Title: "More Than We Can Say": High-Dimensional Social Science and the Conjunction of Small Effect Sizes Date: 2020-01-01 Category: commentary Tags: categorization, epistemology, sex differences Status: draft > But all of us know much more than we can say, and many times we cannot really put it into words at all. > > For example, if we have eaten them, we know what strawberries taste like. We have in us somewhere knowledge—a memory, many memories—of the taste of strawberries. Not just one berry either, but many, more or less ripe, or sweet, or tasty. But how can we really speak of the taste of a strawberry? When we bite into a berry, we are ready to taste a certain kind of taste; if we taste something very different, we are surprised. It is this—what we expect or what surprises us—that tells us best what we really know. > > We know many other things that we cannot say. We know what a friend looks like, so well that we may say, seeing him after some time, that he looks older or no older; heavier or thinner; worried or at peace, or happy. But our answers are usually so general that we could not give a description from which someone who had never seen our friend could recognize him. > > —John Holt, _What Do I Do Monday?_ [TODO: intro framing] https://www.lesswrong.com/posts/Cyj6wQLW6SeF6aGLy/the-psychological-unity-of-humankind neural networks are famously uninterpretable it's easier to recognize an accent than say exactly what vowels are pronounced differently, easier to recognize face sex than say on the basis of what features del Guidice on interpretability