The only caveat I have with this is that the behaviour of virtual patients (or virtual anything else!) depends on how they have been programmed to behave. Programs embody the programmers' prejudices and expectations. Consequently, the results may not be true to life at all. This may, paradoxically, be a bigger problem when the technology makes it look so lifelike.

It is no longer true that virtual characters have strict recipes (are merely programmed) limiting responses to easily discerned patterns in a short set of repetitions. It has been well over a decade since Random Number Generators were first employed to create a range of variables increasing complexity and perceivable reality. The "Uncanny Valley" is considerably more shallow (harder to discern) than that these days and can even be accomplished for very large crowds of individuals.

For reference see this article from Wired regarding LOTR. --- Thinking AI on a Massive Scale ---

We are ALL, all things programmed in a fairly literal sense of the word. More than that the definition of "alive" has always been tenuous and that is becoming evermore diffuse and difficult to nail down as knowledge and experience grows, and "consciousness"? fuggeddabotit since our view is commonly so inescapably anthropocentric. At some point the difference between "alive" and "not alive" becomes so moot it ceases to matter.