Monique, the eager-to-please girl with the chirpy alto, is raising her hand again. But I’m more interested in drawing Maria -- who hides in the back row and avoids eye contact -- out of her shell.
“She don’t wanna talk to you, man,” says Marcus, confidently flip as usual. “She don’t talk to anybody.”
Vince, the pallid kid with dark hair who sits at Marcus’s left, chuckles -- just like he did earlier when Marcus told me he “found” the Mercedes-Benz hood ornament, now draped around his neck, “in the parking lot.”
So I try engaging Francis, the shy but willing young man in camouflage shorts and a T-shirt. I ask him what he wants to learn about. “Uh … music,” says Francis, before launching into a beat-boxing exhibition that he says he learned from YouTube. I compliment him on the routine. Noting this, Monique raises her hand with redoubled urgency.
This is my class.
Well, sort of. I’m not really a middle-school teacher. But then again, the kids are not really middle-school students. They’re not even humans.
They are avatars. Not the blue kind from the James Cameron film,  or even the sort of avatar most often used in higher education: the fantastical, flighted characters that professors and student embody when learning in Second Life. To the contrary, the point of these avatars, created by a team at the University of Central Florida, is to be as realistic as possible.
They have to be if they are going to revolutionize teacher training, says Lisa Dieker, coordinator of the special education program at Central Florida’s college of education. That, after all, is the goal of the TeachME project: to effectively eliminate the trial-by-fire approach to classroom-management training, and replace it with something more instructive and less dangerous.
Dieker and the TeachME team -- which includes members of the university’s education, engineering, computer science, mathematics, and theater departments -- believe they have created a virtual classroom so real-seeming that it could drastically improve how prepared novice teachers are by the time they venture into the blackboard jungle as student teachers -- and in so doing, reduce teacher turnover by weeding out likely candidates for burnout.
Perhaps more importantly, it could limit the students’ exposure to underprepared, ineffective teachers. And, the team assumes, improve learning outcomes.
Without a Script
Here’s how it works: The teacher-in-training stands in a room in front of a projection screen depicting five students in two rows. The student avatars are being controlled by “interactors” -- acting students from the university’s fine arts school and sometimes hired professionals -- who have studied the behavior of the students they are embodying.
The fact that the teachers-in-training are interacting with avatars that are being controlled in real time by humans, as opposed to artificially intelligent personas, is the key to the whole project, says Dieker.
“The first scenario they built, they said ‘Lisa, come in here, stand on this spot, and say this,’ ” she recalls. “And I said, ‘Well, that’s not teaching -- I should be able to walk in and say whatever I want.’ And they said ‘You’re crazy! We don’t have simulators that do that!’ ”
The presence of human interactors eliminates the parameters that would make an artificially intelligent simulation a poor training tool for actual classroom teaching, Dieker says. “Teaching,” she observes, “is not a scripted activity.”
The interactors doing the live sessions are across campus in the university’s Media Convergence Lab, where they can see and hear the teacher-in-training via Skype. A second puppeteer, Angel Lopez, a teacher educator at the graduate school, controls a series of knobs that can prompt non-speaking outbursts, such as giggling -- or, in Francis’s case, beat-boxing.
Lopez can use the dials to “crank up the chaos” if a teacher-in-training starts to rub the students the wrong way. By way of explanation, Dieker calls up the virtual classroom and begins to do everything wrong: She antagonizes the students, and tries punishing minor disruptions by making them write “I will respect the substitute teacher” on a piece of paper 50 times.
“I can only do this for so long,” says Dieker, interrupting herself. “These are real kids! It’s hard to be mean to them.”
Harder for some than others. That is where the TeachME system can detect potentially unfit teachers early, hence reducing teacher turnover, through what the team calls the “after-action review.” If a teacher is consistently rattled by the chaos that can inhere in a middle-school classroom, that person might be advised to pursue another career. Likewise, Dieker says, if a teacher provokes chaos and exhibits no empathy or remorse, that person might not have what it takes either. And for student teachers who have what it takes but also have plenty to learn about teaching in practice, TeachME affords them the perfect environment in which to make mistakes.
Keeping it Real
In order to recreate a realistic classroom dynamic, each of the students is meant to fit one of the four types cognitive psychologists use to classify personalities.
Monique, of the perpetually raised hand, is aggressive-dependent, always seeking affirmation from her teacher; Maria, the unresponsive one, is “passive-independent,” shying away from contact with her teachers and peers; Marcus, of the hood-ornament neck chain, is “aggressive-independent, constantly asserting an authority in contradistinction to the teacher’s”; Francis, the reticent but responsive beat-boxer, is a “passive-dependent.” Vince, the one who really gets a kick out of Marcus’s subversive behavior, is also a “passive-dependent,” but he looks for affirmation from Marcus, not the teacher.
“You’d see those personality types in every middle school in the country,” Dieker says. “We still have these specific types as we move into adulthood, but we tend to blend in other types that allow us to keep our jobs and not get fired and all those things. But in middle school, when kids go through those horrible years of adolescence, they tend to be pretty true to a type.”
Needless to say, adhering to these types in an extemporaneous manner requires some serious skill on the part of the interactor. Kate Ingraham, an doctoral student in the university's instructional technologies program, is the one pulling the strings during my session. She has honed her portrayal of the characters over two years. (Update: An earlier version of this article misidentified Kate Ingraham as a UCF faculty member with a similar name.)
Seeding Teacher Colleges
The rub is that TeachME cannot fit into a neat little package — not yet, anyway.
In order to use the program, there needs to be an interactor in the lab and on the clock. The TeachME team has slimmed down the operation: they used to use five interactors per session (one playing each student); now, in an effort to make the technology cheaper and easier to deploy, a single interactor juggles all five personalities. The students’ body language is pre-recorded by interactors in motion-capture suits; they go on autopilot when not speaking.
Charles Hughes, the director of the Media Convergence Lab at Central Florida, says he is working on ways to reduce the load of the interactor. For example, the team is working to automate more elements of the system, so that interactors can work from home instead of in the lab (although the team is cautious of automating too many elements, so as not to compromise the verisimilitude of the virtual classroom).
Dieker says the goal is to make the system as cheap to implement as possible. From a hardware perspective, the system is theoretically inexpensive. Hughes estimates the hardware would cost about $5,000 to $7,000, and notes that the typical graduate school would probably own many of the necessary components already. The interactors currently bill at about $120 per hour, Hughes says.
There is also the matter of interactor training. Telecommuter or no, an interactor needs to be extensively trained to take on the roles of each student and be able to transition fluidly between those personas. Ingram says she is currently designing a program to train more interactors as more colleges of education implement TeachME pilot programs.
So far, Central Florida is working on partnerships with several colleges to deploy and improve TeachME. Utah State University is running a TeachME pilot that focuses on disabled, illiterate, and autistic students (Ingram has trained herself to play an autistic character, named “Austin”); another partner institution is working on a truncated program aimed at Teach For America, which has been criticized  (and even satirized ) for placing burnout-prone teachers in underprivileged schools; another partner is focusing on students in rural schools. There is no need to build new labs on location; the teachers-in-training at each far-flung college can preside over their simulated classrooms via Skype (which is what I did during my demo from Inside Higher Ed’s Washington office).
The current goal is 10 university partners, which Hughes says he thinks could grow more ambitious once the TeachME team implements more advances this fall that he says “could change the economic model substantially.”
The biggest question at this point is whether the TeachME teacher training system actually improves student learning. “We can prove it changes teacher practice,” says Dieker, but longer-term studies will be necessary to see whether the students in classes taught by TeachME-trained teachers actually learn better, and whether that success gap can be traced to the TeachME system.