I share Glen Whitman's assessment that there simply is no need for a concept of "agnotheism." The coinage has been around for a while, and I've seen it used to refer to a number of different concepts, but I guess Jane Galt gets credit for inventing or popularizing its current meaning as "an agnostic who puts a very, very low--yet non-zero!--value on P(God)."
Of course, my first problem with the term is that it would appear, on its face, to break down to something like "the theory that one can't have knowledge of whether believers-in-God exist." But Glen lays out the more cogent argument -- essentially that Galt's definition amounts to yet one more distinction in search of a difference:
As recently as a few years ago, I referred to myself as an agnostic for just the reason that Jane describes. I reserved “atheist” for absolute certainty of the non-existence of god, and “agnostic” for a degree of certainty less than 100%. But after many conversations with my old roomie Julian, I realized that was silly, because I’m not 100% certain about anything. I’m not 100% certain that unicorns don’t exist, yet I don’t call myself a “unicorn agnostic.” I’m not 100% certain there’s no such thing as ESP, yet I don’t call myself an “ESP agnostic.” When I say I don’t believe in unicorns or ESP, it means that I just don’t have any good reason to think they’re real, and so I proceed on the assumption that they aren’t. If P(unicorns) and P(ESP) are sufficiently small, they don’t merit having representation in your label. Why treat P(God) any differently?
I'd go a step further than that even. The idea that one can simultaneously be "agnostic," and assign a probabilistic estimate to whether God exists, strikes me as incoherent. Presumably, agnostics agree with atheists (and even most theists) that there is no physical evidence of God's existence, so that would seem to rule out assigning a frequency probability. But agnostics also believe (at least, as I've always understood the term) that it is impossible to know whether God exists. So what room does leave for Bayesian inference or filtering?
Sure, anyone can just randomly assign numbers to serve as ordinal placekeepers, but since these numbers don't actually refer to any extant information about the real world, why bother? Would you actually use them to compare the likelihood that God exists to the likelihood that, say, unicorns do? And what could serve as the basis of any such comparison? Unicorns likely would strike most as the more fanciful hypothetical, but hey, at least there we might be able to point to the recombinant potential of horse DNA. Where's the starting point to begin ascertaining the "probability of probabilities" when it comes to God?
Moreover, I would contend the thought that there are actually some statistically significant number of adherents to what is sometimes described as "strong" atheism -- i.e. the stance that P(God) = 0 -- is fallacious to start. I can't say I've never met ANYONE who held that view, but at least in my subjective experience, it's been a tiny minority of those who self-identify as atheists. Most, instead, hold to some variant of "weak" atheism, or what Michael Shermer has called "nontheism."
But I think Shermer's coinage to be just as unnecessary as Galt's. A-theism is "without a belief in God," much as a-pathy is "without pathos." Nothing in the construction of the word suggests a positive rejection of or active hostility toward belief in God. To the extent there are any who hold such a stance, it would properly be labeled anti-theism, much like anti-pathy.
Of course, these semantic games bear an uncomfortably close resemblance to the inane libertarian/ liberal/ classical liberal/ market liberal/ anarchist/ minarchist/ anarcho-capitalist debates that have dragged on for decades in libertoid circles. Perhaps some of this can be attributed to both fringe anti-state and fringe anti-church communities drawing similar types of adherents (or, in many cases, the exact same adherents.)
That being the case, if Megan's "agnotheism" could serve to unite all non-believers under one banner, then I'd be perfectly willing to dispense with any quibbles about the awkwardness of the term and just run with it. But since that's not what she's proposing (and it wouldn't happen, even if she were) I think "atheist" already covers the people she's describing pretty well.