I Guess All of Us Night Owls Are Going to Die, Well Maybe Not

I’m sure you’ve all seen the headlines. A recent study has shown that people who identify as night-owls, were 10% more likely to die early or some variation on that.

Apparently, we’re all supposed to start being morning people for the sake of longer life, except that’s not exactly what the study proved at all.

I’m going to use some quotes from this Popular Science description of the study, which is a bit less dramatic as some other online articles I’ve seen about the study.

That sounds scary, sure—but there are a few limitations worth considering. For one, says Knutson, “we weren’t able to pinpoint and find out why night owls were more likely to die sooner,” so the direct cause of mortality is unknown, creating some murkiness as to what extent night owl lifestyles influenced those deaths.

So what we have is a classic case of “correlation does not equal causation”. Yes, those who identified as night owls were sicker in general, and died more often than those who didn’t, but there’s absolutely no explanation as to why. So, we’ll guess…

“We think,” says Knutson, “it is at least partly due to our biological clocks. We think the problem is that the night owls are forced to live in a more ‘lark’ world, where you have to get up early for work and start the day than their internal clocks want to. So it’s a mismatch between the internal clock and the external world, and it’s a problem in the long run.”

Again, this is only a guess. It may not be staying up late that’s killing us, but getting up early when we aren’t morning people, because the rest of the world believes that early risers are successful. Of course, there’s also this reality:

Unfortunately, the Biobank data only indicated whether someone identified as a morning or evening person, not whether they had a sleep schedule that suited their chronotype. “We know what they’re preferred time to sleep is, but we have no idea what they were actually doing on a day-to-day basis,” says Knutson.

This study didn’t actually go into detail to find out who was actually staying up late, versus who answered a survey and described themselves as night owls. Since we know people are generally not good at knowing themselves, I’m going to go ahead and take all of this with a grain of salt. Also, since the study was conducted in one particular place, it may or may not hold true elsewhere.

Moreover, the data is limited to just British participants, most of whom were caucasians of Irish or English descent. It’s likely the results would be similar for other populations in the Western world, but they could also be substantially different for night owls elsewhere.

What we have here, and the reason I’ve gone to all this trouble, is a situation where there may be some actual causation in there somewhere, but it could also be simple random chance. As we move into the Big Data and AI era, I suspect we’re going to see a lot more correlations, because AI is going to be extremely good at analyzing a ton of different inputs and finding connections. (The stock market mostly goes up on days when the temperature on Wall  Street is between 62-65 degrees at noon, or the Dodgers have a winning record on Tuesdays in May.) I just made up those examples, but these are the kinds of things AI is going to find, and that’s about all it tells us. They are random. There’s no causation there. We would be foolish to act on it going forward, because the randomness is probably temporary. But we see it, and we become biased in our thinking because of it.

I fear that AI systems built around these analytics are going to cause grave harm to people, as we “see” more correlations that are truly just random, and start creating biases around these “facts”.

Heck, we already sensationalize the heck out of the correlations we find today. That’s not going away any time soon.

Follow these topics: Tech