It is not ‘scientific’ to base risk entirely on experienced probability, especially without taking into account the relative scale of the cost.
A 10% chance of something bad happening could be an acceptable risk or not, depending on how bad the bad outcome is. Is it death? In that case, 10% is way too high... eat that old pizza seven times, and you already have a greater than 50% chance of being dead from food poisoning.
This is the problem with using personal experience for low probability/high cost events. Drunk driving is like this; I know too many people who say “drunk driving isn’t dangerous! I’ve driven drunk dozens of times and have never had anything bad happen!”
Most of the time you are fine, but the extreme cost of what happens when you aren’t make it a risk that should not be taken.
Except as the author points out, we have more than just our own personal experience to go on. We have the experience of millions of college students, who, as the author notes, are not being dragged out in body bags by the thousands or even hundreds due to eating day-old pizza. Maybe the evidence isn't organized, easily referenced, or particularly well-controlled, but it exists, and to ignore it is just silly.
We're also talking about two different things here, one being the scientific method, the other being risk analysis. Output from the former can be an input to the latter but they're otherwise orthogonal.
On the subject of risk, I don't think the drunk-driving comparison holds because the numbers are completely different. We can point to innumerable cases of drunk-driving resulting in death. But day-old pizza? I'll be surprised if you can find (just picking a number out of my hat here) even ten cases of day-old pizza killing anyone.
A 10% chance of something bad happening could be an acceptable risk or not, depending on how bad the bad outcome is. Is it death? In that case, 10% is way too high... eat that old pizza seven times, and you already have a greater than 50% chance of being dead from food poisoning.
This is the problem with using personal experience for low probability/high cost events. Drunk driving is like this; I know too many people who say “drunk driving isn’t dangerous! I’ve driven drunk dozens of times and have never had anything bad happen!”
Most of the time you are fine, but the extreme cost of what happens when you aren’t make it a risk that should not be taken.