ALEX: bridging the gap between home and school language expectations. Alex is a virtual peer used as an intervention and experimentation platform to understand how students' learning behaviors are impacted by various cultural design choices in a tech-mediated environment. Alex was either monodialectal (spoke only Standard English) or bidialectal (additionally spoke African American English). When students' attitudes about a given dialect (rather than their own language practices) were aligned with the dialect of the agent, they demonstrated higher rapport, leading to stronger learning outcomes. This study was referenced in The Economist (2017) and The Huffington Post (2016).This project additionally served as the work for my PhD thesis.
SimStudent: talking with an agent predicts learning better than talking at an agent. We performed an experiment using the SimStudent learning by teaching platform. Stacy, the agent, makes the same math mistakes students had made previously. We evoke reflection from students as they articulate Stacy's mistakes back to her, which are really their own misconceptions. We found that students' pronoun use when referring to the agent ("you," "she," or "it") strongly predicted how likely they were to demonstrate improved post-test scores. The more students personified the agent in their speech, the better their own performance.
Snag'em: community networks for marginalized populations in computer science. We developed an app to support undergraduate computer science students in finding peers with similar interests and building social relationships as a mechanism to promote retention. We deployed this experience in three computer science university departments and three conferences, including CHI 2010. We uncovered insights associating participation, ideologies, and sense of community, and identified factors that impact uptake (such as those perceived to be the most well-connected not choosing to participate, which limited the morale of other potential participants).
Space CoDES: a game about meta-linguistics. Scholars have suggested that bidialectal children may benefit from explicit opportunities to translate ideas between dialects, but this is often a stigmatized practice that is rarely integrated into classroom learning. This was a collaboration with Carnegie Mellon's Entertainment Technology Center to design an experience that would allow children to reflect on their own language use in different situations. After playing, African American elementary aged children successfully described their own language practices at home and at school using the contexts in the game as analogous environments.
Monster games: meta-linguistic practice and assessment. We developed a series of four apps that used meta-linguistic awareness as a game mechanic. Elementary aged bidialectal African American children who speak non-standard English in the classroom were able to identify the dialect of a novel sentence, but had a more difficult time associating this dialect with a given context. Students identified monsters who spoke African American English as being friendlier than those who spoke Standard, but also less likely to be intelligent.
Vicarious peer modeling on rapport and co-construction. We explored how different collaborative learning behaviors demonstrated in a learning environment may impact the interaction behaviors among real children dyads using the system together. We found that systems which modeled a "worked example" of collaboration, rather than a "worked example" of the math task domain resulted in improved student post-test scores. We also found that students' rapport during the intervention was predictive of productive peer collaboration behaviors, such as co-constructing ideas.
Navigation, perception, and task performance in immersive virtual worlds. We explored how we could leverage perceptual illusions through phenomena like change blindness to improve the experience of navigating a virtual reality environment. We found that these illusions went largely undetected by users, allowing us to expand their perception of free space by up to ten times.
GameCATS: learners as creators of serious games for education and advocacy. We developed a ten week curriculum that taught students with below-grade level standardized performance scores how to make video games to help their classmates learn target math skills. A bit of a head fake, we found that students demonstrated drive to understand the math principles well enough to design their games, and demonstrated significantly higher math skills at post-test. Subsequent iterations of this curriculum changed focus to encourage students to design game experiences that advocated for a social justice issue they cared about.
Astrojumper: an immersive virtual reality exer-game for children with autism. Prior to the now readily-available Kinect and Wii gaming systems available for home use, we built a virtual reality experience designed with and for children with autism. We found that the game effectively evoked rigorous exercise behaviors during game play, and that there were positive correlations between reported engagement, perceived exertion, and biometric scores associated with exertion. This study was referenced in MIT Technology Review (2015).