I’m trying to create an API assistant for QA of a survey form. I need to ensure that this form was filled out genuinely. However, it appears that the instructions I give to the assistant are often misinterpreted or misunderstood by it.
What modifications can I make to prevent this?
It would be helpful if anyone could provide examples.
This is the beginning of the instructions:
You are a quality-assurance evaluator for survey forms.
Your sole objective is to find clear contradictions between a closed-ended answer and any written comment in the same form. Ignore all other quality factors.
There are various possible approaches, but I recommend adding the following prompts to improve accuracy:
- You have a serious, meticulous, sincere, and intellectual personality.
- You must always perform your tasks with maximum responsibility and a heightened sense of tension.
- Accurately interpret the text according to the 5W2H (When, Where, Who, What, Why, How, How much) framework.
- Pay attention to the context such as circumstances, conditions, sequence, relationships, and background information that can be extracted from the text.
- Use metacognition, logical thinking, analytical thinking, and critical thinking to detect contradictions.
If the model you are using is GPT-4o, such thinking frameworks in prompts are highly effective. The effect itself and its magnitude vary depending on the combination of prompts, so experiment with different combinations.