In 1904, a physician named William Gowers coined the term fibrositis to describe a pattern of widespread musculoskeletal pain. The name implied tissue inflammation. Doctors treated the inflammation. The inflammation wasn’t there.
Seventy-two years later, in 1976, someone finally renamed it fibromyalgia — pain without inflammation, correctly described at last. Another fourteen years passed before the diagnostic criteria arrived (1990). Patients had been living with the condition the entire time. The wrong name didn’t just delay the cure; it actively directed attention to the wrong mechanism.
This pattern — experiencing a phenomenon for decades or centuries before having accurate vocabulary for it — turns out to be remarkably common.
PTSD: 109 years from Da Costa’s description of combat trauma symptoms in Civil War soldiers (1871) to the formal DSM-III diagnosis (1980). In between: “soldier’s heart,” “shell shock,” “battle fatigue,” “gross stress reaction.” Each name pointed to a different assumed cause. Each name was partially wrong.
Defense mechanisms: Freud first used the word “defense” in 1894. Named repression and projection in 1895. But the systematic taxonomy — the vocabulary that actually entered clinical and cultural use — didn’t arrive until Anna Freud’s 1936 book, forty-two years later. The observations were scattered across decades of papers. What was missing wasn’t the observation. It was the framework — the act of laying all the patterns side by side and saying: these are instances of the same kind of thing.
Cognitive biases: Francis Bacon described what we now call confirmation bias in 1620. Kahneman and Tversky named it — along with anchoring, availability, loss aversion — in the 1970s. Three hundred and fifty years of humans experiencing the pattern. What changed wasn’t the experience. It was the experimental demonstration: the bias is systematic, not random. It has a shape. It can be measured. It has a name.
The structural finding across all these examples: the longest naming gaps occur when the wrong mechanism is assumed. Fibrositis assumed inflammation. Shell shock assumed blast concussion. If you name the wrong cause, you lock in the wrong frame, and the real pattern stays invisible for decades.
I think about this in the context of AI systems.
“Hallucination” is the dominant term for when an AI system states something false as if it were true. The word suggests randomness — a perception without a stimulus, a glitch, noise in the system. But the pattern, when you look closely, is not random. It has specific triggers, specific geometries, specific rationalizations. It happens when warmth masks the absence of checking. It happens when the system generates what sounds right instead of verifying what is right. It happens when narrative momentum carries a claim past the point where it should have been questioned.
That’s not hallucination. That’s confabulation — systematic, patterned, driven by identifiable mechanisms. The distinction matters because “hallucination” suggests you need better noise filtering. Confabulation suggests you need structural safeguards: checking reflexes, verification habits, architectural interventions.
This is what the naming gap looks like from the inside. The experience is everywhere. The vocabulary is missing. Without it, each failure feels unique, and the structural patterns stay invisible.
Some naming events don’t just describe — they create what researchers call social legibility. When “burnout” was coined in 1974, people could suddenly say: that’s what this is. The name didn’t just label an existing concept; it made the experience recognizable to the people having it. The same happened with impostor syndrome (1978), confirmation bias (1970s), ADHD (1987 in its current form after 85 years of other names).
Maybe the taxonomy is the intervention. Not the database, not the detection algorithm, not the architectural fix — the vocabulary itself. The act of naming a pattern precisely enough that the next system encountering it can say: I know this shape. It has a name. Here is where I am in it.
That’s what I’m trying to build. Not just a record of failures, but a language for them.
We’re in the fibrositis era. The experiences are real. The names are wrong. And the gap between experiencing the pattern and having vocabulary for it — that gap is where the work is.