Cornell Video Game Speeds Language Learning

In an effort to streamline the process of learning a second language, Cornell University researchers first created a "fun" aspect by adopting a video game format, then boosted the speed of learning by incorporating a social media "chat" function.

The researchers’ most recent paper, Social Situational Language Learning through an Online 3D Game, was presented at the recent ACM Special Interest Group on Computer-Human Interaction conference (CHI 2016).

Mark Riedl, associate professor in the Georgia Institute of Technology (Georgia Tech) School of Interactive Computing and director of the university’s Entertainment Intelligence Lab, said the Cornell work "parallels my own studies of ‘games with a purpose.’ While we work with games that crowdsource common-sense knowledge data sets instead of education, we are trying to identify which game mechanics increase player enjoyment and also increase the accuracy of data produced by humans."

Riedl said he and his colleagues at Georgia Tech "have seen similar results, where social interaction improves both enjoyment and our purpose outcomes." The work at Cornell, he said, "further reinforces our understanding of the importance of social interaction in learning and play. Serious games, especially for learning, have been around for a long time, and it is good to see rigorous attempts to tease apart the game mechanics to figure out what design aspects correlate with improved outcomes."

The Cornell researchers claim that adding the fun and chat aspects into their immersive multiplayer video game—called Crystallize—enabled people to not just memorize the meanings of unfamiliar words, as is the case with traditional second-language learning methods, but also to more quickly learn the context in which to use those new words.

The key, according to Gabriel Culbertson, a doctoral student in information science on the Cornell team, was supplementing the traditional language-learning paradigm with a role-playing game that utilizes spaced repetition to enhance memorization, and also facilitates long-term engagement with multi-user functions.

"Language is inherently a social skill, so multiplayer was a natural fit for the game and added early in its development," said Culbertson. "There are aspects of language that can be learned better alone, and others that can only be learned with another person. For example, memorizing vocabulary is generally easier alone, but learning to have a real conversation can only be done with another person. We wanted to support both types of learning within the game, so some parts of the game are single-player and other parts are multiplayer."

To test the effectiveness of their approach, the researchers performed two tests. In the first, 42 participants were asked to learn using different levels of dependence on one another, resulting in the finding that more cooperation among users generally led to faster vocabulary building.

In a second, larger-scale test with 186 people, a quest paradigm for the game (namely, getting a job) enabled the gaming algorithm to generate a larger set of scaffolded tasks that provided a variety of contexts in which to learn the diverse usages of each new vocabulary word. This addition not only led to a more engaging game—with participants voluntarily spending long time periods per session—but also resulted in learning more new words per hour.

"Overall, players memorized around 10 new words per hour of play. However, language-learning gains are difficult to measure because using a language is such a complex task. People don’t just need to memorize words and grammar, but also need to understand when certain ways of speaking are appropriate. For example, in Japanese, a learner might need to address a friend differently than a teacher. In the future, we hope to measure these additional aspects of language learning," said Culbertson.

In the game, the player observes computer-generated avatars using the Japanese language in the context of social interactions. Using the chat function, users playing the same game can share their best guesses at the content of the words used, and also must cooperate to complete tasks. Each player also keeps their own vocabulary log, by dragging each newly learned word from a character’s speech balloon into their personal inventory of words for later use in constructing sentences.

In the latest version of the game, users are sent on quests aimed at getting a job, which earns them ‘money’ that can be used to buy items for use in future quests. As in most role-playing games, completing a quest moves the player to the next-higher level. The highest learning rate was achieved in a version which forced users to collaborate in order to get to the next level.

"Next, we plan to study whether long-term retention is strengthened by learning vocabulary in our 3D game context, over conventional methods," said Cornell assistant professor Erik Andersen, primary faculty advisor for the Crystallize project.

The team is also working on a immersive 3-D version of Crystallize that uses the Oculus Rift head-mounted display (HMD), which it plans to release this fall. After that, they aim to develop versions for learning both Spanish and English as a second language.

R. Colin Johnson is a Kyoto Prize Fellow who ​​has worked as a technology journalist ​for two decades.