I think this deserves a longer response, and one more informed by (for instance) your newsletter yesterday, but my gut reaction here is: a lot of social science is analyzing stable but temporary systems as if they were permanent orders of meaning and structure. But they're not! The whole idea of a permanent, integrated alliance like NATO is new, maybe even sui generis, and wouldn't have been feasible or even, really, imaginable--and it's not clear to me that it would even come back later in a similar form. (The pooling of sovereignty that NATO represents is somewhat haphazard, and I can envision a 21st-century version embracing something much more than an alliance, with truly shared military capabilities for instance, that would look like a post-Westphalian arrangement.)

The other part of this is that alliances and related phenomena, like tripwires, bear a lot of burdens. They have to signal to allies, they have to signal to publics, they have to signal to adversaries, and they have to signal to the international community. It's clear that the meaning of those signals is contextual: the risk of a nuclear conflict is much lower than it was in 1960, but higher than it was in 1997, and yet the device we employ remains the same.

So how can we theorize about moving targets? The best answer is probably to recognize that our answers are partial and contextual, not universal and timeless. That's one reason this project attracted me: there is no reason why Art 5 should apply to cyber. Really, there's none--the subsequent statements that any attack as damaging as a physical attack would trigger A5 do so against the spirit and text of the article. But on the other hand the NATO alliance was founded in an era in which a lot was agreed to without being written down, and it's also plainly apparent that a commitment without covering cyber would be a very weak instrument, so NATO *has to* cover cyber! But this is in turn a problem for theories of alliances that hold that explicit guarantees are what count. We get into this a lot more in the paper, but this is one reason I just think that our estimates of the effect size are a better guide than our colleagues'.