Cookies   I display ads to cover the expenses. See the privacy policy for more information. You can keep or reject the ads.

Video thumbnail
Ok, correlations and causation footnotes: In the main video I said that when you find
a correlation, it’s natural to look for explanations or causes of it. This is called Reichenbach's Principle.
But sometimes correlations occur just by chance, like those on the website “spurious correlations”
which selectively cherry-picks data points from different stats that randomly happen
to line up.
As an example of a chance correlation, if I flip two coins enough times, eventually
there’ll be a long string of matching heads or tails just by chance, and if I just cherrypick
those flips I can make it look like the coins are super correlated.
But when an apparent correlation is actually random in origin (like in this case), then
if you keep looking at larger and larger samples, the correlation should go away.
This is it sometimes looks like particle physicists have discovered a new particle, only for that
to go away when they collect more data.
Also, you may have noticed there was no mention of feedback loops in the main video – that’s
because, from a causal point of view, feedback loops, like how more grass means more sheep
means less grass means less sheep means more grass and so on – from a causal point of
view, this isn’t actually a loop.
It’s more of a chain, where the amount of grass and sheep now affect the amounts of
grass and sheep next year, and the year after and so on, so from year to year there’s
feedback between the amount of grass and the amount of sheep which we kind of draw as a
loop, but the causal relationship always goes from the present to the future, which we should
draw as some sort of spirally helix thing.