Losing data when you're a professional who should have known what you're doing is one thing.
Losing your life when you're a layman in the medical field is another.
Doctors make mistakes, but there isn't anything preventing that in any event. In theory a doctor can prescribe any drug for any ailment (though there are lots of checks and consequences), and prescribing drugs for things other than their FDA approved use is very common[1]. You have to rely on doctor judgment at some point and efficacy trials don't substitute for competent doctors. And sometimes the whole process goes awry because a drug is approved for one problem, later becomes the only treatment for some people with a second problem, then the FDA withdraws approval because there are now better treatments for the first problem, leaving those with the second problem in the lurch.[2]
[1]http://en.wikipedia.org/wiki/Off-label_use [2]http://nielsenhayden.com/makinglight/archives/007140.html There's some important context in the comments.
"Programmers are quite capable of reviewing the evidence to see if code is effective" — yet, we end up with thousands of people using MongoDB based off fad and hype and losing data by the day.
(Wait, did I just advocate for a government-level code approval board? In the right hands it would probably help if it prevented more Mongos and Hadoops from entering the world.)