But if, as hotpotamus suggests, "consent" is a useful way to look at this, then that's not good enough. You can't randomly do surgery on people just because you believe there's a 90% chance they'll be happy with the results.
Someone is unconscious and dying, you can save them with surgery, surely it’s highly unethical to wait for their explicit consent?!
Every floating point article (including IEEE 754) I've seen treats normal floating point numbers as dyadic rationals.
But if, as hotpotamus suggests, "consent" is a useful way to look at this, then that's not good enough. You can't randomly do surgery on people just because you believe there's a 90% chance they'll be happy with the results.