Few assigned texts generate as much uneven writing as Joan Didion's Slouching Towards Bethlehem. Some students summarize the Haight-Ashbury scenes and never venture an interpretation, while others leap into sweeping claims about the 1960s that the text cannot support. A teacher facing 120 of these essays has to separate careful reading from confident vagueness, and that sorting is slow when done by hand. AI-assisted grading gives a first pass that flags those differences consistently.

The collection is difficult to grade because its power comes from tone, selection, and restraint rather than from plot. Students often describe what happens in the title essay without noticing how Didion arranges fragments, quotes her subjects flatly, and withholds obvious judgment. A grader has to recognize when a student is actually analyzing that technique and when the student is only retelling scenes. Those judgments are subtle enough that fatigue on the sixtieth paper changes the scores.

Many teachers respond by narrowing the prompt so the essays become easier to compare. That helps, but it also trims the interpretive range that makes the unit valuable in the first place. A better approach is to keep the prompt open while grading against criteria that stay stable, such as the quality of the claim, the precision of textual evidence, and the explanation connecting the two. Consistent criteria let an open prompt produce comparable scores.

What a first-pass AI review can actually do

An AI grader working from your rubric can read every essay against the same standard and report where each one lands on each criterion. It does not tire, and it does not soften its judgment because the previous essay was weak. Teachers then spend their time on the papers where a human eye matters most, such as the borderline argument or the student who took an unusual but defensible reading. The result is a faster process that still ends with the teacher's decision.

• Whether the thesis makes an arguable claim about Didion's technique rather than restating the topic
• Whether quoted material is short, accurate, and followed by real analysis
• Whether the essay explains why a detail matters instead of only naming it
• Whether the conclusion extends the argument or repeats the introduction
• Whether sentence-level clarity holds up in the more complex paragraphs

Setting up your rubric before you grade

The quality of automated feedback depends heavily on how specific your criteria are. A line such as "strong analysis" tells an AI very little, while "explains how Didion's selection of concrete details shapes tone" gives it something to check. Spend twenty minutes writing performance descriptions for each level before the first essay is scored. That investment repays itself quickly, because every later paper is measured against language you chose deliberately.

It also helps to include a short note about what the assignment is not asking for. If you do not want biographical summary of Didion or a general history of the counterculture, say so in the rubric. Students frequently fill space with background because it feels safe, and that padding can inflate scores if the criteria are loose. Naming it as a low-value move keeps the focus on interpretation.

Reviewing the results with professional judgment

After the first pass, sample a handful of essays from each score band and read them yourself. If the high band contains papers you would rate lower, adjust the wording of the rubric and rerun the batch. This calibration step usually takes one round, and it builds real confidence in the output. Teachers who skip it tend to distrust the results, while teachers who do it tend to rely on them appropriately.

Keep the human role clear as well. Comments on voice, risk-taking, and unusual readings deserve a personal reply, because those are the moments when a student learns that someone noticed. AI can handle the repeated observations about evidence and organization so that your written attention goes where it changes the writing. That division of labor is what makes a large-class Didion unit sustainable.

Returning feedback students will use

Feedback only helps if students can act on it, so each comment should point to a specific revision. Compare "needs more analysis" with "after the sentence about the runaway teenagers, explain what Didion's flat reporting suggests about her attitude." The second version gives a student a next step, and it also shows what close reading looks like. Comments written at that level of detail are much easier to produce when the first draft of them comes from a tool.

Return the results while the reading is still fresh, ideally within a week of submission. Students who receive concrete comments quickly are far more likely to attempt a revision or to carry the lesson into the next essay. Faster turnaround is one of the clearest practical benefits of AI-assisted grading for a text this demanding. It lets the unit build on itself instead of ending at a grade.