Research suggests that telling Claude what NOT to do ('Don't include lengthy preambles') is generally less effective than what alternative approach?
Select an answer to reveal the explanation.
Short Explanation
Negative instructions make Claude think about the thing you don't want. Tell it what to do instead, like 'Begin your response directly with the answer.' That gives a clear target and works better than 'Don't include lengthy preambles.'
Full Explanation
Prompt engineering research and practice consistently shows that positive instructions (specifying desired behavior) outperform negative instructions (specifying what to avoid). 'Don't use bullet points' is less effective than 'Use numbered lists only.' 'Don't be verbose' is less effective than 'Keep responses under 3 sentences.' This is analogous to human communication — it's easier to follow 'turn left at the stop sign' than 'don't go right, don't go straight.' Telling Claude what to do instead, such as 'Begin your response directly with the answer,' gives a clear target and is the reliable approach. Using ALL CAPS to emphasize restrictions, adding the word 'IMPORTANT' before negative instructions, and repeating negative instructions three times are formatting tricks for negative instructions; they may marginally help but miss the fundamental point that positive instruction framing is more reliable.