Inside Higher Ed
What, exactly, is AI literacy in 2026?
I ask because I keep seeing articles saying that employers want it, but I’d have a hard time saying what it is.
There’s the ability to feed data, whether numerical or prose, into an AI engine and ask it to find patterns. Then there’s the ability to identify hallucinations or errors in its output. Some would argue that awareness of the ecological impact of AI is a key part of the equation. On a more macro level, there’s legitimate concern about the impact of AI on our information ecosystem. What happens when AI engines start feeding on their own output? Veterans of the postmodernist wave will recognize that LLM’s treat language as a series of floating signifiers; it knows which words tend to go together, but it has no referent for any of them.
Worse, what happens when knowledge workers are replaced by AI altogether? (I think of that last possibility as analogous to the “recirculate” cycle in air-conditioning. It’s okay for a little while, but at some point, you need fresh air.) This is the death of the author foretold by the postmodernists, though not quite in the way they meant.
A few months ago, someone mentioned on Bluesky that to understand a technology, you need to know the problem it was meant to solve. AI is being used to solve for wages. It’s meant to do to knowledge work what automation did to manual work, with similar impact on employment.
But I don’t think that’s the AI literacy that employers have in mind.
From an academic integrity perspective, I have a concern that relatively advanced fluency in AI will result in much less fluency in literally anything else. Pushing through the friction of learning both builds skill and helps information stick; when information is handed to you on a platter, you don’t develop either the skill of learning or the knowledge base of information sinking in. Someone who’s good at using AI to fake knowledge in, say, a physics class, may come out knowing almost nothing about physics. That’s particularly true with agentic AI—bots—that can take entire online classes without the student even needing to bother feeding in prompts.
All of that said, I don’t think that channeling William F. Buckley’s famous posture—“Standing athwart history, yelling STOP!”—is an option. We need to prepare students for the world that exists and, to the extent possible, for the world that’s going to exist. Preparing them for the ’90s wouldn’t serve them at all.
In theory, I could imagine using AI for an almost Socratic form of model building. But the technology is moving so quickly that what might look like a useful approach now could be superseded in a few months, let alone a few years.
The obvious piece is error-checking. But in practice, that usually requires relatively broad and deep subject-matter knowledge, which is precisely what AI tends to prevent in people who don’t already have it.
Wise and worldly readers, have you seen (or thought of) a particularly useful definition of AI literacy? I’m looking for something we can use to ensure that our students are ready to take on the world, even if the folks who built the machine had something else in mind. As always, I can be reached at deandad (at) gmail (dot) com. Thanks!
