Balance before length

The 12-, 32-, and 60-item forms cover four dimensions with 3, 8, or 15 items each. We average each wording direction first, then weight them equally. This prevents an extra item on one side of an odd-length form from turning uniform agreement into a preference. Longer forms include the shorter items and add situations.

From answers to a scale

Answers range from −3 to +3; reverse-worded items change sign. The score is 50 + [(forward mean + sign-adjusted reverse mean) ÷ 2] × (50 ÷ 3), rounded for display. We show item and neutral-response counts alongside it. The score is not a norm-group percentile.

Three checks against overinterpretation

We pause type assignment if at least 75% of answers are neutral, 90% repeat the same choice, or all dimensions are between 40 and 60. These editorial rules limit interpretation; they do not judge the person. We test one-step answer sensitivity and suggest alternative characters for middle-zone dimensions. Sensitivity is not reliability or accuracy.

The connection rules are public too

Friendship uses internal weights of 18/30/26/16 for shared dimensions and 7 for differences. Love uses 12/30/26/12 for similarities, 18 for differences in recharge or rhythm, and 7 for other differences. These are editorial rules, not learned predictions. Ties sort by code; five candidates exclude your own type. We never present the weights as relationship success probabilities.

Software tests are not psychometric validation

Automated tests check reverse scoring, bounds, neutral handling, storage, and sharing against the design. Measuring a person accurately is a separate question. Mooa does not yet have representative samples, language-specific validation, test–retest reliability estimates, or validation against relationship outcomes. We do not claim “95% accurate” or scientific superiority.

What better evidence would require

A consented pilot would first check how questions are understood, item distributions, and dimension structure. Retesting, language review, and evaluation in a separate validation sample would follow. IPIP distinguishes reliability and validity using sample evidence. We use that distinction as guidance; we have not implemented an IPIP inventory or inherited its validation. Current answers are not sent to a server.