Submitted, and as would be customary for someline like myself, with comments. Here's what I had typed:
QUOTE(Nottheking)
The key failing: Context sensitivity. NPCs, superficially, act in lifelike manners. However, they methods of reacting to their environments are limited to direct interaction with objects they are programmed to interact with; they sit on furniture, sleep in beds, eat food, and talk with other NPCs or the player.
This seems realistic, until more exotic situations are presented; these behaviors are not changed whatsoever by numerous factors that would impact the behavior of real people, such as weather, crime, etc.
As an example, when nearby an open Oblivion gate, one would reasonably expect nearby people to be responding to it, possibly with emotions such as panic, fear, or perhaps even curiosity. The game's NPCs have no different reactions.
Likewise, NPCs are not reactive to the events that take place around them; major game-related events, such as the completion of a quest, may alter what dialogue they have availible, but their behavior is unnaffected. As an example, if the player enters a busy city street, and one NPC suddenly attacks the player, the NPCs may respond to the attack by aiding the player, but after the battle has concluded, the NPC's behavior, and even dialogue, is not affected.
This would appear to be a keystone in the elements of Artificial Intelligence that is lacking in the game of Oblivion, that would've added the realism level sufficient to truly make the NPCs seem as lifelike characters, rather than as flat "simulation bots."
That aside, I liked the subject that's implied for the thesis. I wonder if, after it's ben completed, and their PhD has been granted, that they might post it online; it'd be interesting to see what they've found out in their research. And after all, all PhDs are glad to get a paper of theirs published, though most institutions don't count "the Internet" as quite a good place to get published as an actual scientific journal.
My only other comment would involve the labelling of the typical ordinal questions; typically, they are labelled from "strongly dissagree" to "strongly agree." That was a bit of an odd naming of the options they gave, so I just treated it as such one usually would've been labelled.