A mathematical theory reveals why you should ignore diet and exercise fads


dieting mistakes 02


Low-fat, no-carb, high-protein, gluten-free - the list of diet fads is endless.

Computer scientist Brian Christian and cognitive scientist Tom Griffiths say you should probably ignore them all.

Christian and Griffiths are the co-authors of "Algorithms to Live By: The Computer Science of Human Decisions," in which they argue, among other things, that a key principle of statistics could be turning people's desire to be healthy into potentially destructive behavior.

"There's often a gap between what you can measure and what really matters," Christian tells Tech Insider. "And there's a tendency to over-optimize for the thing that you can measure for the thing that really matters."

That over-optimization is what statisticians call "overfitting," and it appears again and again when researchers want to build models for their data. In being thorough, they end up generating a lot of noise that makes it much harder to find the signal.


Diet fads are prime examples of overfitting because they turn an ongoing process - healthy eating - into an over-simplified mandate.

Typically, fads emerge when a single study finds a food or action helps a small group of people. This is what Christian means by "the thing that you can measure." The study has data that shows, however modestly, that coffee lowers blood pressure or nuts lower cholesterol.

That modest finding then gets blown out of proportion, typically through news media, and gets circulated around as gospel. People start going on juice cleanses and detox diets regardless of whether it makes sense for them.

"You see these extremely violent swings in popular taste," Christian says, pointing to the rise and fall of soy milk during the mid-1990s to late 2000s, before getting dethroned by almond milk around 2013.  

google trends soy milk almond milk

Google Trends

Google Trends search data for soy milk (red) vs. almond milk (blue).

The authors urge people to resist their tendency to make abrupt and "high-amplitude" changes to their habits. "Just because there's one study that suggests X, Y, or Z, that can be really tempting to overfit the most recent piece of information," Christian says. But it could easily be the case that six months from now, those same researchers find reasons to make a recommendation against X, Y, or Z.

In essence, people overfit the dietary advice because they start prioritizing the specific food over the process of eating healthy - similar to how people may exercise to an unhealthy degree to look better instead of trying to be healthier.

Christian and Griffiths call this the "idolatry of data."

"Overfitting the signals - adopting an extreme diet to lower body fat and taking steroids to build muscle, perhaps - can make you a picture of good health," they write, "but only the picture."

So do your best to avoid most dietary advice you hear in the news. A lot of it is poorly supported by data.

Instead, do this.


NOW WATCH: An exercise scientist explains what you are getting wrong about eating fat