I just finished entering “2 cups of milk” as my day-three breakfast on a tracking-based lifestyle app that shall remain nameless. It is a well-respected, leading app and was recommended by an experienced personal trainer.
Much to my surprise, it complimented my positive pattern of drinking milk in that having milk for breakfast was positively correlated with my having fewer calories for the day.
We've noticed that on days your incorporate milk, you tend to keep your total calories lower.
It's pretty aggressive to report a correlation with only two data points. It could just as easily make such a statement about anything that was different about the two days.
However, the only thing that was the same between the two days was my having milk for breakfast.
We need several separate errors for the program to give this wrong advice.
For years I've been saying that with most apps and websites, we can just reboot, and hence that reliability and such weren't as crucial in the way they are with autopilots or controllers for the power grid.
Maybe I was wrong. Poor lifestyle choices in diet and exercise are among the leading causes of premature death and overall poor health in our country. Giving wrong advice to people trying to do better might be causing more deaths than an occasional plane crash would.
How much wrong advice is it dispensing
Does/should the app company have responsibility or liability for giving wrong advice? How would a proper testing and quality assurance program prevent this?
How much wrong advice is it dispensing? How much worse can it do when it has more than two data points to become confused with?
The world may never know.