Game designers have long employed play testers for the purpose of game refinement [9]. Traditional methods include observational studies of players [10], videotaping players, asking players to talk out loud while they are playing, and question-and-answer sessions [1]. These commonly used methods provide insight into how players react to the game and high-level information on the strengths and weaknesses of a particular level. Although direct interaction with play testers provides useful information, these methods suffer from a few key limitations. First, players may not be able to articulate their impressions accurately and may ignore important design flaws, resulting in incomplete data. Second, since the designers must spend time interviewing each play tester individually, the time required to execute these kinds of playtests does not scale well to large groups of play testers.