Thank you for your response, and I'm so glad you appreciated the critique!
Right, the polygraph aspect – I admit that I forgot about it after the explanation, because my focus shifted to just looking for suspicious statements. But I do remember feeling uneasy when the polygraph was mentioned, because they're so often misrepresented as true "lie detectors," and then being happy to see that you acknowledged the way they actually work. (Maybe this is a weird thing to fixate on, but I have very high anxiety and know I would probably set off a polygraph immediately, so that's on my mind.) Considering that makes me like the ! system more, since it makes narrative sense: SAM is a robot/program, so it's plausible for them to have that kind of mechanism built in, and they're an AI, so differentiating human emotional responses probably isn't intuitive for them but it's something they can learn. That even seems like something that could mirror the player's process of learning when not to press someone, once you add statements that set off "false alarms" on the polygraph.
I'd definitely be willing to go over all the writing stuff I noticed! I can do that in another comment here if you'd like, but I'm very long-winded, as you saw. I could also DM you on Twitter to make it less public, if you don't mind having big blocks of text coming at you there.