The quiet revolution: how AI augments the practitioner

There is a louder conversation happening about AI and skincare. We are not joining it. The louder conversation tends to fall into one of two postures — the breathless one that promises a future in which the practitioner is replaced, and the dismissive one that promises a future in which nothing has changed. Both are wrong, and both are wrong in the same direction: they imagine that the addition of a new instrument changes what the practitioner is for.

I have spent eight years building dermal-imaging systems and the last four studying the ways those systems mislead the practitioners who use them. The systems work. They also fail in patterns the practitioner needs to know about. Both are true. The discipline is in holding both of them at once.

The instrument is the instrument

Computer vision in dermal analysis is, today, reliable enough to be trusted with calibrated questions: how does this region of stratum corneum hydration compare to the same region on this client's previous visit? It is not reliable enough to be trusted with conclusions: what should we do about it? The latter is, and will remain, the practitioner's question.

The error mode of contemporary systems is the same as the error mode of all confident statistics: they are confident in the cases that look like the training data, and they are silently wrong in the cases that don't. The practitioner's job is to know when she is in a case that doesn't.

The instrument is the instrument. The judgement is the practitioner's.

What we are teaching, and what we are not

In the AI-Augmented Practice module, students learn what the systems do well, what they do badly, and how to design a clinical workflow that benefits from the former without quietly inheriting the latter. They also learn — and this is the point of the module — to read the systems' outputs as data rather than as recommendation.

We are not teaching students to defer to the instrument. We are teaching them to treat the instrument as one more sense, the way a clinician treats a stethoscope. The stethoscope is reliable. The stethoscope is also useless without the doctor.

— Katsuko Yoneda, AI-Augmented Analysis