Common pitfalls and solutions: Newbies must know the design errors of prompts
The Assumption Bias Trap
When AI tells you only what you want to hear
Trap symptoms:
- The prompt contains an obvious position or tendency
- The information obtained always supports a specific point of view
- There is a lack of presentation of opposing or different points of view
Coping strategies:
- Self-examination: Reflect on your own possible biases when designing prompts
- Neutral language: Avoid including biases or preconceived positions
- Multi-perspective analysis: Explicitly ask for different perspectives
- Critical thinking: Be vigilant about AI output and cross-validate
The Hallucination Trap
When AI Confidently Talks Nonsense
Symptoms of a pitfall:
- AI provides specific data or facts that cannot be verified
- Output contains technical terms that do not actually exist
- Makes overly specific predictions about uncertain events
Strategies:
- Clarify uncertainty: Encourage AI to state when uncertain
- Fact-checking: Require distinction between facts and speculation
- Multi-source verification: Verify information from multiple sources
- Require citations: Ask for source of information for verification
