It depends on how you came by the information that “at least one of them was heads”.
I’m not sure why no one brings this up, it’s an incredibly important piece of information that’s missing. Try and apply this theory to reality and prepare to be confused.
For example, if you’re blind, and you flip two coins, a machine sees them and says “at least one of the coins is heads” what are the odds the other coin is heads?
In that case you flipped two coins and are given that “at least one of them is heads”.
It’s not 33%! It depends on the rule your machine is following when he talks. If his rule is “If at least one coin shows heads, I will always inform them of such and never anything else” then yes, it’ll be 33%. But if it’s rule is almost anything else, it won’t be!
For example if it’s rule is “I will always inform him regarding one of the coins that I see” Then half the time He’ll say heads, half the time he’ll say tails.
The sample space in that scenario is (assuming each flip combination happens twice for full sample space)
HH says Heads 2 times
TH says Heads 1 time and Tails 1 time
HT says Heads 1 time and Tails 1 time
TT says Tails 2 times
In that more likely case where the machine randomly chose one coin to tell you of, If he tells you “one of the coins is heads” the 2/4 times the other coin will be heads too. 50%, not 33%
Whenever you see the word “given that…” always wonder, “but how is it given?”. (I mean maybe not useful in a math test, but 100% needed in real life probabilities)
All I’m saying is knowing that “at least one of the coins is heads” is a given, is not nearly enough information if you don’t know how it was given.