what AI may not do, with a reason. | §3.1.8 policy with a rationale for why AI 'must, may, or may not be used'; §3.2.4 'when not to reach for AI at all' | Strong |
| Defined AI roles, human instructor | AI roles are named by function with limits; the instructor gives final feedback and grades. | §3.2.3 students object when the only feedback is from a machine | Partial (role-limited AI is our extension) |
| Student choice, with a low-AI path | Students choose how to use AI where it's allowed, and a low-AI pathway is always offered. | §3.1.10 'a pathway that keeps the use of AI to a minimum' | Partial |
| Independent work counts most | Grade weight favors verified independent performance (defenses, portfolios, projects), not only timed exams. | §3.1.6 reconsider grades and incentives; §3.1.2 caution on overweighting in-class exams | Partial (MIT poses the question; this is our answer) |
| Human-only assessments | Each unit includes an assessment only the student can do: oral defense, portfolio conversation, live build. | §3.1.2 'oral exams, semester portfolios, and out-of-class assignments paired with in-class conversations' | Strong |
| Peer and real-audience review | Structured peer review with a review-quality criterion. | §3.1.4 structured in-person social learning; feedback discussions around a class rubric | Partial |
| Visible AI process | Students attach an AI-use note or version history; missing notes cap the score; logging is disclosed. | §3.1.9 version history as 'useful process evidence'; §3.3.9 logging policy | Strong |
| Questions back and mastery checks | Zero-point cold checks, retakeable mastery quizzes, and AI that asks questions back. | §2.4 metacognition; §3.1.6 real-time feedback; §3.1.9 feedback at stages | Partial |
Where we differ. MIT cautions against leaning too heavily on timed in-class exams, so our "independent work counts most" check puts the weight on defenses, portfolios and projects. MIT asks institutions to set clear logging and privacy policies, so whenever a syllabus requires AI logs, it must say what's logged and who sees it. MIT doesn't describe support that fades within a course or role-limited AI; those are our own extensions.
13 additional checks from the MIT report
- Policy posted prominently, standardized format (§3.1.8)
- Rationale tied to learning goals (§3.1.8)
- AI can fabricate facts and citations; students own their work (§3.1.8)
- Students reflect on when AI helped or hurt (§3.1.8)
- No reliance on AI detectors; staged checkpoints instead (§3.1.9)
- Instructor discloses own AI use (§3.2.3)
- Day-one case for in-person interaction (§3.1.4)
- Productive struggle named as the purpose (§2.4)
- Effective, responsible and ethical use taught (§3.2.4)
- Support pointer for AI dependence or stress (§3.2.2)
- Experiential and project work (§3.1.3)
- Integrity process placeholder; detector output alone isn't evidence (§3.1.9)
- Privacy and logging disclosed (§3.3.9)