Using a Large Language Model (LLM) to refine hypotheses

The following assignment is a group assignment in a graduate level geospatial course.  

Details

This week your team will work through a structured approach to hypothesis refinement using human-centered methods, LLM-centered methods, and human-machine collaboration. By following these steps, you will critically engage with both human reasoning and AI-assisted insights to enhance your understanding of the capstone scenario.  

Step 1: Human-Centered Analysis

As a group, brainstorm responses to the following questions. Do not use AI/LLMs for this portion of the assignment.

1. Identify your assumptions.
2. Define your research question.
3. Generate Hypotheses.
4. Identify any limitations that your team faces in testing these hypotheses.  

Step 2: AAG-Centered Analysis

Now that your team has developed initial hypotheses, use a Large Language Model (LLM) (e.g., Copilot, ChatGPT, Claude, Gemini) to generate additional hypotheses to answer your research question.

In your response, include the prompts used and a summary of the hypotheses the LLM generated.  

Step 3: Human-Machine Collaboration and Evaluation

Compare the human-generated hypotheses from Step 1 with those generated by the LLM in Step 2.

1. Critique the LLMs Results
   • Are the AI-generated hypotheses relevant and well-structured?
   • Do they introduce new perspectives your team hadn’t considered?
   • Are there any biases, inaccuracies, or overly generic statements in the LLMs responses?
2. Compare & Evaluate Hypotheses
   • How do the human-generated hypotheses compare with those produced by the LLM?
   • Are there overlapping themes or significant differences?
3. Refine Your Hypotheses
   • Based on your evaluation, refine your hypotheses by integrating the best elements from both human and AI-generated insights.