Designing Writing Assignments to Work with AI
With the growing use of AI, writing instructors are looking for innovative ways to incorporate ethical and productive use of the technology while still fostering the foundational skills of reading, writing, and critical thinking in college-level writing courses. It is critical for students to understand the limits and shortcomings of AI-produced writing while learning how to harness its capabilities to strengthen their own work. As AI models become readily accessible, writing instructors have an opportunity to revisit the purpose of writing instruction and adapt their assignments to help students understand the significance and function of writing in the AI age.
Here are a few suggestions for creating assignments that encourage students to interact with AI thoughtfully.
Generate ideas and outlines. Ask students to generate three ideas or outlines for a writing assignment and evaluate each for accuracy, creativity, and comprehensiveness. Students will observe how AI can support the brainstorming process while recognizing its limitations through the variance across the outlines it produces.
Revise with AI. Ask students to submit a draft to a generative AI tool for revision, then evaluate the edits it suggests. In more advanced versions of this assignment, students can experiment with precise prompts for tone, audience, and clarity, then analyze how well different prompts achieved the assignment's goals. Students will practice receiving AI feedback and developing the judgment to decide what's actually useful.
Critique AI responses. Feed a writing prompt to a generative AI tool and have students critique its output. Students will practice evaluating AI-written text critically, identifying its strengths and weaknesses.
Fact-check AI sources. Ask students to request sources from a generative AI tool for a writing assignment, then cross-check each one against the UCLA Library catalog (or a database like WorldCat for material outside UCLA's holdings). Have them attach screenshots of both the AI output and the corresponding catalog record. A more advanced version asks students to request direct quotes from a specific source, then verify the quotes against the original text. This assignment surfaces real issues of credibility in AI-generated information and builds awareness of AI "hallucinations."
Craft diverse prompts. Ask students to experiment with language, tone, and structure across several prompts and compare the outputs. This trial-and-error process shows students how prompt construction shapes results, building rhetorical awareness they can apply to their own writing.
Mix genres. Ask students to use AI to rewrite an academic argument as a news article, poem, or dialogue. This helps students recognize genre conventions and how those conventions shape meaning.
Disclose AI use. Ask students to document where and how they used AI in a given assignment — which tools, for what purpose, and how the output was revised or incorporated. This builds the habit of transparent, ethical AI use and gives instructors a window into students' process, not just their product. UCLA instructors can point students to the UCLA Library's guide on AI and academic integrity for institutional context on disclosure norms.
Underlying all of these assignments is a broader goal: helping instructors and students understand how large language models actually function, and how that differs from human creativity. LLMs generate text by predicting likely patterns from the material they were trained on; they don't draw on lived experience, intention, or a stake in what they're saying the way a human writer does. Keeping that distinction in view( the "functional" fluency of an LLM against the situated originality of human writing) gives students a clearer sense of what AI is actually good for, and where the real work of writing still belongs to them.