Designing an AI Literacy Curriculum That Works for Every Major
A shared conceptual core plus major-specific practice. Why one general course for the whole cohort fails, and what to assess instead.
An AI literacy curriculum that only works for computer science students is not an AI literacy curriculum. The approach that holds up across majors is a shared conceptual core plus major-specific practice: every student learns the same foundations, then applies them to artefacts from their own discipline.
The problem with one course for everyone
Institutions usually try one of two things. Either they run a single general course for the whole cohort, which is too abstract to transfer into any profession, or they let each department build its own, which produces wildly uneven quality and no shared standard. Both fail for the same reason: they treat "foundations" and "application" as a single decision when they are two.
A two-track structure
Track one: the shared core
Identical for every major. What a model is, why training data determines behaviour, what these systems are bad at, how to evaluate an output you cannot verify yourself, and where the ethical and legal limits sit. This track is assessed the same way across the institution, which is what makes it a standard.
Track two: major-specific practice
Different for every major, built on artefacts the student already recognises. The design student works on a brief. The nursing student works on a case note. The accounting student works on a reconciliation. The task changes; the underlying competency being assessed does not.
What "major-specific" actually means
| Major | Shared core | Applied task |
|---|---|---|
| Business / commerce | Same | Evaluate an AI-generated market summary against source data |
| Design / media | Same | Produce and critique a generated asset set against a brief |
| Health sciences | Same | Identify where an AI summary of a case note loses clinically relevant detail |
| Engineering | Same | Debug and test model-generated code |
| Humanities | Same | Trace and verify claims in generated text |
Sequencing that survives contact with a real timetable
Three stages, in order: understand core concepts, then hands-on practice, then ethical and critical reflection. Reflection last is deliberate. Students who have not yet watched a model fail at something they personally understand treat the ethics session as an abstraction. Students who have just debugged a confident wrong answer do not.
What makes it fail
- No lab time. A literacy course taught entirely as lectures produces students who can define AI and cannot use it.
- Assessment by recall. If the exam asks what a transformer is, the course will be taught to that question.
- Faculty a level behind the students. This one is not fixable with curriculum. It needs a separate faculty track.
- Content that ages out. Anything pinned to a specific tool's interface needs rewriting every term.
Scale reference
EDU360 Cloud maintains a knowledge base covering 29,000+ course directions and 40+ ready-to-deliver practice courses, spanning STEM and humanities majors. The platform has been deployed with 500+ institutions and 400,000+ learners and educators in China.
FAQ
How many contact hours does a literacy course need?
Enough for the practice track to be real rather than a demonstration. The constraint is lab hours, not lecture hours.
Should this be compulsory?
Institutions that make the shared core compulsory and the applied track department-owned tend to get further than those that make the whole thing optional, because the shared core is what creates the institutional standard.
