Explicability
202104102145-explicability

Machine learning models & interpretability (necessitating the use of simplified models) vs accuracy (e.g. kernel methods, "deep" methods, etc.)

enigma-of-reason: human reasoning as means of rationalising ourselves to others & collective problem solving, not as means of performing accurate, logical deductions

We don't really know how we walk, like birds don't really know how they fly. Scientists come in post-facto to try and explain things. antifragile

Metis knowledge is often so implicit and automatic that its bearer is at a loss to explain it.57 A staple of early medical training, I have been told, is the story of a physician who, at the turn of the century, had a spectacularly high success rate in diagnosing syphilis in its early stages. Laboratory tests confirmed his diagnoses, but he himself did not know precisely what it was that he detected in the physical exams that led him to his conclusions. Intrigued by his success, hospital administrators asked two other doctors to closely observe his examination of patients over several weeks and to see if they could spot what he was picking up. At long last, they and the doctor realized that he was unconsciously registering the patients' slight eye tremor. The eye tremor then became a universally recognized symptom of syphilis. Although this insight could be codified, what is instructive here is that it could have been achieved only through close observation and long clinical experience and that, even before then, it could have been known subliminally.

seeing-like-a-statech. 9

Model-free inferences better under low-information regimes, model-based inferences better after having more data.

Bibliography

enigma-of-reason Mercier, Hugo, and Dan Sperber. The Enigma of Reason. ↩︎ 1

antifragile Taleb, Nassim Nicholas. Antifragile: Things That Gain from Disorder. ↩︎ 1

seeing-like-a-state Scott, James C. 1998. Seeing Like a State: How Certain Schemes to Improve the Human Condition Have Failed. Yale University Press. ↩︎ 1