Because it is a lot of indirect reasoning. Literally, decades of occasional information. Even weak patterns can become visible after enough exposure. I have learned even before finding LW that underconfidence is also a sin.
As an analogy: if you throw a coin 10 times, and one side comes up 6 times and other side 4 times, it does not mean much. But if you throw the same coin 1000 times, and one side comes up 600 times and other side 400 times, the coin is almost surely not fair. After many observations you see something that was not visible after few observations.
And just like I cannot throw the same coin 10 more times to convince you that it is not fair (you would have to either see all 1000 experiments, or strongly trust my rationality), there is nothing I could write in this comment to justify my probability assignment. I can only point to the indirect evidence: one relatively stronger data point would be the relative consensus of LW contributors.
Sure, lots of pieces of weak evidence can add up to strong evidence… provided they’re practically independent from each other. And since this issue gets entangled with Green vs Blue politics, the correlation between the various pieces of weak evidence might not be that small. (If the coin was always flipped by the same person, who always allowed to look which side faced which way before flipping it, they could well have used a method of flipping which systematically favoured a certain side—E.T. Jaynes’s book describes some such methods.)
That is, if you say to me “I flipped this coin 1000 times and recorded the results in this Excel spreadsheet, which shows 600 heads and 400 tails,” all I have to believe is that you really did flip the coin 1000 times and record the results. That assumes you’re honest, but sets a pretty low lower bound for your rationality.
And yet it leads you to a 99% probability assignment. :-/
Because it is a lot of indirect reasoning. Literally, decades of occasional information. Even weak patterns can become visible after enough exposure. I have learned even before finding LW that underconfidence is also a sin.
As an analogy: if you throw a coin 10 times, and one side comes up 6 times and other side 4 times, it does not mean much. But if you throw the same coin 1000 times, and one side comes up 600 times and other side 400 times, the coin is almost surely not fair. After many observations you see something that was not visible after few observations.
And just like I cannot throw the same coin 10 more times to convince you that it is not fair (you would have to either see all 1000 experiments, or strongly trust my rationality), there is nothing I could write in this comment to justify my probability assignment. I can only point to the indirect evidence: one relatively stronger data point would be the relative consensus of LW contributors.
Sure, lots of pieces of weak evidence can add up to strong evidence… provided they’re practically independent from each other. And since this issue gets entangled with Green vs Blue politics, the correlation between the various pieces of weak evidence might not be that small. (If the coin was always flipped by the same person, who always allowed to look which side faced which way before flipping it, they could well have used a method of flipping which systematically favoured a certain side—E.T. Jaynes’s book describes some such methods.)
Or your honesty.
That is, if you say to me “I flipped this coin 1000 times and recorded the results in this Excel spreadsheet, which shows 600 heads and 400 tails,” all I have to believe is that you really did flip the coin 1000 times and record the results. That assumes you’re honest, but sets a pretty low lower bound for your rationality.