We don't measure if you're funny. We measure how you use humor — and that radically changes how AI should talk to you.
An AI that responds with irony to someone who doesn't tolerate it is actively worse than one that doesn't try to be funny. An AI that never uses humor with someone who values it sounds robotic and distant. Humor is the most personal and hardest-to-calibrate aspect of natural language — and probably the one that most separates "this is an AI" from "this gets me."
Martin et al. (2003) discovered that what matters isn't whether you have a sense of humor (almost everyone does), but how you use it. They crossed two axes — direction (toward others vs. toward self) and function (beneficial vs. detrimental) — and produced a taxonomy that explains why your best friend's sarcasm makes you laugh and a stranger's offends you.
Each style is relatively independent (inter-scale correlations of .12 to .36). You can score high on both affiliative and aggressive humor — and that tells the AI very different things than scoring high on only one.
Humor to build bonds. Funny anecdotes, witty observations, lightening the mood. The humor of a great host.
Humor as resilience. Laughing at adversity, finding the absurd in difficulties. The humor of someone who finds something funny in the airport queue while missing their flight.
Sarcasm, biting irony, humor at others' expense. Can be brilliant or destructive — often both. The humor of political satire and the roast.
Self-deprecation to gain acceptance. "I'm a mess" so others laugh and accept you. Works socially short-term; long-term, correlates with low self-esteem.
The HSQ (Humor Styles Questionnaire) has been validated across 20+ languages and cultures. IRT analysis by Silvia et al. (2020) confirmed robust reliability with no significant gender bias. Positive styles (affiliative and self-enhancing) predict psychological well-being above and beyond the Big Five (Ruch & Heintz, 2016).
Key finding: men score significantly higher on aggressive humor (Cohen's d = 0.72 in the original study). Women tend to score higher on affiliative humor. These group differences are statistical, not deterministic — your individual profile is what matters.
Key references
Martin, R. A., Puhlik-Doris, P., Larsen, G., Gray, J., & Weir, K. (2003). Individual differences in uses of humor and their relation to psychological well-being. Journal of Research in Personality, 37, 48-75. · Ruch, W. & Heintz, S. (2016). The virtue gap in humor.
The humor profile calibrates three dials: permitted irony level, playful vs. serious tone, and use of humorous self-criticism. The result: an AI that knows when an ironic comment will land well and when it'll sound like an insult.
"To improve your productivity, I recommend establishing dedicated time blocks, eliminating distractions, and prioritizing your tasks with an Eisenhower matrix."
"Look, the truth is half of productivity advice is recycled nonsense from 90s books. But there are three things that actually work (and no, none of them involve waking up at 5 AM): 90-minute blocks, a NOT-to-do list, and accepting that Monday isn't your day."
For someone with high aggressive and self-enhancing humor, the first response sounds like a LinkedIn coach. The second sounds like a real person.
32 questions. 5 minutes. And the AI will know whether it can drop an ironic comment or should keep the tone professional.