From proteinmaxxing and fibermaxxing, today’s extreme fad diets are couched in the language of optimization. Health influencers and celebrities say we need to squeeze as much of these specific nutrients into our meals as possible to maximize how we look or how long we live. While it sounds nice in theory to strive for the best diet, what would that mean? The history of diet optimization suggests that the question itself is ludicrous
To set the stage for optimal dieting, we need to go back to the early 20th century when scientists discovered that a constellation of mysterious illnesses, including beriberi, pellagra and rickets, was caused not by germs but by a lack of certain chemical compounds we now know as vitamins. As soon as a unified theory of vitamins emerged in the 1910s, public health officials sought to get the new science out to families to prevent illness and improve well-being
Hazel Stiebeling — a Columbia University-trained chemist who had written her doctoral thesis on vitamins A and D — was one of the figures who took charge, leading the food economics division of the U.S. Department of Agriculture. Stiebeling modernized the department’s dietary recommendations to incorporate new scientific knowledge of vitamins, publishing her first plans with co-author Rowena Carpenter in 1936. During the Great Depression, their “Diets to Fit the Family Income” proposal was intended to meet basic dietary requirements on a limited budget.
Article continues below this ad
The diet’s “minimum cost” plan ran around $120 per person per year at 1935 prices. It included a pint of milk, three to four servings of vegetables and at least one serving of potatoes each day, adding cereal for breakfast and bread with every meal. The diet suggested consuming citrus or tomatoes 2 to 3 times per week, and meat, fish or eggs 3 to 4 times a week. It even budgeted for some occasional desserts
While most people would see this as a reasonable and modest proposal, University of Chicago economist George Stigler saw government bureaucrats falsely using the language of science to make recommendations about how people should live
Stigler set out to prove that Stiebeling and Carpenter’s diet was far from “minimum cost.”
Combining recommendations from the National Research Council and data from the Bureau of Labor Statistics with some mathematical heuristics, Stigler found a combination of foods that would cost a person only $40 per year, more than 2½ times less expensive than Carpenter and Stiebeling’s plan. He had cracked the code of savingsmaxxing. But what was in this diet? Every day, a person would consume 1 pound of flour, 2 ounces of evaporated milk, 5 ounces of cabbage, 1 ounce of spinach and five cans of navy beans.
Article continues below this ad
That’s not a diet! Or at least not one any reasonable person would be satisfied eating
Stigler, a prickly critic, of course knew this diet was gross and that many subjective factors influenced what every person eats, from taste to company and culture. He grumbled that the nutritional recommendations were based on guesswork, often inflated by 50% just to be safe. The amount of nutrients in cooked food varied depending on its preparation. A Granny Smith apple would have a different amount of vitamin C than a Red Delicious. There was so much complexity that writing down a single optimal answer for everyone was impossible.
Because of this uncertainty, Stigler thought it was inappropriate for the government to recommend what people should eat. He warned that these recommendations, with their appeals to scientific expertise, might be mistaken for rules. The United States was home to people with different preferences, practices and beliefs, and a uniform recommendation couldn’t satisfy all of them
Stiebeling and Stigler’s views on diet recommendations are emblematic of the opposing political philosophies of the two parties that have governed the U.S. since the 1940s. But there are also lessons for us in this fight about diet outside the political sphere. In undergraduate computer science classes about algorithms and operations research, we still teach Stigler’s diet problem as a prototypical example of linear programming
The backbone of modern computational optimization, linear programming is used in tasks as diverse as scheduling flight times to transporting goods from warehouses to shops. Optimization can be useful, but only when the models accurately reflect reality and a single-objective function, like the shortest path, reasonably captures the problem you are trying to solve
Article continues below this ad
About Opinion
Guest opinions in Open Forum and Insight are produced by writers with expertise, personal experience or original insights on a subject of interest to our readers. Their views do not necessarily reflect the opinion of The Chronicle editorial board, which is committed to providing a diversity of ideas to our readership
Read more about our transparency and ethics policies
Real life seldom meets those criteria, and optimization rarely applies to anything we do. This is why optimizing a single objective gives you grotesque answers like a pound of flour and a can of navy beans. Optimization can only fill in what you model, and if you leave out details, maxxing will give you something you didn’t expect and probably didn’t really want
Even the simplest optimization problems are loaded with approximations, idealizations and value judgments. These issues do not go away when we try to optimize more complex and challenging problems. Which is why the way out of the culture of maxxing is to recognize that our goals in life are never reducible to a single mathematical formula
Benjamin Recht is a professor of electrical engineering and computer science at UC Berkeley and an author, most recently, of “The Irrational Decision.” This was written for Zócalo Public Square
