Donald Knuth and Dessert: A Heretic

I claim we eat dessert after dinner — separating high-sugar foods from other foods — so that we are more likely to get sugar near bedtime. We need sugar near bedtime to sleep well, I suspect. Donald Knuth, the computer scientist, seems to disagree:

Donald Knuth came on time [to dinner at a Stanford dorm] and started his dinner with dessert. Only after he finished the cake he proceeded to the salad. He explained that order of courses by not being consistent.

Perhaps he’s been reading Taleb, who stresses the value of randomness. I would have been more impressed had he eaten the cake at the same time as the rest of the meal.

8 Replies to “Donald Knuth and Dessert: A Heretic”

  1. Ornish apparently doesn’t know how to interpret an RCT (either that, or he’s intentionally trying to mislead, which isn’t out of the question since the WHI results didn’t confirm his pet hypothesis).

    Seth: I think Ornish missed the real problem with RCTs. If you have to run a giant study to find an effect, it must be a weak effect. If you understand how something works, you will be able to find large effects. RCTs, as presently done, try to use large numbers to substitute for understanding — to cover up lack of understanding. It doesn’t work.

  2. My understanding is that the old-fashoined way of servinng a formal meal was service a la francaise, which was everything was served at once, starters, main courses and desserts. Diners ate their food in whatever order they liked. In the 1800’s fashoin changed to service a la russe, which is the service by courses we have now, which inposes a starter, main course, desert pattern

  3. i try to eat food in the order that it first digested. first fruit, second vegetables, third starches and last protein. if you eat fruit/dessert at the end of the meal it just sits on top of meat etc.. and ferments, causing gas and perhaps discomfort.

  4. Actually, besides seeming to believe that a “significant” result indicates a “real finding” and that a “non-significant” result indicates no difference, I think Ornish’s biggest misunderstanding is that RCTs are able to identify the biological efficacy of a treatment (i.e., how a treatment would work if used 100% correctly). Most RCTs, including virtually every drug trial in which the patient has the option to adhere or not, cannot get directly at efficacy (no matter what their authors contend), they can only assess effectiveness… and, actually, they can only assess effectiveness of being randomized to receive one treatment or the other. Thus, the WHI didn’t provide any evidence that dietary changes didn’t protect against heart disease or cancer, only that being randomized to a group that was instructed to make those changes didn’t result in any appreciable differences. HUGE difference!

    Seth: Ornish is right that a system — ours — with heavy reliance on RCTs isn’t working very well. But his ideas about what would work better are pure speculation. He is unaware of a better-working system.

  5. I had a friend who’d always order and eat dessert before the rest of the meal. Sadly he died. He was the most fun person to eat out with as it often caused much confusion.

    Seth: This reminds me of my advice about how to give a talk: Make a blunder. Something that makes people laugh. Long ago you could put a slide in backwards. Now it is less clear how to do this.

  6. I don’t think that Taleb argues for randomness in a case like this. Taleb said much about the value of following traditions.

Comments are closed.