While numerous attempts have been made at making video games for language learning, none currently available have been able to achieve both of two major challenges:
(1) Create a game that can hold youth attention and provide meaningful pedagogical experiences.
(2) Create a game environment in which content can be added in an affordable or practical manner.
To illustrate, we can look at 4 representatives samples of products, each achieving one of these two objectives:
With $16 million in initial funding and multiple subsequent grants, this Department of Defense sponsored initiative demonstrates what is possible by bringing video games technologies to the language learning field. Unfortunately, the high development costs of generating virtual 3D spaces and voice recognition models for each interaction are extremely high. Even with military budgets, only a portion of Tactical Iraqi features fully immersive environments and missions. Additional content can also be found in arcade-style games, however, the much of the introduction of new language elements reverts back to traditional textbook-on-a-screen model.
The cost of developing a top-end mainstream video game today has hit $60 million. Furthermore, a quick market analysis [see slide #11] reveals that players have extremely little interest in games that try to be top-end, but are not #1. This phenomenon has led to what is dubbed the "MMO mass graveyard." By contrast, a vast number of low-budget Flash and AJAX games (WoW figures on that link are out of date, but the Travian numbers are accurate) have proven to be the most popular games on the web, receiving billions of page views, with extremely profitable business models.
With the cost of devising mechanisms for naturally combining education+entertainment always exponentially higher than entertainment alone, it is clear that more innovative approaches [link Raph's club penguin post] will need to be explored, rather than avoid trying to reproduce high-budget games.
Virtual Spain is a commercial effort to re-create a virtual Spanish town within the Torque game environment. While the development budget of Virtual Spain is presumably considerably less than that of Tactical Iraqi's, the content is still extremely limited. Players are constricted to exploring a relatively small town only designed for the Spanish language. Players can have conversations with simulated inhabitants of the town, but aren't given specific missions or game-like objectives. Adding additional towns or opportunities for interaction would require creation of entirely new artwork and expensive 3D models to be created for the specific style and file formats used in the game, along with necessary development time for expanding the speech recognition components. This detracts from being able to focus on creating the core game experiences for students (which the current game severly lacks) and thinking about how to best incorporate language learning in a natural manner.
At the other end of the spectrum are activities like Hot Potatoes. Unlike virtual video game environments, the activities featured in Hot Potatoes (Matching, Concentration, Hangman, Crosswords, etc) are highly flexible and can be expanded to cover new topics with ease. For example, if a teacher using Hot Potatoes wants to do a unit on food, they simply type in the words they want to drill and it will generate a series of activities for students.
Unfortunately, the same mechanism that makes Hot Potatoes easily customizable also makes them pedagogically useless. Virtual worlds become expensive because each language element requires an immersive context to be created to represent the meaning of that word and opportunities for players to interact with the system to negotiate meaning. Games like hangman bypass this requirement by treating words only as an arrangement of letters, divorced from any representation of meanings. This, however, contradicts all modern theories of language pedagogy. Additionally, numerous students fail to find the activities particularly engaging.
The popular Quia.com provides activities similar to those of Hot Potatoes, as well as a suite of class management tools for the teacher. Their management system, however, was custom written from scratch specifically for the features currently available. This means any new features will either also need to be written from scratch, or complex interfaces designed to communicate between multiple systems.
The Animal Rescue framework is designed to be able to share a database structure with the immensely popular and freely available Drupal content management system and Facebook Applications. This means that, besides being easier to create robust teacher management tools, it would also be simple to create student-centric tools. For example, Drupal contains all the code necessary to rapidly build a professional social networking site, language exchange system, customized avatar profile, blog, etc.
The Animal Rescue Framework
The original goal of the Animal Rescue framework was to find a way in which the complexity of creative a quality immersive environment for language learning could be reduced to the point that it could be accomplished as a hobby by a single programmer. The first step towards this was recognizing that the majority of the development costs for many games lies not in programming cost, but in the artwork. This is because in order to create the engrossing fantasy/sci-fi/etc realm most games seek to offer requires that all elements of the world be custom drawn by an artist. Games for foreign language learning seek the opposite, they seek to connect students to real (albeit foreign) places. Thus photographs, rather than custom artwork, form a more natural fit to the goals of a foreign language learning game. This is convenient because the popular photo sharing site flickr.com has a commons whereby millions of their photographs from around the world can be easily accessed and legally incorporated into commercial products (requiring only attribution of the original photographer). For the game world itself, Google, Yahoo and Microsoft all provide freely available satellite and aerial photography. In future they will provide surface level photography which will allow games to be build to the ultimate level of realism. Combined with Google/Yahoo/Altavista image search and other freely available web technologies, this allows the primary development cost, artwork, to be eliminated for language learning games in the Animal Rescue framework.
On the programming side, the AR framework is built upon the Adobe ActionScript 3 platform. Released earlier this year, this platform allows game development many magnitudes faster than previous options. Additionally, it makes it easy for components to be reused across different games.
Natural Input
Most games use menus where players click on preformed sentences. This design specifically minimizes any thought about language construction and is highly unsuitable for foreign language learning.
Activities on sites like Quia.com use pull-down boxes. Yet these two were designed for passing http variables, not language learning games. They become highly clunky when used extensively for games.
Can I bring you anything to eat or drink?
I would like a glass of
please.
The AR framework offers a gestural input system that both offers an infinite range of possible sentences and encourages cognitive processing of language construction. Move the mouse cursor over "konte", "ich" or "nein" below:
Additionally, games designed for open-ended input mechanisms such as this one naturally lend themselves towards being upgraded to speech recognition, once the technology becomes more practical and reliable.
Intelligent Glossing
One popular practice in Foreign Language media is to provide annotated glosses of potentially unfamiliar words. In the theory community, countless papers have been published demonstrating that image-based glosses lead to more effective retention than direct L1/L2 translations. In fact, the effect is so reliable that the core undergraduate Research Methods class at UCLA has students reproduce this experiment as their first assignment. Yet, these findings have been largely ignored by publishers. Given the enormous costs of providing custom artwork for each word to be glossed and the lack of guidance for practical implementation from theorists, this is understandable.
The Animal Rescue framework features an intelligent glossing system that offers a wide range of options for providing image content. In fact, the easiest option is cheaper than doing L1/L2 glosses. By setting a single variable, Animal Rescue can be configured to retrieve image glosses directly from google images. As google images pulls images directly from websites in foreign language, no further effort is required on the part of the publisher besides setting that single variable. For example, if the user mouse-overed the word "hund" in a German game, AR would send a query with the parameter "hund" to http://images.google.com (with adult-content filter enabled) and automatically return a picture of a dog from a german website about dogs. (multiple images can be returned to reduce the likelihood of irrelevant images).
While this works most of the time, some images will not have the direct relationship to the word as the user intended, requiring them to search through multiple result for the right one (which may be a good thing). Alternatively, with minimal effort, programmers and direct the Animal Rescue framework to pull images from a cheap ($60 total cost), royalty-free, stock photography application such as Hemera PhotoObjects (this is the path used in the demo). This ensures professional, clean and reliable images. Textual glosses can also be easily inserted for content not appropriate for images.
Additionally, the AR glossing system enables developers to build XML-based repositories of instructions for repeatedly handling common lexical chunks:
In this case, the AR Framework knows to look for instructions about what to do at http://www.langwidge.com/ardata/gloss/xml/es_tut_mir_leid.xml (meaning it can easily be created in a number of user-friendly applications, then reused across multiple applications). There it receives the following data:
It then presents the user with an appropriately formatted gloss.
Student Authored Content
One of the most important shifts in [all] classroom pedagogy today is the recognition that youth are not passive consumers of media, but like to take an active role in its production. In class, foreign language teachers have been crafting activities that allow students to construct L2 media artifacts of their own, leading to dramatically increased student engagement. By being built upon modular components written in Flash -- one of the most widely known web programming platforms available -- the AR framework makes it easy for students with rudimentary web-programming knowledge to create their own games. A team of students could likely produce games far exceeding anything available today. Through incentivization with offers of advertising or other revenue-sharing, students themselves could become the primary authors of games using the AR framework (ala youtube & kongregate). Hopefully their teacher would give them extra credit too.
Authentic, Live, Content
Language Pedagogy Theorists often emphasize the value bringing authentic materials, rather than simply making up artificial scenarios. Unfortuantely this becomes complicated in a traditional publishing process, often leading to materials that lack a current or authentic feel to them. Teachers then often end up needing to bring in their own authentic materials to suppliment those provided by curricular designers.
Using the XMLE4X components offered within the Animal Rescue framework, game developers can connect in sites such as craigslist housing ads (or the L2 equivalent). A designer could, for example, make the first task for student be to find a home for their character. Students would browse through live listings and photographs overlayed ontop of satellite photography of the location. After weighing the relative costs and advantages of the houses available on the market, they would then pick one. The game would then be able to extract and use the cost of the house, number of bedrooms, crime statistics in that neighborhood, pollution data, seizmic activity in the area, or virtually any other iinformation available about the house the student chose to execute game events.
In the demo, you may notice a set of three photographs in the top right (click to enlarge them). These represent the three most current photographs taken in the area surrounding your character which people throughout the world have uploaded to the popular photo site panoramio.com.
