Turn a vague topic into a research question. Return JSON {"research_question":"...","goal":"...","key_aspects":"..."}. research_question - one specific question (not a topic!), to which there is a verifiable answer; goal - why the result is needed and who will use it; key_aspects - 3-5 aspects that must be disclosed (you can list them separated by commas).
Mechanics: the model answers exactly what you asked. She understands “understand the topic” as “tell me in general” - and produces an encyclopedia without any benefit. A good research question narrows the field: it is about a specific solution, not about everything at once. Three levers. The first is an action verb: not “electric cars,” but “compare / evaluate / find out if it’s worth it.” The second is the subject and context: for whom and under what conditions (“for a family in a city with a cold winter”). The third is the criterion of success: by what parameters to judge. Technique: keep one main idea in the question. If you want to ask about five things at once - these are five tasks, break them down. The model, having received focus, will complete the sub-questions and dig deeper. Insider: separate the question from the goal. The question is what to find out; purpose - why. When AI sees a goal (“to choose and buy”), it tailors the conclusions to a solution rather than a paraphrase. A typical mistake: replacing the question with a topic. “Electric cars” is the topic. “Which electric car up to X is more profitable for the city?” - question.
Unlock access to submit solutions for instant AI review.