Providing context sets the stage and guides the GAI towards a more accurate and relevant response that suits your needs. Providing context is part of pre-prompting.
When prompting a GAI, you likely have a sense of what would constitute a successful response. Context helps the GAI generate such a response.
Context can take various forms, making it challenging to pinpoint necessary information. However, it becomes evident when a prompt lacks sufficient context.
"I am writing a paper on climate change for a high school science class. Can you explain the greenhouse effect and its impact on global warming, suitable for a student audience?"
Why It's Good: The prompt provides context about the audience (high school students) and the specific topic (greenhouse effect and its impact on global warming), guiding the response to be educational and appropriate for the specified level.
"Tell me about climate change."
Why It's Bad: This prompt lacks context. Without specifying the audience or the specific aspect of climate change, the response could be too broad or not aligned with the user's needs.
Consider the following aspects in your prompts:
- Background information
- Topic delimitation
- Intended audience
If you are unsure where to start, think about five W's and one H: what, where, who, when, why and how?
The more context you give, the more you can narrow the scope of the response. However, overly lengthy prompts can also be problematic. Keep context clear and concise.
Prompt the GAI with a vague prompt without context. Provide more context iteratively and evaluate the responses. Make your own prompt or start with the bad example above. When is the context sufficient?
Reflect on why context is crucial for GAI from a technical perspective. Consider what information is critical and what might be superfluous for the GAI to generate accurate responses.