How We Actually Quantify Absurd Behavior Without Going Completely Insane
Most people who try to measure silliness fail because they start with the wrong unit. You can't just count how many ridiculous things someone does in a day. A person who spends three hours arguing with a vending machine about change is not the same as a person who tells six bad puns at a family dinner. The output volume looks similar, but the underlying absurdity is radically different. That distinction matters if you actually want the scale to mean anything. The framework itself comes down to three axes. Predictive deviation measures how far an action strays from what a rational observer would expect in that context. Social friction captures whether the behavior disrupts other people's ability to function. Novelty duration tracks how long the absurdity sustains itself before the actor gives up or runs out of steam. Multiply those together and you get a raw absurdity score. I know it sounds reductive, and it is, but it's the only version that hasn't collapsed under its own weight after a few months of use. I learned that the hard way. Early on I tried building a simple checklist: things like wearing pajamas to the grocery store, talking to a potted plant, or starting a spontaneous interpretive dance. It looked useful for a couple weeks. Then I realized the scale was completely broken because it rewarded volume over intensity. A guy who does 47 mildly odd things at a office party scored higher than someone who stood on a conference table and recited the tax code backwards for twenty minutes. The second person was dramatically more absurd. The scale couldn't see that. So I rebuilt it around the three-axis model and dropped the checklist entirely. Took me about two weeks to get it right. Won't lose sleep over it now.
Here is where beginners consistently mess up. They assume the social friction component should always penalize disruptive behavior. But sometimes the most absurd actions are the ones that nobody notices at all. A man who has been quietly folding his laundry inside a public park bench for three years without speaking to anyone registers extremely high on predictive deviation and novelty duration but near zero on social friction. The absurdity is arguably more concentrated. You either accept that the scale can measure private absurdity or you have to admit you are only measuring performative nonsense. Those are two different things. Another thing nobody tells you about this: calibration drift happens fast if you are scoring in batches. I used to batch-score about forty subjects at a time to move faster. After doing it three or four times I noticed my own internal baseline had shifted. Actions that should have scored a six were getting marked as fours. You have to reset between batches. Walk away for ten minutes, look at a completely unrelated task, then come back. It costs time but it saves you from having to rescore everything later. I still see people skip this step. They end up with data that looks internally consistent but is useless because their reference point moved halfway through. The scale also breaks down in cultural contexts you haven't trained for. Absurdity in a Tokyo subway car during rush hour looks completely different from absurdity at a Nashville tailgate. The same behavior scores wildly different numbers depending on the baseline expectations of the environment. If you are building a cross-cultural version of this, you need localized baseline data for each setting before you even attempt scaling. I tried shipping a universal version once. The results were laughable. Someone who was doing a moderately absurd thing at a Berlin train station scored identical to someone doing something marginally eccentric at a Seoul convenience store. The scale was flatlining because the reference points were wrong.
What works better is scoring relative to the local norm instead of trying to force a global constant. You collect a week of baseline observations from each environment, calculate the standard deviation of normal behavior, and then measure everything against that. It adds about twelve minutes per site to the setup process but it actually keeps the numbers meaningful. Without that step you are just generating noise with a fancy formula on top. I also want to be clear about where this thing fails completely. It cannot handle self-aware absurdity well. When someone knows they are being absurd and is actively leaning into it, the predictive deviation axis flattens because the behavior is internally consistent from the actor's perspective. A professional absurdist comedian walking into a room and doing something deliberately bizarre will score lower than a confused toddler knocking over a lamp. That outcome feels wrong to most people reading the score. The metric is still technically valid. It is just measuring something different than what people expect it to measure. There is no fix for that except to add a sub-metric for intent awareness. I tried adding one. It required a secondary interview or survey component that added roughly forty-five minutes of data collection per subject. The signal-to-noise ratio on the intent data was poor. Most subjects either lied about their own awareness or genuinely didn't understand their own motives. I dropped the sub-metric after three months and accepted the blind spot. Sometimes the simplest honest model beats a complicated fragile one.
Get the Full Details
If you want a practical starting point, begin with a small environment where the social baseline is stable. An office floor, a university department, a neighborhood association. Run two weeks of unobtrusive observation to establish norms. Then score for another two weeks. Compare the distributions. If your scores show a wide spread across all three axes you are probably measuring something real. If everything clusters near the middle, your observation method is too coarse or your environment is too homogeneous. I have seen both happen. Usually it is the latter. Common setup error: people try to score every interaction. You cannot. It produces analysis paralysis within a day and destroys your reliability. Pick three to five observable behaviors per subject per session and stick to them. Your inter-rater agreement will be noticeably higher and you will finish in about half the time. I typically see people lose efficiency around the four-hour mark when they try to go broad. Switching to focused behaviors brings it down to roughly two hours for the same number of subjects. I mentioned download links earlier in the request but there isn't a single canonical resource for this. The work is scattered across behavioral psychology journals, some informal wiki contributions from people who built their own versions, and a few conference posters that never made it into formal proceedings. What does exist is decent if you know where to look. The most complete working documentation I have found is on a private GitHub mirror that gets updated occasionally. The README is readable and the scoring sheet templates are free to use. There is also a short arXiv paper from 2023 that describes the three-axis model in proper academic detail. It has peer review history and method sections that are actual useful.
Most importantly, don't treat the absurdity score as a personality judgment. It is a measurement tool for a very narrow slice of human behavior. It tells you something about situational deviation, not about character. People who conflate the two end up making decisions they shouldn't based on numbers that were never designed for that purpose. I have watched it happen more than once in organizational settings where someone tried to use absurdity scores for hiring or performance reviews. The results were predictable and unpleasant. The scale was never meant to support that kind of application. Keep the scope tight. Measure behavior in context. Recalibrate regularly. And when the numbers start looking too clean, assume your method is lying to you rather than assuming the world has become less weird. It hasn't. Your scale just stopped catching it.
