E3 is busy this year. It took literally an hour to park even at the very start of the day. I ran a workshop on in-game advertising on Tuesday (which I'll post about later), so I couldn't attend the Education Arcade this year. Last year I covered the conference extensively.

Nevertheless, I managed to catch one session, on using Neverwinter Nights for education. Click through for my notes (special thanks to Cyprien Lomas and Mark Wagner for letting me in on their SubEthaEdit notetaking session).

Case Studies: Neverwinter Nights
Philip Tan, Media Development Authority (moderator)
Russell Francis, Oxford University
Alice Leung, BBN Technologies
Peter Gorniak, MIT Media Lab

Philip Tan, Media Development Authority
Philip introduced Neverwinter Nights, a fantasy role-playing game. NWN comes with the Aurora Toolset, an editing tool for the game. Aurora is designed for non-programmers to put together an NWN level. For those who do know how to program, more complex interactions are possible.

Russell Francis, Oxford University
Russell talked about using MIT Comparative Media's Revolution NWN mod with homeschoolers. Revolution is an NWN mod created to teach students about 18th century colonial Williamsburg. Russell argued that the game allows students to learn about aspects of social history; it is an educational resource to be used among others. For example, in the workshops of each of the 7 characters, players learn about their tasks and goals.

Among the learning objectives of the game, Russell included geography, trade and employment, social class, gender relations, slavery, family obligation, and communication. In one hour of class play, Russell wanted to see how the students would model the complex social system of an 18th century colony.

Of special interest was how to develop a games-based pedagogy around Revolution and apply it to learning outcomes. Russell borrowed criteria from the New London Group's multiliteracies pedagogy (which, incidentally, Jim Gee was a part of). This included four criteria, roleplay leading to situated learning, drawing out tacit knowledge, production, and critical framing.

He was interested in how the game could telegraph social roles in the 18th century colony, especially the restrictions based on rank or social status. Russell found that the play experience underwrote a meaningful discussion of abstract concepts afterward. To demonstrate this, he had the students synthesize this learning in other ways. He by having students write a diary of a character in the game, but then moved to a machinima diary -- a great success. Russell also pointed out that the machinima diaries and their constituent artifacts could become platforms for further learning or discussion email to friends, or share the assets for creating machinima diaries or other artifacts.

Alice Leung, BBN Technologies
Alice shared her experience using NWN as a testbed for culture and personality research. NWN allowed her group to study behavior without starting from scratch to build a game environment. They wanted cooperative team play, non-violence, and required a reasonable and immersive environment for the representation of the contemporary world. They found that NWN was highly modifiable and allowed them to realize these goals.

Leung also took advantage of the NWN user community, finding free third party data-logging software that allowed them to record every action a person takes, their location, and their communication with one another. This was critical for later assessment.

Peter Gorniak, MIT Media Lab
Peter demonstrated his impressive modifications of NWN for use in situated language understanding in the cognitive machines group at MIT Media Lab. Previously, the group had been using robots to test machine understanding of language, and Peter showed us Ripley the Robot, who was capable of performing commands like "touch the one on my left" or "pick up the green one."

But robots present numerous other challenges, including just getting them built and functional. For this reason, Peter moved over to games, and specifically NWN, of which he was an avid fan.

Peter articulated three problems of situated speech and demonstrated his very cool hacks of NWN to create AI agents in the game to try to perform well in each situation.

Situated speech covers cases where we can't know what the speaker means unless we know the situation they are in.

Type 1: Ambiguity of form (what was said?)
Demo: a barely intelligible phrase, "attack the barrel" (which most of the audience heard as "attack the sparrow"). The agent interpreted the statement based on proximity to objects in the environment, and attacked the nearby barrel.

Type 2: Ambiguity of reference (what are they referring to?)Demo: "unlock this chest" (standing by a wooden chest). With two chests in the room, the agent disambiguated by the position of the speaking character.

Type 3: Ambituity of intention (what do they want)
demo: "can you help me with this?" (no nouns in this request). The agent had to know what to do: help opening the locked chest or, for example, the attacking demon? Peter built a plan to model the current goals of a level and its state, for example, if a lever opens a door, and the player stands by door and says 'can you help me with this', the computer agent throws lever to open door.