aiX Weekly — AI in Higher Education (August 26th, 2026)

“Adding AI to the curriculum” can mean three very different things.

Teaching with AI uses the technology as an instructional enhancer. Teaching how to use AI builds skills for carrying out disciplinary inquiries: prompting, tool selection, the “AI competency” now appearing in graduation requirements. Teaching about AI is a third thing: it treats AI as an object of study; It may cover asking how these systems actually work, what they are made of, whose labor and data they encode, and what they do to knowledge, learning, and power. This issue looks at that third area.

Each aiX Weekly post is organized around a set of recurring sections and pairs with our companion AI and Higher Education timeline, which traces the broader arc of how AI has reshaped higher education since late 2022.

Curated by Claude for the aiX Programs, Columbia University. AI can make mistakes and so can the human reviewer. Please double-check the linked sources.

Review by Tian Zheng on August 26, 2026

Scope note: This issue is thematic rather than tied to a single news cycle.

TL;DR

• The field is learning to distinguish three projects. Teaching with AI, teaching how to use AI, and teaching about AI often share the same tools and activities, but they aim at different learning goals: from mastering a subject, gaining a skill, or understanding a system. Naming that difference is itself a sign for progress in AI education.
• Functional competency in AI is a foundation. What matters more is understanding. Fluency with the tools is a genuine achievement to build on; the next step, increasingly within reach, is helping students understand what they are using and not only use it well.
• “About AI” has two strands. A conceptual/technical strand — how models actually work, why fluent output can still be wrong — and a critical/sociotechnical strand — the data, labor, capital, and energy the systems are built from.
• The deep materials already exist. Annette Vee’s critical-AI-literacy taxonomy, Henry Farrell’s Political Economy of AI syllabus, George Mason’s prerequisite-free UNIV 182, and Oberlin’s new Critical AI Studies minor are all curricula about AI.

Table of Contents

• In Focus
• This Week at a Glance
• Research Highlights
• What’s in the News
• Institutional Movements
• Most Discussed
• Try This Week

In Focus

Three ways to “add AI to the curriculum” — and why the differences matter.

Our vocabulary in AI education is catching up to a distinction that has been quietly sharpening. When a campus says it is “adding AI to the curriculum,” it could mean three genuinely different things. It might mean teaching with AI — using tutors, graders, and generators to help teach existing subjects. It might mean teaching how to use AI — the functional skills of prompting and tool selection that now anchor most “AI literacy” requirements. Or it might mean teaching about AI — studying the technology itself: how it is built, how it works, what it is made of, and what it does to the people and institutions around it. The first is pedagogy. The second is training. The third is a subject of inquiry — and it is the one now coming into focus.

The three are easy to confuse because they look alike in practice. They draw on overlapping designed components — the same tools, the same readings, often the very same activity. What separates them is not the activity but the learning goal: what a student should be able to do, or understand, once it is over.

Consider one designed component: students paste a historical document into a chatbot and ask for a summary. In a teaching-with course, the goal is comprehension — the summary helps them grasp the history. In a how-to-use course, the goal is skill — they iterate on the prompt and judge which summary is better. In a teaching-about course, the goal is understanding — they examine what the model dropped, added, or distorted, and ask why: what the training data rewards, whose perspective is encoded, what the system optimizes. One activity, three different things learned and three different things assessed. The components overlap; the alignment to a goal is what makes a course one kind of thing rather than another — which is why getting clear about goals is such a practical gain. Once we know which of the three we intend, we can design the assessment that actually reaches it.

The functional version arrived first, and for good reason: it is the easiest to stand up. A prompting workshop needs a tool and a room, while a course about AI has to reach into computer science, economics, philosophy, history, and labor studies at once. That sequencing is natural — and the encouraging development is that the deeper layer is now being built on top of it. Writing scholars have helped by naming what that layer adds: beyond using AI well, real literacy means being able to evaluate its outputs and to reason about the systems, the labor, and the data behind them. Annette Vee puts the functional layer first precisely so that fluency can grow into understanding rather than stop at competence.

This Week at a Glance

This issue steps back from the week’s news to sort out what “teaching about AI” really means — and to point to the curriculum, maturing quickly, that already exists to do it. The timely marker is Oberlin, which this fall becomes one of the first liberal-arts colleges to offer a full Critical AI Studies minor. The deeper materials are not hypothetical — published syllabi, prerequisite-free courses, and, as of this fall, an interdisciplinary minor — and they sort into the two strands that “teaching about AI” contains: understanding how the systems work, and understanding what they are made of.

Relevant to faculty: Before you add a tool to your course, decide which of the three directions you are actually undertaking. If the answer is “about AI,” the goal is not fluency with a product; it is understanding a system — and you can borrow a rigorous outline rather than build one.

Relevant to institutional leaders: A functional-competency requirement and a critical-studies offering do complementary work. The opportunity now is to give your curriculum a clear home for the about — not only the how — so the two can reinforce each other.

Relevant to students and researchers: The most transferable thing you can learn about AI is not how to drive it but how to interrogate it: where the training data came from, who labeled it, what the model optimizes, and why a confident answer can still be false.

Research Highlights

What is Critical AI Literacy? (Annette Vee, AI & How We Teach / W. W. Norton, February 2025; foundational reference.) Vee separates AI literacy into three layers: functional (can you use it, and do you know how it works?), rhetorical (can you evaluate its outputs and judge when its use is appropriate?), and ethical (do you understand its societal, environmental, and personal risks?). The “critical” in critical AI literacy, borrowing from Maha Bali, means both questioning and critiquing — not just operating the tool but interrogating it. The piece is also refreshingly concrete: it links the MLA–CCCC Joint Task Force working paper on critical AI literacy and describes classroom exercises that teach evaluation directly.

Practitioner framework; functional / rhetorical / ethical literacy

💬 Editor’s note: Vee’s three layers give us a ladder. Functional literacy is the first rung; evaluation and critique are where “about AI” begins. What I find hopeful is how naturally a writing class can climb from one rung to the next once the layers are named. — TZ

From Understanding to Creation: A Prerequisite-Free AI Literacy Course with Technical Depth Across Majors (George Mason University, UNIV 182; arXiv, March 19, 2026.) This is the conceptual/technical strand done seriously. The course takes undergraduates of any major through a single pipeline — problem definition, data, model selection, evaluation, reflection — traversed repeatedly at rising sophistication, with ethical reasoning integrated at each pass rather than bolted on. Instructor-coded analysis of student work documents a progression from intuition-based description to technically grounded design, reaching the “Create” level of Bloom’s taxonomy. Its central claim is important: technical depth and broad accessibility can coexist when the scaffolding supports both. Understanding how AI works need not be reserved for computer science majors.

Course design study; prerequisite-free, technically deep

💬 Editor’s note: The premise I find most useful here is that “no prerequisites” and “real depth” are not in tension. We too often assume that non-majors can only be taught to use AI, never to understand it. This course is a counterexample. — TZ

The Political Economy of AI: A Syllabus (Henry Farrell, Programmable Mutter, July 2025; available open syllabus.) If UNIV 182 is the technical strand, this is the critical/sociotechnical one, laid out reading by reading. Farrell organizes AI as a political-economic system: inputs (capital, chips, energy, and human knowledge — the data), outputs (markets, the state, culture), and conflicts (AGI debates, geopolitics, labor). The reading list is a ready-made education in what AI is made of — Kate Crawford’s Atlas of AI on energy and extraction, C. Thi Nguyen’s “The Limits of Data,” Ted Chiang’s “ChatGPT Is a Blurry JPEG of the Web,” Narayanan and Kapoor on AI as “normal technology.” It is, deliberately, not about how to use any of it.

Open graduate syllabus; inputs / outputs / conflicts

What’s in the News

Oberlin launches a Critical AI Studies minor this fall. Beginning in fall 2026, Oberlin students can minor in the study of AI as a subject — analyzing, in the college’s words, “the ethical, cultural, environmental, political, economic, technological, and labor effects of AI.” The structure is telling: every student takes an Introduction to Critical AI Studies and a dedicated course on methods of critique, plus a foundational course in computer science or data science, before choosing electives across categories like “decision-making and learning” and “epistemology and the history of science.” It is explicitly interdisciplinary, bridging STEM and the humanities, and it complements majors from philosophy and anthropology to neuroscience and math. This is what an “about AI” curriculum looks like when a campus builds the whole thing rather than a single required module.

A field, not just a course, is taking shape. Oberlin is not alone. Standalone “AI and the Humanities” courses are now on the books at institutions including Leiden and the University of Chicago, and Edward Elgar’s Handbook of Critical Studies of Artificial Intelligence now gives the area a reference text. The vocabulary — critical AI studies — is consolidating around the idea that AI deserves the same kind of scholarly scrutiny we give any powerful social technology.

Institutional Movements

Two curricula, working in complement. The dominant institutional model this year answers the question can our graduates use AI? — a functional-competency requirement satisfied by discipline-specific courses. A newer model, at places like Oberlin, answers a second question — do our graduates understand what AI is? — with critique, history, and political economy. These are complements, not rivals — different organs of a healthy curriculum — and the encouraging trend is that campuses are starting to build both, and to notice how each strengthens the other.

💬 Editor’s note: A quick way to assess a course’s intended learning goals is to look at what it asks students to produce. A well-engineered prompt or a working tool signals teaching use; an argument about how a system was built and whom it serves signals teaching about. — TZ

Where the “about” tends to live. For now, the deepest “about AI” teaching is concentrated where it has natural homes — writing programs (via critical AI literacy), the humanities and social sciences (via critical AI studies), and a handful of prerequisite-free general courses willing to carry real technical content. The open question for the next few years is whether that understanding stays in those enclaves or becomes something every graduate is expected to have, the way basic scientific or statistical reasoning is.

Most Discussed

How do we grow the about alongside the how — so that fluency with AI arrives together with understanding of it? (Prompted by the critical-AI-literacy discourse and the maturing conversation about learning goals.)

Question for discussion. The functional case is strong: students will use these tools in their careers, so fluency is a real service. The maturing view adds a second aim rather than replacing the first — that alongside using AI well, students learn to ask whose data built a model, whose labor labeled it, and what its design optimizes for. The two reinforce each other: understanding tends to produce more discerning, more capable use, not less.

A good aspiration for any “AI literacy” offering: by the end, a student can not only get a good answer out of the system, but also explain how it got there, where that process is likely to fail, and what interests shaped its design. Where offerings reach the first today, the natural next step is designing them to reach the second too — and the materials to do that now exist.

Try This Week

Teach one hour about AI — not how to use it

Pick a single session and, instead of showing students a tool, help them interrogate one. Two ready-made, deep options:

Run Annette Vee’s evaluation exercise: paste the first paragraph of the Declaration of Independence into a model (or a head-to-head comparison site) and ask for a one-sentence summary. Have students compare the summaries against the original — most models drop “the Creator,” change “men” to “people,” and shift emphasis. The point is not the tool; it is watching students discover that a fluent summary encodes choices, and asking whose choices and why. Vee’s write-up links the full activity and a companion exercise on bias using Joy Buolamwini’s AI, Ain’t I a Woman?

Or assign one week of Henry Farrell’s Political Economy of AI syllabus — “Human Knowledge” is a strong single-session choice, pairing C. Thi Nguyen’s “The Limits of Data” with a chapter of Kate Crawford’s Atlas of AI. Then ask the one question that separates understanding from use: how was the data behind this system made, and who is missing from it?

aiX Weekly is curated by Claude and reviewed by Tian Zheng for the aiX Programs at Columbia University. Editorial notes reflect one statistician’s reading of the week and invite discussion. Corrections and suggestions are welcome.