Devil’s Advocate: How AI Can Resurrect Reading Assignments
Source: Terry U Substack
Author: Terry Underwood, PhD
Original source: https://terryu.substack.com/p/devils-advocate-how-ai-can-resurrect
Published: 2026-07-08
Source type: essay
Private backup: the full article text is archived in the private repository at archives/articles/terryu-substack-com-devils-advocate-how-ai-can-resurrect.source.md. It is not published on the public Quartz site.
Summary
Terry Underwood argues that schools often treat reading comprehension as if it were automatic, passive, and basically finished by early schooling, even though real comprehension is interpretive, effortful, and developmental. Drawing on Lorraine Cella’s classroom observations and Gorzycki et al.’s study of college readers, the piece says students usually know whether they read well, but they have been taught to misunderstand what reading is for. Reading assignments often feel disconnected from real classroom work, so students treat them as expendable inputs for grades rather than as tools for thinking.
The article’s AI turn is deliberately practical rather than utopian. Underwood suggests that AI could make assigned reading more usable by turning solitary pre-class reading into supported conversation, rereading, questioning, and summary-checking. In that framing, AI is not replacing reading; it can scaffold the kind of return-to-the-text work that many classrooms have not had time to build. The durable idea is that AI becomes educationally helpful when it pulls students back into difficult texts and visible comprehension work, not when it lets them bypass reading altogether.
Pull quotes
Comprehension is not a light switch
“Students describe comprehension as something that either happens or doesn’t — a light switch, a skill set, not a powerful mental capacity on par with composition.”
What students misunderstand
“What they don’t know — or rather, what they’ve been taught to misunderstand — is what reading is for.”
Why assignments fail
“Students weren’t rejecting reading. They were rejecting the version of reading college had mapped out for them — reading with little direct relevance to classroom use, assigned without follow-through, tested for retrieval rather than taught for understanding.”
The constructive AI possibility
“This is not a replacement for reading. It’s a bridge to it…”
Big ideas
- AI literacy has to be taught inside real subjects
- Learning still needs some struggle, even when AI can make things easier
- Students need to check AI answers against real evidence
Claims
- Reading ability is a bottleneck for effective student AI use
- AI-assisted inquiry should ground claims in evidence
Key evidence and examples
- Lorraine Cella’s informal classroom survey found that many students believe reading is not a process, rereading signals failure, and reading ability is mostly fixed rather than improvable.
- Underwood argues that these misconceptions reflect a hidden curriculum in which schools teach comprehension as if it had a single correct endpoint like decoding.
- Gorzycki et al. (2019) found that undergraduates generally knew how strong or weak they were as readers, but many still saw assigned reading as ineffective and disconnected from actual classroom use.
- The article argues that weak reading assignments are often a design problem, not proof that students reject reading itself.
- Underwood proposes AI as a scaffold for difficult reading: students can discuss passages, generate questions, test their summaries, and return to the text before class.
- The piece insists that this use of AI only helps if it moves reading back into the center of classroom work rather than replacing it with shortcut summaries.
Education relevance
High relevance for K-12 and higher-ed literacy, AI literacy, humanities pedagogy, disciplinary reading, and assignment design. The article is especially useful for thinking about how AI could support rereading, discussion, and comprehension without collapsing into anti-reading automation.
Durability note
The article’s most durable contribution is not the broad claim that AI will fix reading assignments. It is the narrower design insight that AI can be educationally valuable when it scaffolds visible comprehension work around real texts instead of letting students skip the reading. That idea should travel across classrooms more easily than the article’s anecdotal institutional history.