Firstly, a thank you to Berian for graciously allowing me to post here. Secondly, an apology: this first post is a shocking six months late. This was going to be an exclusive scoop for an exciting, current science item that was making waves in the blogosphere, and sparking passionate responses of every shape and hue. Instead, this post comes to you after the fact, a retrospective look at research that made the headlines…
The story begins with an idea. It came to me as most ideas do, in a place and time when you should be thinking about something else. I was sitting in a lecture about the Search for Extraterrestrial Intelligence (SETI), and its current successes/failures (feel free to comment on them!). I had been working on Monte Carlo methods at the time: for those not in the know, Monte Carlo methods use stochastic (probability-driven) techniques to attack problems where other analytic or numerical methods cannot gain a foothold.
It then occurred to me that Monte Carlo methods could be applied to questions of SETI. We could create a mock Galaxy, with stars and planets, and watch life’s progress on a planet-by-planet basis. Of course, we don’t have a complete census of every star and planet in the Galaxy, but we do have enough information to construct probability distribution functions (PDFs) for each variable: stellar mass, galactocentric radius, planetary system architecture, etc. Every time we create a new star system, we sample from these PDFs so that the mock Galaxy is statistically identical to ours. Then, we can evolve life using stochastic equations, which lets us account for extinction events and other probabilistic elements of the evolutionary process.
By running the code several times, we could ascribe sampling errors to our results (albeit massively underestimating them). Of course, we don’t know how life forms on other worlds, but we can compare different theories of Life and Intelligence on a level playing field, with realistic potential niches and environments (see the paper for more).
You may have spotted the main weakness of the code: because the method relies on observational statistics as input (the distribution of stellar masses, the structure of the Galaxy, exoplanet statistics, etc.), the simulation of potential niches is biased by our ignorance. At the time of press, we’ve identified less than 400 exoplanets, and the identification process itself is fraught with bias. That aside, even if we could input every star and planet in the Galaxy individually, we still clutch at straws when it comes to simulating Life’s origin, and are forced to make broad assumptions that, although sensible and based on good science, could easily be wrong.
However, this crisis is also an opportunity: the results will improve as our data improves. As we learn more about the planets (and how life on Earth came to be) we can quickly feed that knowledge into the code, and gradually move towards the truth. The algorithm is an alternative to the Drake Equation, with the added advantage that spatio-temporal information (great phrase) is more intuitively incorporated. An alternative, not a replacement – we need to use all the tools at our disposal, and the Drake Equation is beautifully simple, and easy to use (and life is being breathed into it by Claudio Maccone in the form of the Statistical Drake Equation).
This work resulted in a maelstrom of media attention: I’ll save further thoughts on my experiences with the public/media for another post. The field of astrobiology is not without controversy: I encourage you to debate with me via comments. Thanks for reading!
Letters to Nature 
Welcome, Duncan! It’s good to read a bit more about this work, and even though the priors (I like to call inputs priors, because, after all, everything is prior to something else; I expect Brendon will have more to say about that) are somewhat unconstrained at present, the method certainly seems extensible, so we should keep an eye out in the future.
I like your selling the method as an alternative to that awful Drake equation, though I’ve not heard of this Maccone chap—what’s he about? Will you be meeting him when you go to Prato next week?
Very interesting. I agree with Berian (and xkcd: http://xkcd.com/384/) about the Drake equation.
I’m very interested in how you place life on planets. Presumably, a given planet is assigned a timescale for the evolution of intelligent life. Or at least a timescale is chosen from a distribution that depends on the suitability of the planet for life. That part seems to be the real weakness to me. Assumptions about planets aren’t going to change too drastically as data improve, in my opinion.
I’d be interested in your opinion on the section on the evolution of life in “The Cosmological Anthropic Principle” by Barrow and Tipler. If I remember correctly, they argue that the fact that
time to evolve intelligent life (=4Gyr) ~ lifetime of sun (=10 Gyr)
suggests that the average timescale to involve intelligent life is much longer than the time it took on earth (i.e >> 4Gyr). The lifetime of stars applies a selection effect – any intelligence that manages to evolve before its star explodes will observe 1.) That they arrived near the end of the life of their star 2.) That they evolved unusually fast.
At least, I think that’s what they said. Any thoughts?
Thanks for the comments, guys. I think that the Drake Equation is elegant, but kind of difficult to use (how do you ascribe values to the parameters without massive generalisation?). I would compare it to the Ancient Greek’s attempts at natural philosophy: aesthetically gorgeous, but archaic (and almost useless).
Berian: Claudio Maccone is a well established SETI researcher, who has upgraded the Drake Equation by allowing each parameter to be a random variable. If you go nuts and allow the equation to have an infinite number of terms, and take the log, then you get a sum of random variables, and voila! The Central Limit Theorem lets you identify the distribution of N as lognormal. A lovely piece of statistics, (even if we can’t ascribe means and variances to the random variables). It also shows that the separation of civilisations follows a similar distribution!
Luke: We allow planets to have life if they satisfy certain criteria, which are user-defined. The typical criterion is the stellar habitable zone (a range of orbital radii where liquid water can exist on the planet’s surface). Others can be specified too (star masses, planet masses being within a certain range, etc). I would point you to the paper for the stochastic evolution algorithm, as it gets a little complex and long-winded.
Your stellar lifetime selection effect applies here too: we apply the Biological Copernican Principle (and assume that life on Earth is not abnormal), so life takes around 4Gyr to evolve intelligence (with a spread defined by the stochastic evolution algorithm). This implies that intelligence will only arise on stars below a certain mass (otherwise the stellar lifetime is too short).
Of course, it’s dangerous to apply selection effects (or any of the above arguments) on one sample! Hopefully, that will not remain the case for too long.
>>I’d be interested in your opinion on the section on the evolution of life in “The Cosmological Anthropic Principle” by Barrow and Tipler. If I remember correctly, they argue that the fact that
time to evolve intelligent life (=4Gyr) ~ lifetime of sun (=10 Gyr)
suggests that the average timescale to involve intelligent life is much longer than the time it took on earth (i.e >> 4Gyr). <<
This is basically correct – however, it's not really evidence that the average timescale is extremely long per se, just that it cannot be short. The likelihood function from the data of 4Gyr is basically flat all the way from ~10 Gyr all the way out to infinity.
You're right about selection effects. It's very hard and it's quite easy to get the wrong answer in this field! A worthy paper that I recommend to anyone interested in anthropic selection effects is the one by Radford Neal (search for radford anthropic on arxiv – apparently an update is in the works too).
Does the code estimate time scales based on any particular theory of chemical evolution? RNA first vs. the Cairns-Smith theory of clay catalysation? I’d guess that these theories aren’t exactly the most quantitative …
What is the average timescale, for an Earth like planet in your model? How does it compare to 4Gyr?
[…] 24, 2009 by dh4gan I promised in my first post to talk about my experiences with the press (and I now promise to talk about something non-aliens […]
Sorry for the belated replies to your comments. We don’t strictly model which route life takes to form. What is hardwired into the code is Copernican mediocrity: we assume that Earth is a typical biosphere, so the evolutionary timescale comes in (typically) at around 5 Gyr (this fluctuates with the number of resetting events that the planet is suffering).
If you allow resetting mechanisms to take effect, then the evolutionary timescale (without resets) can become very short (if the resetting frequency is high, to prolong the actual timescale). There’s been some really good work in showing that biological timescales can be regulated by these resets to correlate with astrophysical timescales.
[…] } I promised in my first post to talk about my experiences with the press (and I now promise to talk about something non-aliens […]