REDIMENSION

Swap the variables you describe things with, turning labels into continua, averages into distributions, and traits into states, so hidden differences become visible.

You've started describing your partner as "unsupportive." It's a single label covering a whole person, and it's making every disagreement feel like confirmation. Yet if you're honest, they're generous with time and terrible with words, attentive on weekends and absent during their work crunch. The label hides all the variation that actually matters.

REDIMENSION means swapping the variables you use to describe something so hidden differences become visible. The cue: your current labels poorly discriminate between outcomes, treating unlike situations as identical. Run the standard swaps: turn a binary label into a continuum (how supportive, in which situations, on a scale), turn an average into a distribution (great some weeks, awful others is a different problem than uniformly mediocre), and turn a trait into a state ("is unsupportive" becomes "acts unsupported when stressed"). Pick the new dimensions that improve your ability to notice, measure, or influence what's happening. It works when the new axes separate cases the old label treated alike: suddenly "during deadline weeks" looks fixable while "always" never did. Don't grab a variable just because it's easy to measure, and don't pile on dimensions that change no decision.