Why the Gambler’s Fallacy argument does not apply to cosmological natural selection

My email to Philip Goff, in response to his Scientific American essay, in which he argues against multiverse explanations for our complex, fine-tuned universe

A few years ago, during lockdown, I was invited to join a small email group discussing the role of consciousness in the universe (and the relationship of consciousness to the material universe). The group included people like the controversial and delightful biologist, author, and heretic Rupert Sheldrake, deep-space propulsion expert Kelvin F. Long of the Interstellar Research Centre, and panpsychist philosopher Philip Goff, and was enormous fun, with exactly the right mixture of brilliance and lunacy to spark an explosively creative conversation. (For example, Rupert’s fascinating and provocative paper, Is the Sun Conscious?, published in the Journal of Consciousness Studies, came out of these discussions.) I didn’t agree with everything my fellow group members believed or argued for, but I had a hell of a good time debating a wide range of interesting issues with a group of highly eccentric and original thinkers.

I stayed in touch with Philip, and at one point he sent me a link to his article for Scientific American, in which he dismisses multiverse theories (arguing that they can’t explain the complexity and fine-tuning of our universe). 

I wrote an email in response, which I think makes a good mini-post, and so I have pasted it in below: the argument forced me to clarify my thoughts on this subject, and articulate them succinctly. (Also, I said I would give you glimpses behind the scenes of the writing process; well, it is from email conversations like this that much of the book emerges.)

Here’s Philip’s original article: Our Improbable Existence Is No Evidence for a Multiverse.

For those of you too lazy, sorry I mean busy, to read Philip’s article, I’ll just paste in the opening to the Wikipedia definition of the Gambler’s Fallacy.

“The gambler's fallacy, also known as the Monte Carlo fallacy or the fallacy of the maturity of chances, is the belief that, if an event (whose occurrences are independent and identically distributed) has occurred more frequently than expected, it is less likely to happen again in the future (or vice versa). The fallacy is commonly associated with gambling, where it may be believed, for example, that the next dice roll is more than usually likely to be six because there have recently been fewer than the expected number of sixes.”
–Wikipedia

This is one of those situations where I think the demented wrongness and inhumanity of a ChatGPT4o/Dalle/AI image works quite well. The prompt was “Please draw for me a woodcut with hundreds of dice being thrown. Some of the dice are large in the foreground. I would like all the dice to show sixes on all their faces.” (Yeah, I tried a bunch of prompts, and there seems to be no way to get it to understand the idea of sixes on all the faces of the dice.) By the way, if you hate AI images, no problem. Feel free to continue hating them. I shall continue to enjoy playing around with them, because I like exploring important and disruptive new technologies, and trying to understand them, even if I don’t always approve of them, or their effects on the world. We can still be friends.

And here is my email to Philip…

Hi Philip,

Thanks a million for getting back to me, and for sending me that link. Apologies for not replying before, as I was suddenly inundated with other writing deadlines, and I wanted to clear some time in which to read the article, and think about it before responding.

I love the article you sent me, because I agree with every word of it. And every word of it, in fact, supports my case.

All the multiverse options you discuss involve all the other universes having essentially random properties. And all such theories do indeed fall to the gambler's fallacy.

"How many times the gambler has rolled that night has no bearing on whether the next roll will be a double six."

But if we live in an evolved universe, where the offspring universes have only slightly different fundamental properties to those of their parents, and thus differential chances of reproductive success, the number of times the gambler has previously rolled the dice has an ENORMOUSLY consequential bearing on how likely it is that we live in a complex, fine-tuned universe. The difference between two sequential rolls still won't be of any detectable significance (just as the evolutionary difference between two sequential generations of bacteria aren't of any detectable significance); but the difference after a long enough sequence of rolls where evolution is a factor at each step will be HUGE, as we see with DNA evolution.

By analogy, if you were an observer with a very limited field of view and the only thing, the only blob of matter, which you knew to exist, which you could perceive, was a giraffe – one giraffe, that's all you knew of that exists – then explaining why this blob of matter was so complex would be extremely difficult, without a theory of evolution. You could postulate that there were many other blobs of matter, inaccessible to you, and that most of them were probably much blobbier, and didn't contain kidneys and eyeballs and lungs and a heart and a long neck and so on. And that you had just been incredibly unfeasibly lucky in your particular blob of matter, in its marvellous accidental random complexity. But that's not a satisfying theory, and it is completely missing the reason for the complexity. Indeed, it's denying there is a reason. It's just a highly unlikely chance arrangement of matter.

Similarly, a theory like Wheeler's initial multiverse theory where universes give birth to randomly different universes doesn't really get you anywhere, because, yes, the gambler's fallacy kicks in. Sure, you might get a blob of ridiculously unlikely complexity at some point, but there is no reason to, ever.

No matter how long the random generation of random universes goes on, a giraffe never gets more likely. First universe in the sequence, last universe in the sequence, all equally unlikely to be complex.

But the brilliance of Smolin's tweak (and I cannot understand how nobody seems to really get this – you are not alone in writing entire articles about multiverses without mentioning it, there was similar article in Aeon last month) is that the longer the evolutionary generation of universes goes on, the more likely a particular universe will be large, stable and complex. Evolution EXPLAINS the size, the stability, and the complexity we see in our specific universe, in precisely the way a random multiverse theory cannot. (And, given that our universe contains an estimated 40,000,000,000,000,000,000 stellar mass black holes, it is clearly descended from a long line of highly reproductively successful universes!)

Of course, Smolin then fails to see the broader implications of his own theory, as they roll through field after field, but that's another conversation for another day.

Feel free to give feedback /disagree / poke holes in my argument. In fact, I would love to have a quick chat with you about this sometime, because I think it's supremely important, not just for my work but for yours.

Fond regards,

-Julian

AFTERWORD:

I should probably also mention here that Philip had a fall recently in which he fractured his skull and had two brain bleeds. Let me take this chance to wish him a full recovery (which, fingers crossed, looks very likely); I greatly look forward to arguing with him full strength again, when it is once more safe for him to shake his head in violent disagreement with me.