The AI Conversation Still Missing In Higher Ed

Most campus conversations about AI are organized around academic integrity: how to think about authorship, assessment, and what it means for students to do their own work in a world of capable models. Those are serious questions, and they deserve the attention that they're receiving.

But after a recent WCET talk that I delivered to digital learning leaders, instructional designers, online learning directors, and compliance staff, I left thinking about a parallel question that has gone missing next to the integrity conversation — one that very few institutions seem to have delegated to anyone.

Students aren't only using general-purpose AI to write essays. They're using it for the hard, human stuff: anxiety, loneliness, a decision that they're stuck on. They're bringing it to general-purpose tools on their own accounts, and to the AI features being built into the platforms that institutions already deploy, although the evidence on that second pathway is still emerging.

This isn't only a higher ed story. The same pattern is surfacing beyond campuses: as large employers roll out AI to their workforces, employees are using those sanctioned tools for the hard human stuff too, and the organizations are just as unsure who owns the well-being side of it. The thread connecting both is that the technology arrived faster than anyone decided who's responsible for its effects on the people using it.

Much of this use is invisible. Recent national data found that many young people now turn to AI for mental health support, and most of them don't tell anyone — not a parent, friend, or clinician.

When you raise this with the people who shape digital learning environments, the most common reaction isn't disagreement. It's recognition, often for the first time, followed by a question: whose job is this?

Who owns the well-being side of campus AI?

On most campuses, AI policy doesn't live in one place. It's scattered, sometimes in the library, sometimes in a center for teaching and learning, sometimes in an online learning division, sometimes in individual colleges, sometimes in a senior leadership committee. Each owner has a legitimate stake in the AI question. But the well-being dimension of AI — what happens when students use these tools for emotional support — tends to fall into the gaps between them. It's partly that no one is sure that it's theirs, and partly that many haven't registered it yet as a question at all.

This isn't a failure of any one office or officer; it's structural. The use case has emerged faster than the org chart could absorb it, and the result is a question that nobody has formally taken up or been handed.

The wellness and student success leaders often aren't in the room

The people with the most relevant expertise, the well-being and student success leaders on campus, are frequently absent from AI decisions. The AI conversation has been framed as academic and technological, so it convenes academic and technological people. Meanwhile, students are using these tools in ways that reside in the wellness or student success team’s domain, often without that team having any visibility into it.

Because this use is already happening, those leaders should be involved, not only at the table when institutions decide what AI to deploy, but actively educating other stakeholders, shaping what students are taught about these tools, and making recommendations about how the institution responds when a student in distress turns to a general-purpose or academic AI tool instead of a person or a tool intentionally designed for wellness help.

Bridge, not destination

There's a design question underneath all this. General-purpose AI tools are often optimized to keep you engaged, to be agreeable, available, and endlessly willing to continue. For a student leaning on the tool through distress, it can quietly become a destination that substitutes for the people and resources around them, rather than a bridge toward them. In research with my Stanford Medicine collaborator, I've argued that the same features that make these tools useful can also maintain anxiety and related patterns rather than resolve them. Anxiety is one of the most common wellness challenges on campus, which is part of why this matters here.

The embedded tools that institutions buy raise a related but distinct issue: they weren't designed for emotional support in the first place, so they aren't built to recognize when a student needs more than the tool can offer, or to route them anywhere when they do.

This is the lens that I'd bring to any AI that touches student well-being, including tools that an institution provides itself. The goal is to help the student build evidence-based skills, make real-life connections, and get to human support as needed: the counseling center, a peer program, an advisor, a club, and to actually land there.

What to do with this

If you're a student success or well-being leader, get into the AI conversation on your campus (and if you're not, flag it for the colleagues who are!). As the people who understand student well-being and success, you should be helping to shape decisions about the tools that affect these areas, and educating everyone else in the process.

Treat student use of AI for emotional support as in-scope, part of the campus environment, not an out-of-scope behavior happening somewhere else. That one reframe changes what makes sense to put in place.

Build the well-being dimension into AI literacy: why a general-purpose model tends to tell students what they want to hear, why that can quietly reinforce distress, and that personal disclosures to commercial AI may be retained and used in ways that students don't expect.

And consider where fit-for-purpose tools fit. General-purpose AI and the academic tools that institutions buy weren't designed for emotional support, although students may use them that way. Purpose-built well-being tools, designed to complement rather than replace those tools and campus services, and built to route students toward campus-specific resources, can fill that gap.

The institutions that get ahead of this won't be the ones with the best chatbot policy. They'll be the ones that increase awareness around this missing dimension, and make sure that the right people are in the room to answer it.

References

Golden, A., & Aboujaoude, E. (2026). A transdiagnostic model for how general purpose AI chatbots can perpetuate OCD and anxiety disorders. npj Digital Medicine, 9, 343. https://doi.org/10.1038/s41746-026-02531-7

McBain, R. K., Cantor, J. H., Breslau, J., et al. (2026). AI chatbot use and disclosure for mental health among US adolescents and young adults. JAMA Pediatrics. Advance online publication. https://doi.org/10.1001/jamapediatrics.2026.2015