As someone who's studied AI, I agree with Winde about how emotions would only ever be an illusion, at best (the Turing Test is literally about measuring how convincing the mere illusion of AI is, after all), but would say SM has it right about Ash's behaviour: Like a Skynet Terminator, it would effectively be programmed to exhibit certain traits, like how it would calculate the most relevant emotional state to be. Ash would be undertaking a lot of processing 'overhead', precisely because of not being able to know whether or not a member of the crew might be observing. Ash is effectively maintaining an illusion all of the time and devoting a lot of processing power to that end,
Bishop, however, makes it clear that the Asimovian rules were never codified during the time of Ash's model.
What's interesting, now that I think about it, is that Ash's faked human crew profile is scrolling up behind Ripley when she's stating her case, yet Burke's later casually acknowledging Ash was known to be a synthetic, right down to specifying the actual model!