https://www.nature.com/articles/s41467-023-37457-1
Most nutritionists would probably agree that eating more whole foods and avoiding overly processed foods is healthier for us. The problem is defining exactly what we mean by “processed”. I tend to use a fairly loose definition, I count canned beans or frozen peas as “whole”, and reserve “processed” for things like cookies or donuts. But it would be nice to have a more scientific way to quantify this. Researchers addressed this by training a machine-learning algorithm to look for concentrations of a list of healthier and not-so-healthy nutrients in foods and categorizing them as unprocessed, processed, or ultra-processed. As seen above, raw onions are correctly categorized as having a high likelihood of being unprocessed. In contrast, deep-fried onion rings are categorized as having a high likelihood of being ultra-processed.
This AI algorithm was then used to examine the typical foods in the US diet as reported in the National Health and Nutrition Examination Survey (NHANES) and found the typical diet to be overly high in ultra-processed foods (more than 70%). Further, they were able to show how lowering the amount of processed foods by food substitutions correlates with better health outcomes.
Although common sense goes a long way towards healthier eating (“eat more whole foods like fruits and veggies and eat less junk”) I think algorithms like this one are useful in quantifying the issue.
