The Things That Don't Fit

Yesterday, ScienceDaily reported on a survey of hundreds of astrobiologists. It found remarkably little agreement about whether recent observations constitute evidence of extraterrestrial life.¹ Researchers at Durham University asked scientists to assess claims concerning K2-18b, an exoplanet about 120 light years away, and Cheyava Falls, a rock on Mars. Only 6.6 per cent thought the biological molecules that might have been detected in K2-18b’s atmosphere »probably represented« alien life. Roughly two-thirds disagreed, while the rest were undecided. The scientists were more persuaded by the evidence from Cheyava Falls, although there was still no consensus.

»A trace tells us that something happened. It doesn’t tell us what happened«

Scientists disagree all the time. The interesting thing here is that they’re looking at the same evidence and disagreeing about what kind of thing they’re looking at. It’s a classification problem, and classification seems straightforward. We see something, compare it with things we already know, and put it into the appropriate box: rock, bird, machine, bone. But what happens when the box doesn’t yet exist?

Let’s say someone finds a footprint in the forest. We establish it exists. We measure it, photograph it, analyse its shape. But if nobody has ever seen the animal that made it, the footprint doesn’t come with a »Yeti« label. It might indeed be a Yeti. Or it might be a bear, a human, a geological accident, or a hoax. The Yeti footprint and the K2-18b molecule have the same logical problem. A trace tells us that something happened. It doesn’t tell us what happened.

This is the basic idea behind Bayesian reasoning, developed by the mathematician Thomas Bayes in the 1760s.² Suppose we find the footprint and want to know whether a Yeti made it. Then what we shouldn’t ask is whether a Yeti could make such a footprint. We should ask whether something else could make it too. Bayesian logic lets us weigh a hypothesis against its alternatives, rather than judging it in isolation. Evidence becomes powerful when the observation is much more likely if our hypothesis is true than if it is false.

The same problem applies to K2-18b. Finding a molecule associated with life on Earth is not the same as finding a molecule that could only plausibly have been produced by life. The question is not whether life could produce the observation, but how many non-biological explanations remain. If we don’t know the possible causes, we don’t even know the size of the box we’re trying to classify the observation into.

This helps explain why the evidence from Cheyava Falls was more persuasive to the scientists. There’s an actual piece of Martian rock to examine: its minerals, textures, chemistry, and geological context. The K2-18b finding is much more indirect: a spectrum of light that has passed through a distant atmosphere. That gives us many more steps between observation and explanation.

Not every strange encounter is a classification problem. This year’s Ryan Gosling film Project Hail Mary gives us the opposite case. Its protagonist encounters Rocky, an extraterrestrial organism radically unlike terrestrial life. Yet Rocky presents little difficulty as a classification problem: it communicates, responds, behaves like an agent. The alien is strange, but its category is not.³ K2-18b reverses this. We may have detected something produced by life, without being able to observe the life itself.

So what would we actually need before we could claim to have found alien life? The first requirement would be that the observation survives independent measurement. Then that plausible non-biological explanations have been tested. Ideally, several independent clues would converge. Stronger still would be a biological explanation that makes predictions: if this is life, we should also find x, y, and z. If those predictions came true, the evidence would become much harder to explain away.

This is where the astrobiologists’ disagreement gets interesting. It’s tempting to read the 6.6 per cent as a failure of evidence, and the two-thirds rejection as scientific caution. But the disagreement may be telling us something sharper: the observation is real, and the classification simply isn’t settled. The pitch-black monolith in director Stanley Kubrick’s 2001: A Space Odyssey is the purest fictional version of this problem.⁴ Nobody in the film doubts that it exists. What nobody can settle is everything else: what made it, where it came from, why it’s there. It’s an object that’s absolutely real and completely unclassifiable at the same time.

References