AI as Reader: An Experiment in Artificial and Human Interpretation
Faculty Reflection
This assignment transforms the typical relationship students have with generative AI. Instead of using AI to write papers, students make AI itself the object of a literary-critical experiment. They investigate how AI reads literary texts, how its interpretations are influenced by their questions, where it makes mistakes, and what they, as human readers, can do better. The assignment is structured around a central pedagogical insight: the best way to understand the limitations of AI is not to prohibit its use, but to engage with it critically and systematically. Students compare classical German texts (Goethe, Schiller, Winckelmann) with texts exploring the "pathological conception of humanity" (Kleist, Hoffmann, Moritz), examining whether AI can meaningfully distinguish between these different intellectual traditions. The emphasis throughout is on process over product. Students document their thinking at every stage—from their initial AI-free readings through multiple rounds of prompting, fact-checking, and reflection. They learn that interpretive work is not about information retrieval but about questioning, comparing, and revising. The assignment culminates in a reflection connecting Kant's Enlightenment imperative—"Have courage to use your own understanding!"—to the challenges of thinking critically in an age of generative AI. Assessment focuses not on producing perfect interpretations but on making visible a demanding intellectual process. A poor AI response becomes excellent material if students can explain why it's problematic, how they discovered that, and what they did in response.
What Students Create
Students produce a Process Portfolio documenting their investigation of how AI interprets literary texts. They compare their own readings with AI-generated interpretations across two German texts—one from the Classical period and one exploring pathological humanity. The portfolio includes initial AI-free readings, annotated AI responses, a prompt log tracking how question changes affect interpretation, fact-checked AI claims, comparative reflections, and a self-designed experiment testing AI's ability to distinguish between different conceptions of humanity.
Why This Assignment Matters
This assignment positions AI as an object of critical inquiry rather than a writing assistant. Students develop essential skills for the AI age: distinguishing between plausible interpretation and factual error, recognizing how AI responses depend on question framing, and understanding their own interpretive work as a process that goes beyond information retrieval. By systematically testing AI's limitations with literary texts, students learn to use their own understanding—practicing Enlightenment thinking in a new technological context.
How It’s Graded
Assessment Criteria (100 points total)
1. Quality of the Process (30 points)
27–30 points: The work process demonstrates genuine exploration. Prompts are deliberately changed; the student challenges the AI, tests assumptions, follows unexpected results, and documents how the investigation develops.
23–26 points: Several meaningful iterations are present, but the investigation remains somewhat predictable.
19–22 points: The AI is used primarily as a question-and-answer system; there is little genuine iteration or change in the student's own strategy.
0–18 points: The portfolio documents very little traceable process.
2. Close Reading and Verification (25 points)
23–25 points: AI claims are checked precisely against specific passages from both parts of the seminar. The analysis convincingly distinguishes between plausibility and textual evidence.
20–22 points: Good textual analysis, with occasional superficial justification.
17–19 points: Textual evidence is present, but it is used more to confirm claims than to genuinely test them.
0–16 points: Judgments about the AI are rarely supported by the primary text.
3. Critical Reflection on AI (20 points)
18–20 points: The reflection develops nuanced insights into the possibilities and limitations of AI and derives them from the student's own experiments. Differences between the two texts and the two parts of the seminar are used productively.
15–17 points: Good reflection, but at times general or only loosely connected to the experiments.
12–14 points: Primarily general statements about AI.
0–11 points: Little critical engagement with the student's own use of AI.
4. Development of Your Own Interpretation (15 points)
14–15 points: The portfolio clearly shows how the student's own interpretation of both texts changes, deepens, or is deliberately confirmed over the course of the process. The comparison between Part I and Part II leads to new insights.
12–13 points: Development is clearly evident.
10–11 points: The original and later interpretations remain largely disconnected from one another.
0–9 points: Little development of the student's own interpretation.
5. Documentation and Transparency (10 points)
9–10 points: Complete, clear documentation. The development of the prompts and the student's own decisions is fully traceable.
8 points: Minor gaps.
6–7 points: Essential parts of the process are only partially documented.
0–5 points: The process by which the work was produced is difficult to reconstruct.
Basic Principle of Assessment
I am not evaluating how well you can operate the AI. I am evaluating how well you read, ask questions, check, compare, and reflect on your own thinking.
A perfect final product with an invisible process of creation is therefore worth less than an imperfect result that makes a demanding intellectual process visible.
The Ultimate Goal
The ultimate goal of this assignment is not to determine whether AI is intelligent.
The more interesting experiment is:
What do you learn about your own reading and thinking when you compare it with the reading and thinking of an AI?