What Role Do Probabilities Play In Explaining Things?

I am engaged in an interesting debate on my blog about the nature of probabilities and past events. I am coming to the brilliant Cadre bloggers and readers for some help.

The topic of my friendly debate is the origin of life (i.e. abiogenesis). Here is my claim. When it comes to explaining the cause of event 'X', it is natural to consider the likelihood of competing explanations, and eliminate the ones that have low likelihoods (i.e. probabilities).

This seems like an intuitive point. However, my debate opponent refuses to concede this point.

I offered the following analogy.

Event 1. Lucy wins the powerball lottery in 1995.

Conclusion: wow, isn't Lucy lucky?!

Event 2. Lucy wins the powerball lottery in 1996.

Conclusion: whoa, what are the odds!! Lucy is really a lucky person.

Event 3. Lucy wins the powerball lottery in 1997.

Conclusion: okay ... this is starting to get a little fishy. Maybe Lucy has a little more going on than just luck.

Event 4. Lucy wins the powerball lottery in 1998.

Conclusion: okay ... that is it. Let's launch an investigation. Lucy is cheating ... she has to be.

What drove me to that conclusion? The probability of Lucy winning powerball once is remote ... but four years in a row go beyond remote. It no longer suffices to attribute the fact that Lucy won four years in a row to luck.

I am comparing the odds of two things ... the likelihood that pure random chance as an explanation works, and the likelihood that it does not.

My friend responded:

Further, the lottery example doesn't fly too well anyway, because the "winning numbers" would have to have been specified by Lucy a priori. The better comparative example for design versus evolution is rather than pointing out that Lucy won in those four years, just say that the numbers in 1995 were 32-14-23-08-15-02, and in 1996 the numbers were 21-18-07-11-23-14, and in 1997 the numbers were 01-17-34-29-27-06, and in 1998 the numbers were 30-14-18-07-19-26. The odds that those four sets of numbers should be picked are astronomical. But what you ignore here is that four sets numbers had to be picked. The probability argument against evolution is akin to hitting a golf ball and saying "Wow, of all the zillions of blades of grass it could have landed on, how unlikely that it should land on this one," while ignoring the fact that it had to land somewhere.

And this bears repeating about the Lucy example: that the likelihood of her cheating is much greater than the likelihood of her legitimately winning four times does not prove anything. It's merely suggestive, and gives us a place to start looking for evidence. But until we find such evidence, we really don't know anything at all about it.

--------------------------------------------------------------------------------------------------------

What do you think? Do probabilities play a role in inferring (or rejecting) explanations or don't they? Should we just accept that the unlikely happened? Is my lottery example flawed?

What say you?

### Comments

Yes ... he has mixed the two.

But for the sake of argument, let's say he did not mix them up.

Is the lottery example flawed? Is it a fair analogy that at some point, the unlikelihood of a given explanation leads the investigator to abandon that explanation as viable?

Second question ... why are you not a fan of probability arguments?
Weekend Fisher said…
So his argument, on the lottery front, is "someone had to win". And yeah, he's playing a little fast-and-loose ignoring that the same person is always winning but then again it's just an analogy for your point.

If his argument is that "someone had to win", then he should be able to demonstrate how it applies to abiogenesis, that life-from-nonlife was "bound to happen" in the same way that somebody winning the lottery was bound to happen.

Take care & God bless
WF
BK said…
I think your friend misses the boat on two points. In the area of design, as WF points out, there is no reason to believe that the arising of life was bound to happen. Moreover, the odds are certainly relevant to discovering the best explanation for an event if the event requires a series of steps each one of which is highly improbable.

Look at his example. First, he says:
"you ignore . . . that four sets numbers had to be picked. The probability argument against evolution is akin to hitting a golf ball and saying 'Wow, of all the zillions of blades of grass it could have landed on, how unlikely that it should land on this one,' while ignoring the fact that it had to land somewhere."

But the fact that it landed on a blade of grass out of a million possible landing points is not what design looks to. That is akin to the old idea of drawing the bullseye around the place where the arrow lands and saying: "See, it wasn't design that it landed there." However, design says "for life to occur event A, event B, event C, event D, event E . . . event Z, all had to occur and each is highly unlikely (many of which are as unlikely as picking out a particular blade of grass to land a golf ball on). Now, if each of those events were a 1,000 to 1 shot, then what are the odds of all of them happening naturally?"

He closes with the idea "the likelihood of her cheating is much greater than the likelihood of her legitimately winning four times does not prove anything." I agree it doesn't "prove" anything. However, the question isn't proof, but which does the evidence point to as being the more likely? Given the odds of winning the lottery four separate years in a row are astronomically small and require repeated beating of long odds is extremely strong evidence that she is somehow manipulating the system. Likewise, the fact that all life (which it has never been shown exactly how it might arise naturally) that we can imagine requires an environment that can only be created by a very rare combination of events each of which appears to be extremely remote strongly suggests that the game has been fixed and did not arise by pure chance.
Saint:

Thx for the clarification. It does not sound like you are against *any* probabilistic arguments ... just those which are levied against potential chemical pathways in abiogenesis theories.

Some design oriented arguments (not just biology, but cosmology etc.) appeal to probabilities. The gist of it is ... when you pile one lucky coincidence on top of another lucky coincidence on top of another one ad infinitum ... the plausibility of blind dumb luck begins to get thin. The inference is, that something other than simple blind luck is going on ... like maybe all these "coincidences" really are not coincidences.

That is the larger context of the discussion ... versus just abiogenesis and chemical pathways.

It is not a proof of design, per se ... but an inference that leads us in a direction ... i.e. evidence.

Weekend Fisher:

Good catch. His conclusion begs the question ... "it had to happen" is a presupposition, not an argument.

BK:

Nice summary of the argument. I wish I could better convey this to my friend. Perhaps he gets it and is unwilling to admit it. I am running out of analogies to simplify this argument. Perhaps the answer is the Columbo question you present ... specifically,

"the question isn't proof, but which does the evidence point to as being the more likely?"

I never actually used the word "proof" ... he inserted it ... which indicates to me, anyway, that he is not really getting my argument at all.

Perhaps pointing this out will do the trick. Thx.
Andrew said…
I am comparing the odds of two things ... the likelihood that pure random chance as an explanation works, and the likelihood that it does not.

You are exactly correct to do this. Your opponent is just mistaken in his method.

A trap Christian apologists often fall into is to say "This is really really unlikely, therefore it cannot have happened by chance." That's just wrong. What you are allowed to validly do is to compare the likelihood of the possible causes of the observed result, and infer the result as evidence for the most likely of the causes.

However, when it comes to the origins of life I would not be inclined to take seriously any modern probability estimates (we just don't know enough about how it might have worked), so I have to regard the whole discussion as fundamentally flawed here. (The statistical inference process above works nicely in an anthropic coincidence argument though.)
Anonymous said…
When something of seemingly impossible probablity occurs, people abandon the improbable in favor of the probable unless and until evidence gives them reason to think otherwise. The naturalist, however, as I've found, is inconsitant with his own conclusions—i.e., he always depends on non-naturalistic evidence to explain naturalism. And your friend in this situation is the same. He/she depends on something that has not been evidenced naturally (he/she has not experienced a situation where someone won the big lottery a great many times in a row in the span of a few years without cheating) and yet allows for this to be a possiblity on equal footing with what the evidence does show (Michael Larson, for example, in Press Your Luck). In the process of trying to argue for naturalism, your friend has negated it. You've won the debate already.
BK said…
Just a quick response to Andrew's final paragraph. I agree that we have to be careful about placing too much stock in the "modern probability estimates" (as Andrew put it) of life arising naturally on earth. I agree that the data we have on these probabilities is limited. However, I would note that some of the probabilities are not that difficult.

Assuming the heterogeneous nature of the universe, we can estimate the odds of several of the factors which we believe to be essential for life to arise on earth. For example, we know that life would be much harder if our planet circled a red dwarf or red giant. We also know that life would be much more improbable if we circled a binary star or a neutron star. We also know that life is much more unlikely if the planet orbits the star outside of the "life zone" where water can remain on the surface of the planet in liquid form. The odds that earth would end up circling the right kind of star in the life zone can clearly be estimated based upon probability.

There are other things that are unlikely, and I welcome anyone to research these issues more thoroughly. But for now, it is important to recognize only that simply because some of the probabilities are estimated, does not mean that the estimates in many cases are groundless or extremely uncertain.