Last week my former PhD supervisor Geraint Lewis visited us here in Auckland. The main reasons were to work on gravitational lens modelling with myself and my PhD student David Huijser, and also to attend a cosmology conference. Those things went well, but as per usual we also drifted off into many conversations about more crazy ideas to do with life, the universe and everything.
Years ago a paper came out by Charlie Lineweaver and Tamara Davis, about the formation of life on Earth. There is a popular argument that goes, “We know life formed pretty early on in terms of Earth’s history. Therefore abiogenesis is easy and life is common in the universe”. The Lineweaver and Davis paper made this argument quantitative, so we could say exactly how strong the evidence is in this direction. Unfortunately the paper also contained an unintentional highly informative prior that made the argument much stronger than it should have been. So I wrote up my response which was rejected by two journals. The basic conclusion is that the early formation of life does provide some evidence that life is “easy to form”, but not enough to convince a skeptic who thinks life is extremely rare. Some other authors independently came to the same conclusion.
Geraint is interested in generalisations and more realistic versions of this argument. But I’m not confident in any model that I write down until I understand the anthropic complications surrounding the argument (we couldn’t find ourselves on a planet where life didn’t form, etc). So I’ve given myself the task of reading and understanding Radford Neal’s tour de force on the subject before I deign to try anything.
Plausibility Theory 
I have a growing worry about Neal Radford’s article, though it might just be my own misunderstanding. I was going to ask you about it when next you appeared in Sydney. Here goes.
It seems like he wants to say that if someone want to calculate the probability of “I am going to roll a six”, in order to specify “I”, one must give enough information to pick out exactly that individual in the universe. One could imagine an angel who visits our universe and needs to calculate the probability that I will roll a six. To know who this I is, we need to start specifying the exact arrangements on the curtain and the location and speed of the ant in the kitchen so that the angel don’t get mixed up with someone who just looks like me on a planet very much like Earth elsewhere in the universe.
This is why Neal is so troubled by very large and infinite universes – no amount of information about my surroundings is enough to specify which “Luke Barnes” I am.
We cannot rely on “individual essences”, says Neal, to pick out an individual as these are mysterious and unscientific. This is why Neal says that the information is non-indexical – there must not be an “I” in probability sentences.
But this doesn’t seem like a reasonable restriction to me. There is nothing mysterious about indexicals – I, here, now, today etc. Their novelty is only that they pick out the speaker (or speakers location, or speaker’s present time) of the sentence in question. The angel need not examine every room and person and ant in the kitchen. She need only ask who the speaker is.
So I think that indexical information is permitted for probability theory, contrary to what Neal says. Apart from that, I agree with a lot of the article. In particular, I can see the advantages of a Bayesian approach over SIA+SSA-style reasoning.
“The angel need not examine every room and person and ant in the kitchen. She need only ask who the speaker is.”
I’m inclined to agree, but my grasp of all the details isn’t very strong. Presumably if the universe is sufficiently large there will be more than one speaker, which might create an ambiguity which would be resolved by FNC conditioning on “there are N very similar people calling themselves ‘I’ and asking a particular question about the universe”, or something. I’ll need to think about it more.
Getting away from SSA and SIA has always seemed the most compelling point of the paper IMO. I used to think the Doomsday argument was correct, based on what I didn’t realise was SSA-SIA, but I don’t think that anymore.
I will be appearing in Sydney for the first two weeks of February, and would be happy to discuss this more if you’re around. I feel like half an hour chatting would probably be more effective than hours of solitary thinking.
Good timing: http://arxiv.org/pdf/1412.4382.pdf
The more I read on this topic the more I feel like it’s just not my area.
It’s a plan!
Hi Brendon. The questions regarding the origins of life on Earth and life in the Universe are compelling. Perhaps this is a bit off-topic since one of the main aspects of the paper you are considering is how to reason in the face of the anthropic principle and a paucity of information.
Here are some references to other works as well as my half-baked thoughts related to life in the universe.
Bob Zubrin has an excellent article making the case for panspermia
R. Zubrin (2001), JBIS, 54, 262-269 Interstellar Panspermia Reconsidered
http://www.jbis.org.uk/paper.php?p=2001.54.262
where he notes that the fact that life arose in such a short time after the bombardment phase, and that there are no pre-bacterial forms present (along with arguments about time scales required for panspermia and the ability of bacteria to remain dormant in space for tens of millions of years) suggest that life arose elsewhere.
While, David Thomas later notes that much of Zubrin’s argument relies on the missing evidence of pre-bacterial life forms, which may either be represented by the nanobacteria discovered in the last decade (and possibly present in Martian and other meteorites), or by organisms that have not yet been detected on Earth (alive or in the fossil record), I find a few other observations by Zubrin to be compelling.
https://www.academia.edu/873799/Interstellar_panspermia_reconsidered_again_in_defense_of_life_s_evolution_on_Earth
First, one would expect (and perhaps this expectation is wrong) that it is far more difficult to create a bacterium out of an organic soup than it is to create a multicellular organism out of a colony of single celled creatures. If this expectation is correct, then it is shocking that bacteria appeared on Earth within 300 million years after the bombardment phase, but then took 2 billion years to evolve into eucaryotes with simple plants and animals appearing another 1 billion years later. This could be explained by the fact that at the sub-bacterial level the relevant time scales for replication, experimentation and exploration are rather short, but in the case of plant and animal reproduction, the time scales are much longer slowing the evolutionary rate. Regardless, some prior information here may be of use in performing the inferences intended.
Another consideration is that (aside from viruses, which may be relevant) there is no evidence that abiogenisis has happened here since the formation of the first life forms. It may be that a biotic environment tends to prevent the establishment of any newly abiotically generated life. Despite this, one might expect a dramatically different abiotically generated lifeform to not directly compete with bacteria and establish a foothold. Either way, a biotic environment is simply an environment and both the climactic and biotic aspects of Earth’s environment have gone through dramatic changes over the last 4 billion years. It may be that the fact that abiogenisis has apparently only happened here once might be able to put some reasonable constraints to the time scale at which one would expect abiogenisis to occur in general.
This suggests to me that it is indeed very possible that the abiogenisis characteristic timescale mu is very long. This could mean that Earth was lucky with life evolving very early on, or life arrived at Earth a short time after the bombardment period.
If the abiogenisis characteristic timescale is relatively long, this would imply that the generation of life in the universe is slow. However, the probability of panspermia may actually be appreciable, which means that life would only have to take hold in a few places to populate the galaxy since infection is exponential. For example, Zubrin estimates that if an average star system with a planet like Earth could infect one other planet in another star system every 100 million years (that is 35 planets in 3.5 billion years) and this process was repeated, then there would be 2^35 = 34 billion planetary systems with life. Three to five nucleation points occurring 3.5 billion years ago could conceivable populate a fair portion of the galaxy by now.
Thanks Kevin, lots to mull over!
Thanks for this elaborate comment Kevin.
Pingback: Inferences about Life in the Universe | Elata Fundamenta