Serious Games Summit DC 2005, Day 2Derek Wischusen: Expanding Teachable Moments in Serious Games using AI (Jerz’s Literacy Weblog)

Began with a useful grid that recognizes that realism is not the only way to evaluate a game, and that realistic games are not the only kind of training games out there.

Wischusen presented two binary pairs:

Fun/Training

and

Realistic/Unrealistic

Goal: Fun -> Training

(I like the recognition that realism is one option, but there are many ways to be “unrealistic,” just as there are many ways to be nonwhite or un-American. Expressionism, impressionism, romanticism, abastraction etc. are all artistic alternatives to realism.)

Key messages:

Training isn’t “just in time” or “just in case”

Simulation isn’t training

Networks, virtual environments, games, process measures are not training

Training is “relevant practice with feedback”

Assessment and Feedback

Simulation/game provides the practice environment. Practice alone does not make perfect. Intelligent assessment and feedback is essential for effective training to occur.

Perfect practice makes perfect. To learn how to do it correctly, you need assessment and feedback to tell you when you’re not doing it correctly.

While the good instructor is always the best source of training, there are rarely enough instructors to meet training needs; and never enough to provide one-on-one training.

(Hmm… A embedded assumption is that one-on-one training is best.. but “best” is a value statement. If you factor in ROI, then one-on-one instruction may be inefficient. Are all instructors “good” at one-on-one training? The reference to a “good” instructor is sort of begging the question. )

The system can do pre-testing to identify the knowledge you bring into the system, personalizing the delivery according to the needs of the learner.

“Cognitive fidelity is as important as graphical fidelity for both training quality and end-user acceptance.” (Hmm… that phrasing privileges graphical fidelity, but the speaker noted that not all training simulations requires that kind of fidelity.)

Presentation of Virtual Interactive Pattern Environment and Radiocomms Simulator (VIPERS) – an Air Force Research Laboratory project designed to give student pilots practice performing a labor-intensive landing procedure. Typically simulated with a duct tape pattern on the floor, with the students calling out their status and responding to commands as the routine requires. A technique better practiced on the ground, but students have limited access to fancy simulators.

(Hooray – in an aside the presenter noted that 2D environments can provide just as good results as rich 3D simulations, and cautioned the assumption that realistic graphics are always the best option.)

Synthetic Teammates for Realtime Anywhere Training and Assessment (STRATA)

DARWARS – offers a massively multiplayer online war, that permits different training systems to plug into it. Emphasizes “headwork” (teamwork, not flying skills).

Giving the trainee a situation that contains an error; will the trainee note that the setup information contains an error that makes accomplishing the goal impossible.

AI agents should occasionally make errors, since people do make errors in the field. That’s part of the “realism” in the training environment.

Virtual Environment Cultural Training for Operational Readiness (VECTOR)

Building a 3D environment that teaches soldiers to interact with people in a foreign environment, so that the locals are more interested in helping the soldiers than hurting them. The AI agents respond to trainee actions based on the emotional state of the agent, according to rules determined by cultural subject-matter experts. Within the exercise is a synthetic instructor, who gives objectives, provides feedback on both military and cultural protocols. The training involves interacting with the locals, getting important information while following cultural rules.

VIGILANCE – HazMat training, skilled support personnel (SSP) who assist the first responders. They may operate heavy machinery. The concern is that our SSP haven’t been trained to assist an EMT during a HazMat response.