Maieutic Prompting: Definition and Examples

Maieutic Prompting is an advanced prompt engineering method directly inspired by Socrates' maieutics — the art of giving birth to minds through questioning. Applied to large language models (LLMs), this technique involves asking the model to generate a response, then decomposing it into sub-claims that it must justify or refute itself, thus creating a self-verified reasoning tree.

Concretely, Maieutic Prompting works in two steps. First, the model produces an initial explanation or answer. Then, it is asked to question each element of this answer: 'Is this true? Why? What would be the alternative explanation?' The model thus generates a tree of propositions with their justifications and counter-arguments, making it possible to identify logical inconsistencies and converge toward a more reliable answer.

This approach is particularly effective for complex reasoning tasks where the model might initially produce plausible but incorrect answers. By forcing self-examination, Maieutic Prompting significantly reduces hallucinations and reasoning errors. Research works, notably those by Jaehun Jung et al. (2022), have shown that this method improves LLM accuracy on reasoning benchmarks compared to standard approaches.

Maieutic Prompting differs from other techniques like Chain-of-Thought in that it does not merely ask the model to 'think step by step', but actively requires it to confront its own statements, explore contradictory hypotheses, and build verifiable logical coherence.