I think the main problem is piling up the estimates. If I use anthropic reasoning to conclude that I am from an extremely populous country like India, I am wrong, but 2 billion Chinese and Indian people would be right, and the majority of people making the estimation would in fact be from larger than average countries
But if someone starts drawing conclusions like "due to the correlation between population size and landmass, it is therefore unlikely any country is more than 20% larger than mine", there is a 98% chance they are not Russian and therefore incorrect (and there is a 100% chance that their country is dwarfed by either population or landmass by some other country). The probability of them being the largest country is roughly the same as the probability of them being from a small island group like the UK or Japan
If the principle falls apart when applied to the human populations whose population density relationships its estimated from, how can we assert 95% certainty that the circumference of planets supporting very different civilisations will be no more than 20% greater than Earth?
> "due to the correlation between population size and landmass, it is therefore unlikely any country is more than 20% larger than mine" there is a 98% chance they are not Russian and therefore incorrect
Your mistake was to stop using statistics to refer to groups forming a distribution, and referring to ONE particular country, Russia.
The argument only works when you keep things in the realm of statistical distributions. An argument that works should be:
"due to the correlation between population size and landmass, it is unlikely most other countries are more than 20% larger than mine".
You would be right almost every time! Yes, it's a weaker argument but still extremely interesting.
It's a weaker and more plausible claim, but it's also a different claim from the one advanced by the author "we can say with 95% confidence that another planet with intelligent life, such as our nearest neighbour, will have a circumference no more than 20% greater than that of the Earth".
As far as I can see we can't even predict that for distributions of individuals on earth (the article suggests the median human lives in a country as populous as Pakistan; a random other human has an 18% chance of being Chinese which is quite a bit more than 10x the size of Pakistan by landmass) and that's long before we add ancillary assumptions like alien species' size distribution matching earth's and their tolerance for population density being no greater than mean human population density (which requires them to be less tolerant of dense populations than many self-sufficient human regions!)
"we can say with 95% confidence that another planet with intelligent life will have a circumference no more than 20% greater than that of the Earth".
The author shouldn't have added "such as our nearest neighbour" as that just adds confusion.
The fact that this prediction is not 100% accurate when considering Earth's countries does not invalidate the argument. I see a lot of people doing this: showing one practical example where it doesn't work and calling it BS.
Please try to do as the author suggested: plot your own data against many world statistics... you will see that while yes, some of those statistics fail for you (e.g. you might be 90% taller than everyone else) when taken all together, they should all indicate you're pretty close to the middle in the majority of them... and knowing this, hopefully you can see how, yes, this prediction by the author might be BS, but given what we know, it's the only prediction we can make which has a good chance of being true.
> The fact that this prediction is not 100% accurate when considering Earth's countries does not invalidate the argument.
I mean, it does , because he's moved from making general applications of the anthropic principle to very explicit claims about confidence intervals, relationships between variables and shapes of the distributions which don't match even the figures for the part of reality that actually is observable. I'm not just saying "but there are exceptions". I'm saying "given actual human population/landmass distributions, it appears obviously wrong to say that a random person has a 95% probability of living in a country no more than 20% larger than the home country of a randomly selected person from another country, and so a claim the confidence interval is that narrow for planets and alien species is quite extraordinary"
(it's moot that much of my own data also works pretty badly for anthropic reasoning, and I'm almost convinced that the applicable version of the anthropic principle for some of those stats is "if an individual is willing and able to make observations about the anthropic principle, the probability they are exposed to Western culture and in the top decile for access to education, cash and free time is ~1" :-) )
> given actual human population/landmass distributions, it appears obviously wrong to say that a random person has a 95% probability of living in a country no more than 20% larger than the home country of a randomly selected person from another country...
You are just refusing to believe statistics has any value. It's just maths, not opinions.
I mean, when we say a fair coin will turn up heads exactly 50% of the time, we mean it in a mathematical way.
Do it 3 times, and you might get 3 tails... based on your argument, the theory should be put into question... but of course it's not how maths works.
Do it 1000 times, and you will see how the normal distribution becomes aparent, with a mean nearly exactly a 0.5. Do it 1 million times, and the mean becomes even more evident... you can keep going on forever, it will ALWAYS become more evident. This is not useless just because it doesn't work after a few times. You can be extremely confident it would work after a large amounts of tries.
Thanks, that makes sense. Do you know if there is any "mathematical trick" to quantify the likelihood of me being wrong (without knowing the population numbers of the other countries)?
But if someone starts drawing conclusions like "due to the correlation between population size and landmass, it is therefore unlikely any country is more than 20% larger than mine", there is a 98% chance they are not Russian and therefore incorrect (and there is a 100% chance that their country is dwarfed by either population or landmass by some other country). The probability of them being the largest country is roughly the same as the probability of them being from a small island group like the UK or Japan
If the principle falls apart when applied to the human populations whose population density relationships its estimated from, how can we assert 95% certainty that the circumference of planets supporting very different civilisations will be no more than 20% greater than Earth?