> assuming people that give wrong answers lie uniformly in the answers
This assumption doesn't seem right, because age of HN users is not uniformly distributed, and we can assume (yes, I'm exchanging one assumption with another ;)) that people who lie won't pick their own age as a "lie". Therefore the biggest age group should be underrepresented, and the others, on average, overrepresented. :).
And the joke answers are likely to get picked a lot more, like under 10 or over 90. So that doesn't give us any information as to how many answers are likely fake for the others, except that we can assume it's probably a lot less.
On the other hand you can vote more than once for some reason, so some people could have just voted for every age. If you couldn't vote for more than one thing though, the obviously-untrue answers would filter out a lot of the noise.
This makes me wonder about the polls that say 18% of Americans believe the Sun goes around the Earth. Is it really that high, or are polls inaccurate plus or minus 18%?
This assumption doesn't seem right, because age of HN users is not uniformly distributed, and we can assume (yes, I'm exchanging one assumption with another ;)) that people who lie won't pick their own age as a "lie". Therefore the biggest age group should be underrepresented, and the others, on average, overrepresented. :).