Embodied perception is the term Nita Farahany uses for one of two kinds of knowledge that, she argues, have long been conflated under the single word "expertise." In her framework, set out in AI Isn't Replacing Expertise. It's Showing Us What We Value. (Farahany, February 2026) (February 2026), AI's arrival forces a separation between pattern recognition — matching new input against a library of prior cases, at which AI may now exceed humans in narrow domains — and embodied perception, an integrated, tacit, body-mediated form of knowledge she contends may be structurally inaccessible to systems without a body.
Pattern recognition
Farahany describes pattern recognition as matching what one sees against a large store of prior experience, and argues AI is increasingly better at it than humans in narrow domains. Her examples are drawn from sports judging. The Judging Support System in artistic gymnastics can identify roughly 2,000 elements with about 90 percent accuracy. AI-driven figure-skating analysis at the 2026 Milano-Cortina Olympics, using 14 8K cameras, computes airtime, landing speed, and rotation heat-maps in real time. She notes that AI does not get tired, is not anchored by the last routine it watched, and can train on more examples than any human can.
Farahany draws a parallel to Herbert Simon's chess research, which found that grandmasters do not think harder than novices but see differently, perceiving positions in large meaningful chunks built from tens of thousands of games. In her account, AI replicates and often surpasses this layer of expertise.
Embodied perception (tacit knowledge)
The second kind of knowledge, in Farahany's framework, is harder to name, may genuinely require a body, and seems to integrate sensory input simultaneously rather than sequentially.
Her central example is the chicken sexer. In the 1920s, Japan's Zen-Nippon Chick Sexing School trained experts to sort day-old hatchlings at 800 chicks per hour with 98 percent accuracy, yet the experts could not explain how they did it. Farahany cites cognitive scientist Richard Horsey, who found that even the best sexers said they "just knew." Formal instruction — lectures, diagrams, decision rules — could not train others. Trainees learned only by working alongside a master, picking up a chick, guessing, hearing yes or no, and repeating thousands of times; they eventually became experts but still could not explain how. By Farahany's account, the chicken sexer does not just see the chick but holds it, feels it, smells it, and integrates these simultaneously, and a gymnastics judge perceives the difference between a routine that is technically correct and one that is alive.
Farahany invokes Michael Polanyi's The Tacit Dimension, which names the structure as "we know more than we can tell" and holds that experts perceive wholes before they can identify particulars. She situates her argument in a tradition running from Hubert and Stuart Dreyfus through contemporary enactivism, which holds that embodiment is constitutive of, not incidental to, the highest forms of expertise. On this view, AI — having no body, no fatigue, and no felt experience of the world — may be structurally unable to access this layer, not because of a technical limitation more data will solve but because of what it is.
The Lindsey Vonn case
Farahany's grounding example is from the 2026 Milano-Cortina Olympics, where 41-year-old Lindsey Vonn skied the Tofane course with a fully ruptured left ACL, crashed within seconds, and was airlifted off the mountain with a broken leg. Cameras measured her speed at impact, but, in Farahany's framing, could not score what it meant for her to push out of the gate in the first place:
That's what we watch the Olympics for. Not to see a body satisfy a checklist. To witness courage and artistry, the way an athlete pours herself into the fraction of a second between a rule's specification and its execution. It's our awe at the felt experience.
The distinction and its stakes
Farahany argues that observers have conflated what is measurable with what is meaningful and called the package "expertise." She points to figure skating's scoring overhaul after the 2002 Salt Lake City Olympics, which tried to systematize artistic impressions into discrete component scores — an instinct, she says, to decompose wholes into measurable particulars when human bias is found. If Polanyi is right, she contends, decomposing a whole into particulars can destroy the understanding itself.
Her prescription is to let AI handle the measurements — accurately, fairly, without fatigue, and with detectable and correctable bias — while protecting the space where the only appropriate response is awe. The goal she names is not better checklists but honesty about what is being celebrated.
Farahany generalizes the chicken-sexer and gymnastics-judge cases to "elite sports, medicine, law, and every other domain where machines are arriving." In each, she identifies the pattern-recognition substrate as the separable layer and the embodied perception of meaning as the genuinely irreducible one, arguing that domains which conflate the two will be threatened by AI while domains that learn to separate them will be made more honest.
The Jordan Chiles counterpoint
Farahany uses the 2024 Paris Olympics Jordan Chiles bronze-medal case to make the opposite point: AI scoring Chiles' tour jeté full would have automatically credited the element, avoiding the four-second-late inquiry, the procedural collapse, the medal stripping, and the months of racist abuse that followed. For element-counting and rotation-detection, she argues, AI is simply better than human judges. Her point is not that AI is bad at judging but that pattern recognition and embodied perception need to be separated and assigned to the appropriate kind of judge.
Relationships
- introduced-by: AI Isn't Replacing Expertise. It's Showing Us What We Value. (Farahany, February 2026)
- depends-on: Polanyi The Tacit Dimension (1966); Dreyfus Mind Over Machine (1986); Herbert Simon's chess research
- related: Expertise Framework (Autor–Thompson), Cognitive Friction, LLM Fallacy
- instance-of: AI Labor Disruption
- contradicts: the AI-replaces-expertise framing common in industry messaging