Literacy levels in most developing countries remain shockingly low and formal education is making little progress. MILLEE improves literacy through language learning games on cellphones --- the "Personal Computers of the developing world" --- which are a perfect vehicle for new kinds of out-of-school language learning. Games bring children into rich, immersive environments where they can acquire and use language naturally, while encouraging them to transfer their language skills outside the game.

The MILLEE research project focuses on developing scalable, localizable design principles and tools for language learning. The challenges are (i) to integrate sound learning principles, (ii) to provide concrete design patterns that integrate entertainment and learning, and (iii) to account for cultural and learning differences in children in developing regions. In this thesis, we will describe a framework called PACE that addresses these challenges and eight rounds of fieldwork that had contributed to its development. We will also describe a tool to expedite audio-only learning (Pimsleur Generator), a very important niche for developing regions. We discuss our most recent work which patterns learning games after local children's traditional village games and the benefits this approach offers. Finally, we describe the complex adoption ecology in developing regions, and how MILLEE preserves learning principles while supporting rich localization and customization at multiple stages in the adoption hierarchy.