A professor’s objective is to generate graduate-level, analytical discussion questions on Moby Dick. Gemini generated the following output:
- “Here are your discussion questions:
- Who is the main character in Moby Dick?
- What is the name of the whale?”
The questions are fact-based recall and clearly do not meet the professor’s objective. What is the most effective and efficient next step the professor should take?
The most effective and efficient next step is to use the Gemini chat interface to provide a refinement prompt, explicitly asking the system to make the questions more analytical and appropriate for a graduate-level audience.
Generative AI models operate through an iterative dialogue process. If the initial output fails to hit the mark, starting over or abandoning the tool wastes valuable time. Instead, treating the first response as a draft allows you to issue course corrections. Providing clear, comparative guidance—shifting from simple factual recall to deep thematic analysis—tells the system to narrow its focus, adjust its tone, and pull from its training data on advanced literary criticism.
The alternative options create unnecessary friction or miss the core advantage of the technology:
- Searching the web for harder questions treats the model like a standard search engine, which only returns generic text fragments from other web pages instead of generating tailored prompts for your specific seminar.
- Discarding the chat entirely ignores the conversational memory of the workspace. Building upon a flawed answer often helps the AI understand what not to do, helping it pivot toward the correct style much faster.
- Moving the content into NotebookLM adds an unnecessary step. While NotebookLM is excellent for source-grounded exploration of personal files, the main Gemini app is fully capable of elevating its own reasoning and text generation through iterative prompting.
Providing a direct correction calibrates the system, giving you the rigorous, multi-layered questions your graduate students expect.