In the case of spaghetti sauce, at least, we cannot always explain what we want. Especially when asked about our preferences, we are more likely to list what is currently known and popular than ask for something that hasn’t yet been done or isn’t widely available (even if that is what we would prefer).
How do you like your spaghetti sauce? Thin, spicy, chunky, traditional …
The point of the talk is variability — one product cannot meet the needs of all users. In the food industry that revelation ultimately led to the 500 kinds of sauce we now face at the supermarket. (See Barry Schwartz‘s talk for more on choice overload.)
Gladwell’s talk was a good reminder to me to stray away from open-ended questions when conducting user research. A question like “What kind of features do you want in this system?” is only going to yield a list of features that are already well-established, even bad ones.