There's a lot of reasons to disagree with that, but one of the biggest is how impractical it'd be to do that. Let's go with a conservative estimate of 1000 people taking the AT-SAT each round. The total scored time (e.g. the actual tested portions) of Scenarios take up a total of 70 minutes. The amount of time necessary for review, then, is nearly 1,200 hours worth of people staring at a screen. And that's off of a conservative estimate; since pretty much everyone who fills out the application correctly gets to take the AT-SAT, the actual tests given could be numbering much higher.
Then there's the subjectivity, as mentioned above. It's not quantifiable in any way; it'd be based entirely on, "Well, that guy looks nervous." No statistician in their right mind would create a test designed to measure aptitude based on something so vague and ill-defined. I'm not even sure how you'd come up with agreed-upon guidelines to interpret that kind of thing... or how you'd train people to do it.
The simplest explanation is the most obvious one... cameras are there for the same reason they're in most places: to watch for impropriety. Cheating on the AT-SAT, that kind of thing. That way if a Robinson proctor says he caught someone with a calculator (or whatever) and the test taker disputes it, there's video proof.
