As always, I had planned to discuss two articles this week, but then I got swept up after reading just one, and my plans changed. ASU's quiet rollout of an AI course builder that pulls from faculty content without their knowledge raises questions large enough that I decided to give it the whole issue rather than splitting attention. We'll spend the deep read here, then close with three Sparks worth your time on AI anxiety among students, persistent myths in faculty AI conversations, and the Department of Education's final word on graduate loan caps. 

AI Course Creator 

From Faculty Concerned About ASU’s ‘Frankensteinian’ AI Course Builder | Inside Higher Ed 

ASU has built a teaching machine based on their own faculty’s instructional work. 

Our Thoughts  

Oh, ASU…you’re breaking my heart. Why you would invisibly use faculty work to build an AI learning module creator is beyond me. Don’t you know that people are already questioning the value of higher education? I don’t mean to call out ASU; they’ve done some amazing things for higher education. But a chatbot built in secret from the instructional materials of their faculty is probably not one of them.  

Let’s start with the lack of transparency around this tool. First, faculty didn’t know their content was being remixed via this AI app. While ASU’s intellectual property policy probably allows the institution to do this, a better path might have been to follow the historical norms of use by informing and collaborating with the faculty before their face ends up in a product with a five-dollar monthly price tag. ASU broke that historical balance by not going through faculty governance. Given the pressures higher education is facing, they may not be the last institution to try. Second, why all the secrecy around the model underpinning Atom? If it’s Anthropic’s Claude, that’s a fine choice given their recognition in the market as having a strongly performing LLM. Finally, and maybe this is specific to Inside Higher Ed, but why did the ASU spokesperson decline to answer any questions from the respected news outlet? Instead, we simply get a non-answer statement. Each of these was a choice that ASU made, and together, they paint a picture of an institution that perhaps isn’t too thrilled about Atom being known about publicly. What does that say?  

Regular readers know that I’m finishing my doctoral work in teaching, learning, and technology, so the questions Atom raises aren't abstract for me. Audrey Watters' Teaching Machines documents more than a century of attempts to automate instruction, beginning with Sidney Pressey's 1920s mechanical testing devices, running through B.F. Skinner's behaviorist programmed learning, and arriving at today's adaptive platforms. The pattern is consistent across every generation. The technology gets sold as "personalized learning,” but on closer inspection, it treats the student as a container to be filled and the teacher as either an obstacle to be removed or a content source to be extracted from. 

Paulo Freire named this more than fifty years ago in Pedagogy of the Oppressed (whose bright red cover is staring at me from my bookshelf). He called it the banking model of education, where knowledge becomes deposits that an authority figure transfers into a passive student. Freire's alternative was dialogic education, where teacher and student build understanding together through questioning, conversation, and shared inquiry. The teacher matters in that model as does the cohort, the relationship, and the trust. Atomic is the banking model rebuilt for an AI economy. It strips the teacher out entirely, takes their content as the deposit, and delivers it to a paying user with no dialogue and no relationship. Calling it "personalized" doesn't make it personal. It makes it efficient.  

Watters' larger argument is that every previous attempt to automate teaching failed not because the technology wasn't good enough but because students learn better when there's a human at the center of it. The teacher matters. The dialogue matters. The cohort matters. None of those things show up cleanly in efficiency metrics. It’s hard to measure the impact of a good classroom conversation that impacts a student and helps them grow as a learner and as an individual. When institutions look for cost savings, these are often the first to go via larger class enrollments or increasing adjunctification. This is also why each new wave of teaching technology arrives promising to replace what couldn't be replaced last time. Humans are expensive; robots considerably less so.  

Additionally, I think we might be dramatically underestimating the impact this would have writ large. Public confidence in the value of a college degree is at historic lows. If higher education's response to that crisis is to make itself look more like a content delivery platform, we've conceded the only ground we have left to defend. The thing postsecondary institutions do that nothing else does is human teaching and learning at scale. When we replace that with a chatbot that scrapes our own faculty's work to generate five-dollar modules, we dismantle our own value proposition. We do the work of our opposition for them. 

For example, let's look at Ostling's concern about decontextualization. He worries that bad actors will use Atomic to generate out-of-context clips of faculty teaching about race, gender, or conflict in the Middle East, and use those clips as ammunition for harassment campaigns. While it may seem like paranoia, it isn't. Last summer, the Oversight Project, a Heritage Foundation spinoff, filed public-records requests for syllabi from roughly 70 UNC-Chapel Hill courses, hunting for terms like DEI and LGBTQ+. Faculty whose materials surfaced faced sustained online harassment; one professor eventually deleted his X account. A platform that takes faculty teaching out of the framing the faculty member built around it creates the conditions for that harm whether ASU intended it or not. And that's not a feature you can patch out later with a software update. The harm has already been done. 

While I truly hope it doesn’t happen, I expect to see this model replicated at other large institutions with the infrastructure to build it and externally through vendors who'll pitch ASU-style products to provosts and CIOs over the next eighteen months. Sadly, most of those vendor versions will offer less transparency than Atomic, not more. I guess the question isn’t really whether this is coming. The real questions are whether your faculty intellectual property policy and your Canvas terms of use allow this. If you don’t know the answer to those, I’d recommend starting the conversations now before the academic year ends and not wait until the Fall. AI development is rapid; don’t give it a summer head start. 

Sparks 

• College students are changing course in search of 'AI-proof' majors. But no one knows what they are (Associated Press) - Students are switching majors in real time over AI fears, and the advisors they'd normally turn to don't have answers either. Brown's Christina Paxson admits as much, suggesting that "the fundamentals of a liberal education are probably more important than learning how to code in Java right now." If you work in advising or career services, this one is worth reading carefully; the students profiled here are the conversations you'll be having all summer. 
• 5 AI Myths and Why We Must Move Past Them (Inside Higher Ed) - Two faculty members take aim at the outdated AI advice still circulating on teaching and learning websites, including the persistent myth that you can reliably spot AI-generated writing. A useful gut check for anyone whose campus is still running professional development built around assumptions that stopped being true a year ago.
• ED Rejects Call to Expand Access to Higher Grad Loan Caps (Inside Higher Ed) - The Department of Education finalized its rule on graduate and professional loan caps, sticking with the narrow list of 11 programs eligible for the $200,000 limit. Advanced practice nursing, physical therapy, social work, architecture, and dozens of other licensed fields land in the lower $100,000 tier. The rule takes effect on July 1, so graduate enrollment managers and financial aid offices have a busy summer ahead.