Correlation is not causation… but what about an absence of correlation — what does that tell us?
(This post is a continuation of my three-part series on the Correlation/Causation flaw, one of the Big Three LSAT flaws. If you haven’t already, check out part 1 and part 2.)
For all the details please see the video version of this post, in which I go into more detail, but here are the basics: The fact that two things are not correlated with each other may imply that neither of them causes the other, but on the LSAT an argument that assumes as much is flawed. Moreover, the issue in such problems is almost always that there might actually be a causal relationship, but it’s being masked or hidden by something else.
That “something else” takes various forms, but the most common one is at bottom a sampling problem: The two groups we’re comparing aren’t actually similar, and the difference makes it harder for us to see causation.
An example will help: Suppose that the Orinoco Company has two facilities, one at Braceville and one at Freetown. We are told that at Braceville, employees have been required to wear backbraces for the past year while at Freetown there has been no such requirement. We’re also told that back injuries are not less common at Braceville than in Freetown. The conclusion of the argument is that back braces don’t help prevent back injuries.
What could be wrong with this argument? Well, it could be that our measurement of no correlation doesn’t really help, for example because the policy at Braceville has been effect for only a year but back injuries take five years to show up. Or perhaps Braceville employees don’t actually follow the policy. But usually on an LSAT problem this sort of thing won’t be the issue. Instead, what we’ll see is the possibility that back braces really do help prevent back injuries but for some reason it doesn’t show up in the stats, probably because without the backbraces the Braceville employees would look even worse. Maybe Braceville is a warehouse in which employees lift lots of heavy objects while Freetown is an executive facility at which employees don’t lift anything heavier than a coffee cup. Maybe employees who are prone to back injuries are preferentially assigned to Braceville. Or maybe the Freetown facility teaches employees safe lifting techniques but Braceville has no such training.
This sort of not cause argument is fairly predictable and varies much less than the usual correlation/causation type. The conclusion is almost always explicit to the effect that A does not cause or not help with B; the issue (whether we’re weakening the argument, strengthening it, or finding a necessary assumption or a flaw) is usually whether causation really is there, producing correlation, but something else, something that cuts the other direction, masks the effect.
Watch the video (and its two companions) and let me know what you think.