it's a useful heuristic of mine to massively distrust anyone who makes silly claims like assigning probabilities at 1 decimal point to things like this:
> 1 Opposition: Opposition forces in Syria (Liwa al-Islam) carried out the chemical attack. 96.4%
> 2 Syrian army: The Syrian army carried out the chemical attack. 3.6%
though it helps when they also explain the process with nonsense like:
> The few prior instances of chemical attacks against civilian targets contain factors that make them poor comparisons to the attack in Syria, leaving motivation as the primary factor. An analysis of the motivating factors behind such an attack results in the Syrian army being twice as likely to launch such an attack compared to an Opposition group, i.e. 67%-33%.
I really hate this style of analysis. It's transparent, yes: it's a prior being updated by evidence to produce a posterior, and you see where the rabbit goes in the hat. But this just puts a sciencism gloss on what are basically qualitative arguments.
A short version of their argument is "You might think for any number of reasons that the Syrian government launched the missile, but other peoples calculations of the launch location were wrong and ours were right so it couldn't be the government." Their prior begins at 95-5 for the government doing it, then it becomes 69-31 for the government doing it when they show a video of an anti-government guerrilla wearing a chemical hazard suit, then it becomes 7-93 in favour of the opposition once they decide they know the launch site. The launch site evidence is supposed to mean the Syrian government's probability is divided by 3 and the opposition's probability is multiplied by 15. That's the whole argument, that's where the rabbit is in the hat. After that point, no other evidence can adjust anything.
Most of the other evidence is just conjecture: they assign a 90% confidence to the evidence point "No western countries shared their evidence that the Syrian government did it" and they feel this fact is enough evidence to half their confidence in the Syrian government doing it. As proof for this, they literally say "The US says they have signals intercepts saying the Syrian government did it, but what if they actually have signals intercepts from the Syrian government asking if they did it or maybe they did admit to doing it but they were lying" This is not evidence.
This really seems like crank people trying to apply a rational choice / scientific skeptic veneer to a process of qualitative reasoning. And frankly I flicked through another ten or fifteen pages on the blog and almost everything discussed was discussed in the exact manner LessWrong/Slate Star Codex commenters would, and many of the issues being discussed were the exact issues discussed on those sites.
(I have no prior at all on the Syrian gas attacks, as I don't pretend to know anything about Syria or chemical weapons)
Unfortunately, it's also standard for planted false stories.
It's notable that Hersh won the Nobel prize for producing photographic evidence of war crimes.
None of his anonymously sourced articles has ever been confirmed true