How do I write a good question?

Kindred is at its best on questions that are genuinely contested: questions where thoughtful, well-informed people disagree, where the evidence pulls in more than one direction, and where the honest answer is a weighing rather than a fact. It builds the strongest case on every side, grounds them in real sources, and leaves the judgment to you.

It is less suited to questions with a single settled answer (a date, a definition, a conversion) and to questions of pure personal preference, where there is no shared evidence to reason about. For those, a quick search or a calculator will serve you better.

Three qualities make a question tractable for a multi-perspective analysis.

It is contestable. There is a real disagreement to map. “Should cities replace single-family zoning with denser housing?” has serious arguments on each side; “What is the population of Tokyo?” does not.

It is specific enough to ground. A question tied to a concrete decision, population, or context can be grounded in sources and tested against precedent. “Is remote work good?” is hard to ground; “Does fully remote work help or hurt early-career employees' long-term advancement?” gives the analysis something to hold on to.

It carries real stakes. The most useful analyses sit where a decision, a policy, or a belief actually turns on the answer. Naming the stakes in the question helps Kindred frame the right one.

The same underlying topic can be asked in ways that produce sharper or vaguer analyses.

Too broad: “Is AI dangerous?” Sharper: “Do the near-term risks of large language models in hiring decisions outweigh their benefits?”

Leading: “Why is a four-day work week obviously better?” Neutral: “What is the case for and against a four-day work week for knowledge workers?” A neutral phrasing lets the steelmanning work on every side; a leading phrasing asks Kindred to argue with itself about your premise.

Two questions at once: “Is nuclear power safe and affordable?” One at a time: ask about safety, or about cost, or about a specific trade-off between them. A single clear question keeps each phase focused.

The first reasoning phase, Issue Framing, pins down what is actually being asked so the rest of the analysis answers the real question and not a nearby one. A precise question gives that phase a clear target; a vague one forces it to guess at your intent, and every later phase inherits the guess. If the framing comes back narrower or broader than you meant, that is a signal to rephrase and run it again.

You can also set emphasis without changing the question, by choosing a domain lens. A lens shifts which disciplines lead, not which question is answered.

Browse the domain lenses→

Before you run an analysis, a good question usually:

• names a real disagreement, not a settled fact;
• is specific enough to point at evidence and context;
• is phrased neutrally, without assuming the conclusion;
• asks one thing rather than several at once;
• makes the stakes, or the decision behind it, clear.

You do not need a perfect question to start. Run one, read the Issue Framing, and refine from there. The framing phase tells you how Kindred understood you, which is often the fastest way to find a sharper question.