How the questions are written, and the rule behind all of them

Both partners get the same question, at the same time, and both answer it before either can read the other. That single fact constrains what a question can be far more than it first appears.

The rule

Every question addresses the reader in the second person about their partner in the third person. Never in the first person, as though one partner were asking the other.

So: "What small thing does your partner do that makes you feel loved?"

Never: "What small thing do I do that makes you feel loved?"

It reads as a minor stylistic preference and it is the difference between a question that works and one that breaks.

If a question is written as though one person is asking it, the other person cannot answer it. Both people receive the same text.

Why it breaks

Imagine the first-person version arriving on both phones simultaneously. Each partner reads "what do I do that makes you feel loved" and has to work out who "I" is.

The answer is: whoever is holding the phone. Which means each of them is being asked to describe themselves, which is not the question anybody intended, and the two answers will not be about the same thing.

The mechanism only works when a question means the same thing to both readers and produces two answers that can be read side by side.

This is enforced by a constraint in the database rather than by good intentions, because in sixteen languages and several thousand questions, good intentions are not sufficient.

What else a question has to do

Be answerable in two or three sentences. A question that requires a paragraph gets skipped on a Tuesday, and skipped questions are how habits die.

Produce two different answers. "Do you love me?" is a bad question because both answers are known in advance. "What did I get wrong about your week?" is a good one because neither person can predict it.

Be about the present or about a change. Questions about the past are exhausted quickly in a long relationship, which is why most published lists of couple questions stop being useful after a fortnight.

Not be a trap. Any question where one answer is obviously correct produces compliance rather than information, and compliance is worse than silence because it looks like an answer.

Not require a confession as the price of entry. Questions that can only be answered by disclosing something painful get skipped, and repeated skipping trains a couple to treat the whole thing as optional.

The proportion of light ones

Roughly two thirds of what we write is ordinary, and some of it is silly.

This is not a failure of ambition. A product that asks something significant every day becomes an obligation, and obligations get abandoned in about three weeks. The serious questions work because they are surrounded by unserious ones, so no individual day carries weight.

Translation

Every question exists in sixteen languages and the rule has to hold in all of them, which is harder than it sounds because languages differ in how they mark person and formality.

A question that is neutral in English can land as intrusive in a language with a formal register, and a phrasing that is warm in one language is clinical in another. These are not translation problems in the usual sense, they are rewriting problems, and the rule about perspective is the fixed point they are written around.

What we do not do

We do not write questions designed to produce a particular answer.

We do not write questions about the app, or about how the relationship is going, because both of those invite performance.

And we do not read what anyone answers, which means we cannot tune the questions on responses. We can only look at which ones get skipped, and even that is coarse. The questions are written the old way: by people, and revised when they are obviously not working.