Close Reading of Complex Literature: What About AI?

Jan 24, 2025

English teachers assign interpretive essays about literature to teach students to read closely and notice meaningful details, to construct evidence-based arguments, to develop original insights about texts, to engage in academic discourse about literature, and to think critically about challenging material.

The Pedagogical Problem

Recent research by Chang, Cramer, Soni and Bamman (2023) reveals a significant problem in the offing. Their study demonstrates that AI algorithms have effectively "memorized" most commonly taught texts through exposure in their training to countless online analyses.

Their findings reveal clear patterns in what AI has "memorized." The most thoroughly memorized texts include public domain classics (like "Alice in Wonderland" with 98% accuracy), science fiction/fantasy bestsellers (like "Harry Potter" at 76% accuracy), popular contemporary fiction (like "The Hunger Games"), and—guess what—books frequently taught in schools.

This memorization correlates strongly with web presence, drawing from book review sites like Goodreads and Amazon, study guide websites like CourseHero, Quizlet, and LitCharts, academic sites like academia.edu, archive sites, and student help sites. Crucially, the researchers found much lower memorization rates for works by Black authors, Global Anglophone texts (English-language works from outside US/UK), and less commercially successful literary works.

Through systematic testing using a "name cloze" method, they found that "for some commonly taught texts, GPT-4 can correctly identify specific characters from context over 70% of the time, even when given minimal information." More importantly, they discovered that "ChatGPT and GPT-4 have seen books in proportion to their popularity on the web, and their performance on downstream tasks is tied to that popularity." In other words, AI can generate sophisticated literary analysis of standard curriculum texts while potentially marginalizing diverse voices and perspectives.

Commonly proposed solutions are likely to fail. Banning AI use is largely unenforceable and doesn't address the underlying challenge. Requiring personal connections to the literature can be easily circumvented by having AI generate the analysis and students simply sprinkle in a personal anecdote or two. Moving to in-class writing doesn't develop the sustained thinking skills teachers want students to gain.

Instead, teachers might better serve their learning goals through several alternative approaches. The interpretive process can be made visible by having students document their understanding as they read over time, requiring them to explanation how they arrived at their insights and valuing expressions of uncertainty and confusion as part of authentic interpretation.

Valuing originality and nuances of thought over polish and paragraphing can be done by creating assignments that reward noticing new things in texts, encouraging challenges to standard interpretations, and valuing rough but genuine insights over annotated conventional analysis.

Building interpretation through dialogue is crucial. This means using class discussion to develop ideas collaboratively, having students respond to and build on each other's observations, and creating interpretive communities where ideas evolve through interaction. We

Focusing on metacognition by having students reflect on how their understanding changes as they read and reread is crucial. Documenting intentional rereading for metacognitive conversations is a powerful strategy. Semantic revision is significant both in reading and writing, but traditional close reading places a premium on finding a pattern and pursuing it. Asking students to analyze their own interpretive processes and valuing growth in thinking over final products leaves little room for bot work.

These approaches align with what makes human interpretation valuable: The personal struggle with uncertainty, the experience of discovery, the courage to challenge conventional readings, and the satisfying evidence of growth over time. As the researchers note, AI excels at reproducing patterns but cannot experience these aspects of human interpretation.

The goal isn't to make assignments "AI-proof" but rather to revitalize the elements of interpretation that are inherently human. This might mean fewer traditional analytical essays and more documenting and reflecting on the interpretive process itself. Through these approaches, we can maintain the core value of literary analysis while acknowledging and adapting to the realities of AI capabilities.

Judith Langer’s Model

Drawn from Langer's influential research studying students while reading literary texts, teachers can base pedagogy on four key stances readers take toward a text as they develop understanding. These stances are recursive rather than linear, meaning readers move between them throughout their reading process.

The first stance is "Being Out and Stepping Into an Envisionment." Interestingly, Langer chose a physical analogy to express the sensation of breaking into a text and finding a path through the linguistic wilderness. When we “envision,” we develop a scenario which will unfold; since this stance is a non-negotiable part of reading (each new chapter, each new spell of reading) it’s worth rehearsing and refining.

In this initial stance, readers try to make contact with the text world using their knowledge, experiences, and surface features of the text. For example, when reading "I See You Never," the text Langer chose for her study, students would notice character names like Mrs. O'Brian and Mr. Ramirez, try to picture the setting of a California boarding house, and draw on any knowledge they have about immigration or landlord-tenant relationships. As Langer explains, this stance is about "gathering information and associations in an attempt to build text world."

The second stance is "Being In and Moving Through an Envisionment." Here, readers are immersed in their developing understanding, using both their growing sense of the text world and their personal knowledge to build deeper meaning. In Langer's study, students reading "I See You Never" moved beyond just identifying characters to understanding the complex emotions and relationships - noticing Mrs. O'Brian's mixed feelings as a landlady and authority figure who also cares about her tenant. Readers in this stance are actively "using local envisionments and personal knowledge to build and elaborate understandings."

The third stance is "Stepping Back and Rethinking What One Knows." This is where readers use their growing understanding of the text to reflect on their own knowledge and experiences. For example, students might use their understanding of Mr. Ramirez's situation to rethink their own assumptions about immigration or consider how they've experienced saying difficult goodbyes. As Langer notes, readers in this stance "use growing understandings to rethink previously held ideas, beliefs, or feelings."

The fourth stance is "Stepping Out and Objectifying the Experience." In an odd way, this stance hearkens back to the distinction literary critics have traditionally drawn between close reading (seeing into the text) and distant reading (looking away from the text).

Here, readers distance themselves from their immersion in the text to analyze their understanding, the text as an object, or their reading experience. They might evaluate the author's craft, consider the larger themes, or reflect on how the reading experience affected them. This stance involves the reader who "distances self from text to examine, evaluate, or analyze the reading experience or aspects of the text."

What makes these stances particularly valuable is that they aren't sequential steps but rather different ways of engaging that readers move between as part of the reader’s craft. A reader might start in the first stance, move to the second, jump back to the first when encountering something confusing, move to the third when making a personal connection, and so on. This recursive movement between stances allows for rich, multifaceted understanding to develop.

Langer's research suggests that recognizing and supporting these different stances might be more valuable than focusing solely on final interpretations. This aligns well with addressing the challenges of AI in literature classrooms - while AI can generate polished analytical products, it cannot engage in this dynamic process of moving between stances to build understanding.

Stance Activities

Classroom activities that can help students develop a personal repertoire of strategies for each stance while clarifying engagement with literature:

For "Being Out and Stepping Into an Envisionment" have students keep a running record of their initial questions, confusions, and attempts to orient themselves as they begin reading. These could track questions about characters they're meeting, responses to unfamiliar words or references, initial predictions or assumptions, places where they need more information.

For "Being In and Moving Through an Envisionment" create opportunities for students to document how their understanding evolves. For instance, stop periodically to have students write or talk about how their view of a character is developing, ask them to note moments when they changed their mind about what's happening, have them track how their questions from their initial stance are being answered or complicated.

For "Stepping Back and Rethinking What One Knows" design reflection prompts that help students recognize when the text is causing them to reconsider their own experiences or assumptions: "What has this text made you think about differently?" "How has your understanding of [theme/issue] shifted?" "What experiences of yours look different after reading this?"

For "Stepping Out and Objectifying the Experience" create opportunities for analytical distance through activities like discussing author's craft choices and their effects, comparing different readers' experiences with the same text, evaluating different possible interpretations.

The key, according to Langer, is that these activities shouldn't be treated as sequential steps but rather as different ways of thinking that students can move between. A class discussion might start with students in the second stance (developing their understanding) but move to the third stance when someone makes a personal connection, then back to the first stance when that raises new questions about the text.

This approach helps address the AI challenge because it emphasizes the dynamic process of understanding rather than just the final product where AI shines. The value lies not in arriving at a polished interpretation but in documenting and reflecting on how understanding develops through these different stances.

Assessment

Traditional rubrics focused solely on final written products miss the most important evidence of student learning to read literature: their movement between different stances of understanding. Here's how we might assess student engagement with literature differently.

Instead of just evaluating a final essay, teachers might assess students' documented journey through the stances. For example, a student portfolio might include evidence of Initial Engagement (Being Out and Stepping In): preliminary questions and confusions, early attempts to make sense of characters or situations, initial hypotheses that may later be revised.

Evidence of Developing Understanding (Being In and Moving Through) can include their tracking textual material that deepened their understanding of a symbol, an event, a character; moments when they revised earlier assumptions; growing awareness of complexities in the text.

Evidence of Personal Connections (Stepping Back and Rethinking) can be found in student reflections on how the text changed their thinking or caused them to rethink an experience; intertextual connections or links between known people and fictional characters; recognition of how their perspective shifted.

Evidence of Analytical Distance (Stepping Out and Objectifying) can be assessed through informal writing or oral presentations to express to and support different interpretations; analysis of author's craft; reflections on their own reading process.

What makes this assessment approach powerful is that it values uncertainty and privileges the courage to suspend closure and to revise, and growth over polished final products. A student who wrestles with meaning, even if their final interpretation isn't as sophisticated, might demonstrate more learning than one who produces a polished but formulaic analysis.

Literature as a Human Project

Judith Langer’s model of envisionment building does not suggest that argumentative essays positioning the reader as critic should not be used. On the contrary, one experience as a literary critic (vs a literary reader) per semester as a culminating project could be more impactful than many separate assignments.

To implement this approach needs the practical expertise of teachers. The approach can’t be ordered from a curriculum shop. How could this kind of assessment work within your grading requirements and time constraints? What aspects would need to be modified to be realistic in your classroom? Though AI has crashed the classroom without invitation, look at it this way: It could be an opportunity to move learning and literature to higher ground.

Learning to Read, Reading to Learn

Discussion about this post