“Fortunately for serious minds, a bias recognized is a bias sterilized.” -Benjamin Haydon
You might look up at the night sky, at the vast canopy of stars we can see, and ask, exactly, what we’re seeing?
Thousands upon thousands of stars, of course, even with just your naked eye. These stars come in all sorts of different sizes, temperatures, and distances, and what we see in the night sky is largely determined by a star’s brightness and distance from us.
All the stars you can see with your naked eye belong to one of the seven color classes of stars.
With dark skies, one can easily — even with just a naked eye — tell apart the bright stars by their different colors. It turns out that all of the colors — from the red and orange M- and K-stars to the bright blue O- and B-stars — are well represented in our night sky.
But when we perform large surveys of the stars, such as in our solar neighborhood, we find something else entirely.
Within twenty light years of our Sun, there are just over 100 stars.
And exactly zero of them are O-stars.
And exactly zero of them are B-stars.
But if we look at the K- and M-stars, they make up over 90% of these stars. In fact, three out of four stars are the reddest, coolest, M-class stars, including the closest star to us.
That would be Proxima Centauri, at just 4.2 light years distant from our Sun. The closest M-star to us, Proxima Centauri is invisible even with binoculars, and even with dark skies, a small, 3″ telescope would unable to find it.
Because nearly all M-stars are incredibly dim, less than 1% as luminous as the Sun. Compare that to the O-star Alnitak, one of the stars on Orion’s belt.
Despite being 700 light years away, Alnitak is one of the brightest stars in the sky (coming in at #31), because it’s about 100,000 times as luminous as the Sun!
But O- and B-stars are incredibly rare, with less than 0.2% of all stars making up those types, while about 75% of the stars in existence are M-stars. Yet if we relied solely on our naked eye observations, we see far more O, B, A, F, and G stars than are actually there. Why is that? Because of bias.
In the case of nearby stars, we see more of the stars that are easier to see, because our eyes are biased towards bright stars.
Bias works in other ways, too. Take a look at the Whirlpool galaxy, above. This visible light image traces out very clearly the stars in Messier 51. And you might think that we are seeing the light from the brightest stars — the O- and B-stars — that make up this galaxy.
But that’s not true.
This is where the O- and B-stars live! O- and B-stars, because their lifetimes are so short, trace out the regions of galaxies where stars are newly being formed. By looking in ultraviolet light, we only see the hottest, bluest stars. As you can see, they cluster together in a very different manner than the stars in the visible light images do.
Well, guess what? On the largest scales in the Universe, the way galaxies cluster together is also biased!
The above simulation shows how matter in the Universe — based on what we know of gravity, dark matter and normal matter — ought to cluster on large scales.
But that is not necessarily the same as how bright galaxies cluster together!
Moreover, different types of objects — quasars, blue compact galaxies, or luminous red galaxies — have different biases from one another.
In other words, when you look at how galaxies actually cluster together, it is related to, but not equivalent to the underlying distribution of total matter.
After all, look at the difference between what the gas (made of normal matter) and dark matter (which isn’t) do in a video simulation on the scale of the Milky Way galaxy!
When you’re looking for extra-solar planets, don’t be surprised that we find the biggest ones and the ones closest to their parent star: that’s also bias.
And when you look out at the large scale structure of the Universe, don’t be surprised that the brightest, most luminous galaxies cluster together disproportionately largely compared to the underlying dark matter density! The entire Universe may be biased, but as we learn exactly how, we can use this biased data to figure out exactly what it’s doing!








