I have been recently entertaining myself with a 3-day non-stop binge of Theist vs. Atheist debates, On the atheist side: Richard Dawkins, Christopher Hitchens, Daniel Denett, Sam Harris, P.Z. Myers. On the theist corner: Dinesh D'Souza, William Lane Craig, Alistair McGrath, Tim Keller, and (unfortunately) Nassim Nicholas Taleb. One of the interesting points that comes up, often by Hitchens, is what I call the "Bodycount Argument". The atheist will claim: "Look at all the deaths caused by religion: Crusades, Inquisition, Islamic fundamentalism, Japanese militarism, Conquests of the New World" and the list goes on and on. Then the Theist will claim: "Well, look at the Nazis, the Fascists, the Soviets, the Khmer Rouge...". the Atheist then tries to reverse some of that, e.g. the Fascists were the catholic right wing, the SS were mostly confessing Catholics and Hitler had churches pray for him on his birthday, and, most tenuously, that the Soviets had the support of the orthodox church and used the pre-existing structures set up by the Czar to establish their power.
Some of that retort is convincing, some is not so much. You cannot really blame Soviet, Cambodian and Chinese massacres solely on religion. While they do at least manage to bring it to a tie, I suspect that the atheists follow this argument up suboptimally. My instinctive reaction would be "ok, so you proved that except for religion, communism leads to mass slaughter too. I have no problem doing away with both". But the Theists have a stronger form of their argument in which they claim that the crimes of communism are -because- of atheism, so a simple one-line retort won't work in all cases. We need to lay a deeper foundation for that claim to be convincing.
Enter single points of failure. The rudimentary definition, usually given in terms of computer networks, is that a single point of failure is that component which takes down the entire system when it fails. While the term has originated in computer science as far as I can tell, it can be applied to human networks as well. The strategy of Alexander the Great, at the battle of Issus, was instead of trying to defeat the entire, vastly ournumbering, Persian army in combat, to attack the Persian king Darius directly. When he was able to make him flee, the entire Persian army fell into disarray, with one side executing an orderly retreat, but the left flank completely disintegrated while being pursued by Alexander's cavalry. So while the term is new, the concept has been long known and has been used to great effect.
What I want to argue, is that all the examples cited by Theists and Atheists alike, are instances of a single point of -moral- failure. Here, instead of the system disintegrating or stopping to operate, it goes into a sequence of actions that when examined by an outside human observer, or even the participants themselves at a latter date, seem to be immoral, irrational, and akin to madness. The common point in all the examples is that a central organization, supported by a specific fanaticizing ideology, ordered the massacres to occur, and the people at the lower ranks, implemented those orders, despite perhaps individually knowing better.
My explanation of this, is that the lower-ranks had in effect outsourced their moral sense to their leadership. As with all centralised structures, when things go well, they go -really- well (assuming aligned incentives, greedy algorithms generally will not be as optimal as top-down ones), but when they go bad, they can be disastrous. The bigger the power of the network, the bigger the consequences. It is not hard to imagine why the outsourcing happened. Humans are tribal. I think very few, having observed the weekly rituals called 'football games' (whatever your definition of football is) would disagree. But humans are also moral. We have a rough set of rules that we tend to follow relatively consistently. What is of interest in these cases, is that an individual's tribalism completely overrode that individual's personal morality. And this happened repeatedly and reliably, throughout the ranks of each of these human networks.
Coming back to the original argument, if indeed tribalism trumps morality, and the above give us good reason to believe it does, then the theist argument that god put morality inside us comes into question. It does not explain why god saw fit to make our morality less powerful a motivator than our tribal instincts. But the biological explanation stands confirmed: If morality is a mechanism that was useful for intra-tribe interactions, then it would -have- to be suspended when the tribe was facing another. One can imagine the pacifist tribe being annihilated by the non-pacifist tribes around it or, lest I be accused of arguing for group selection, the individual pacifists being attacked both by their own tribe or the enemy tribe. Tribalists may disagree about who gets to live and who gets the resources, but they don't disagree about tribalism.