Prompt strategy
The prompts we are about to use provide high-level guidance for Copilot, and the outputs/results allow further tailoring of Copilot’s responses to match the specific dataset and analysis needs.
The key aspects of the prompting approach are:
- Define the task. Clearly instruct the AI assistant what task we are solving.
- Break down into steps. Breaking the data exploration down into logical steps (like data loading, inspection, summary stats etc.)
- Providing context/intent for each prompt to guide Copilot (like requesting numeric summary statistics)
- Sharing previous results as input. Sharing outputs and results from Copilot’s code snippets to further guide the conversation (like printing the summary stats)
- Refine, iteratively refining prompts and conversing with Copilot in a back-and-forth way
Therefore, we will use the TAG (Task-Action-Guidance) prompt pattern described in Chapter 2. Let’s describe this...
