In an prosaic social context, I am happy to use Dobbins' stronger definition: I definitely say that I disbelieve in the existence of god. In this context, I am using the social definition of "god": the sort of being that characters such as Yahweh, Jesus*, Allah, Krishna, the Buddha*, Ngai, etc. purportedly represent. None of these entities actually exist; I believe that these characters are fictional on the basis of evidence and reason. I might be mistaken, of course, but I definitely do believe, and I would rationally defend that belief, that they do not actually exist. In a prosaic context, I agree with Dobbins: the facts warrant a statement of definite disbelief.
*To the extent that explicitly deistic attributes are essential to these characters. In a similar sense, the character of Abraham Lincoln in Benjamin P. Thomas's Abraham Lincoln: A Biography represents a real person, whereas the character of Abraham Lincoln (Benjamin Walker) in the film, Abraham Lincoln: Vampire Hunter is fictional.
However, things get a lot more complicated when philosophers consider an idea. Many atheists, myself included, have studied a considerable amount of philosophy, and there are many philosophers who have examined and defended atheism at the highest professional academic level. In a philosophical context, the precise meaning of words becomes critically important; the unqualified word, "god," becomes unacceptably ambiguous. The sense noted above, beings like Yahweh, etc., i.e. beings with personality, desires, preferences, and who intervene in the physical world to effect their will, is only one sense. There is also the deistic god, a god who sets the world in motion with a set of physical laws and then does not intervene further. This sort of god is not so much disbelieved as dismissed. While it would be nice to know, even if such a god existed, it would have so little impact on my daily life that in the absence of any evidence (even if such evidence could be adduced) deciding one way or the other is a waste of time. Finally, there are the "gods" of Sophisticated Theology™. For example, Jerry Coyne (who reads Sophisticated Theology™ so I don't have to), quotes David Bentley Hart's book, The Experience of God: Being, Consciousness, Bliss:
To speak of “God” properly, then . . . is to speak of the one infinite source of all that is: eternal, omniscient, omnipotent, uncreated, uncaused, perfectly transcendent of all things and for that very reason absolutely immanent to all things. God so understood is not something poised over against the universe, in addition to it, nor is he the universe itself. He is not a “being,” at least not in the way that a tree, a shoemaker, or a god is a being; he is not one more object in the inventory of things that are, or any sort of discrete object at all.It seems clear that Hart's definition of god is not the sort of... concept?... that I can have any belief one way or another regarding existence. To be philosophically rigorous, the stronger, definite statement of disbelief is too narrow to encompass all these different definitions of "god"; the broader, and admittedly weaker, definition of "atheism" as a lack of positive belief succinctly covers all these cases.
In addition to ambiguities in the meaning of "god," there are also ambiguities and subtleties in the word "believe." In a philosophical sense, a person can believe or disbelieve only propositions, i.e. statements that can coherently be either true or false. (Philosopher Theodore Drange explores this concept in some depth in his 1998 article, "Atheism, Agnosticism, Noncognitivism.") If "God exists" is a proposition, then I can definitely disbelieve it. However, if "God exists" is not a proposition, as Hart seems to claim, then I can neither believe nor disbelieve it. In a similar sense, I can neither believe nor disbelieve the statement, "Colorless green ideas sleep furiously," nor can I believe or disbelieve emotitive sentences such as "Yay!" or "Boo hoo!" Again, confronted with a vast range of ways that theists present the propositional status of "God exists," I can be both precise and compact only by asserting that I lack a positive belief about the existence of God.
In addition to senses of god that are not propositions, there are senses that are propositional but cannot be known. To illustrate this principle, consider the statement, "There is [present tense] a ninja hiding in the room." First, this statement is hard to prove: ninjas are, by definition, far more skilled at hiding than I am at detecting them. More importantly, though, even if I discover a ninja in the room, he or she is ipso facto no longer hiding. Neither discovering nor failing to discover a ninja, therefore, is evidence for or against the proposition. While the statement is propositionally, semantically, and even scientifically unproblematic, it is fundamentally unknowable by definition. While I might be able to come to a definite belief on indirect evidence (it seems unlikely that any ninja would want to hide in my office), if I am going to be rigorous (or if I am considering a statement where indirect evidence is unavailable), I have to simply deny any belief.
Another philosophical subtlety comes from the way that scientific naturalists such as myself view knowledge. First, in the scientific naturalist account, without exception, all knowledge — i.e. all propositional statements about reality — is always provisional. All knowledge is conditioned on evidence, and any individual human as well as all human society, has at any time only a small, finite subset of the very large and possibly infinite body of available evidence, and all knowledge, therefore, is subject to revision given new evidence. Because all knowledge is provisional, it's unnecessary to explicitly condition knowledge statements with provisionality. The sentence, "I believe (or know) that two bodies experience an attraction described by general relativity, which can be closely approximated at low densities as a force proportional to the product of the masses divided by the square of the distance," does not gain any additional meaning by adding provisos noting that further evidence might change my opinion. Because there are no statements about reality that are believed non-provisionally, we don't need to distinguish between provisional and non-provisional beliefs, and the linguistic distinction is dropped as redundant. In my own writing, I try to avoid the word, "certainly," replacing it with "definitely," but my vocabulary was shaped by convention, not scientific rigor, so I occasionally err. To the obtuse or unaware, unconditioned statements about knowledge sometimes appear to be stating facts with certainty rather than definiteness. Thus, even towards conceptions of gods that I disbelieve, I definitely disbelieve, i.e. I have made a decision, but I do not certainly disbelieve.
A more important consideration, however, requires looking a little more deeply into how scientific naturalism works. Because all knowledge is provisional, it is always statistical, at least conceptually. (I have to egregiously simplify here, but I hope to capture an essential feature about scientific knowledge.) In a statistical model, we create a "null hypothesis," which represents a default belief about the world, and an "alternate hypothesis" which represents the negation of the null hypothesis. For example, I might say that the null hypothesis is that the average height of men in the United States equal to 179 cm, and the alternate hypothesis is that the average height is not equal to 179 cm; on average they are either shorter than or taller than 179 cm. Note that the null hypothesis is probably not precisely correct; even if the average height is very near 179 cm, it is probably not exactly 179.000000000 .. 000 cm (we can measure length very precisely). (This imprecision is not really problematic; close enough is close enough, and if I'm designing a car or a house, for instance, I don't need to know the average height to nanometer precision.) In addition to being not precisely correct, the null hypothesis is usually not directly provable, it is only disprovable. If I measure the height of 300* men, and find that their average height (sample mean) is 180 cm, with a standard deviation of 10 cm, then I know with about 95.8% confidence that the average height of all men is not 179 cm. Note that I do not know that the average height of all men is 180 cm; I have "proven" (provisionally) only the alternate hypothesis, which is that the average height is not 179 cm. The best I can say is that I have good evidence for now considering 180 cm to be the new null hypothesis when talking about the height of American men.
At this sample size, the different between the normal and t distributions is negligible.
This method impels a curious terminology that any competent professor of statistics will impress on her students: you say you reject the null hypothesis or you fail to reject the null hypothesis; you do not, on pain of durance vile, ever say you accept the null hypothesis. Similarly, you say you have sufficient evidence to conclude that the alternate hypothesis is true, or you have insufficient evidence to conclude the alternate hypothesis is true; you never say you conclude that the null hypothesis is true. Strictly (very strictly) speaking, therefore, a scientific naturalist never actually believes the currently specified description of the world, a systematic collection of null hypotheses; she believes, instead, that she has insufficient evidence to conclude that the world is different from this current specification.
Note that "insufficient evidence" applies equally to edge cases as well as to non-edge cases. In the above example, if I had measured the height of only 250 men, I would be only 94.3% confident that I can reject the null and conclude the alternate hypothesis (that the average height was not 179 cm) was true. Because by convention I will reject only if I am 95% confident, I will fail to reject the null and conclude that I have insufficient evidence to conclude that the average height is not 179 cm. Similarly, if I find the average height of my sample to be 179.1 cm, I will be only 56.2% confident that the null is false, but I will still just say that I have insufficient evidence. (If I measured 30,000 men, however, a sample average of 179.1 would give me 95.8% confidence to reject the null.)
In practice when we repeatedly test and fail to reject some specification of the world, especially when our failure to reject is not borderline, we have good reason to believe the world really is at least very close to the specification. Still, when pressed, and in ambiguous or uncertain circumstances, scientific naturalists tend to retreat to "insufficient evidence" semantics.
I hope you'll forgive me, gentle reader, when I tell you we atheists really don't care that much anymore about the philosophical subtleties I have wasted so much of your time describing to you. As far as most atheists are concerned, the philosophical and scientific debate is over, decided. No matter what definition of "god" you choose (that is not intentionally metaphorical nor does unacceptable violence to the meaning of the word, "god"), your definition is meaningless, non-propositional, unknowable, or rejected by the evidence. We make a nod to the philosophical subtleties by making the most general statement — we lack a positive belief about god, which includes disbelief in some definitions of "god" — when concision is more important than detail.
Atheism is not primarily a philosophical position; it is a political position. Our position is that all god talk (that is not intentionally metaphorical) is not just nonsense, but pernicious nonsense. Religion is not just a weird thing that some people do in private; it has profoundly negative effects on our societies, cultures, and nations (and what positive effects it might have would be at least as good, and usually better, if the god talk were eliminated). We define atheism broadly not just as a nod to the philosophical subtleties, but also to be as inclusive as possible to people who reject god talk for a variety of reasons, with various degrees of philosophical sophistication. We want to include as an "atheist" someone who is not particularly interested in philosophy, who just doesn't know whether or not Yahweh and Jesus are real, but who finds offensive and absurd, as we do, the notion that, for example, the leader of an organization of supposedly celibate men, an organization that has gone out of its way to protect and defend men they know have raped children, has anything whatsoever legitimate to say about how consenting adults employ their genitals or women employ, or refuse to employ, their uteruses. If you can say only that you lack a positive belief about god, and that people who say they do have any sort of positive belief thereby gain no moral or scientific authority whatsoever, you're one of us.