How do we evaluate intelligent systems that are designed to create fun experiences? Subjective evaluations are possible where one asks how much fun one has. However, more objective measures are also possible such as:

• Affect, measured through paper tools or through biometrics
• Engagement
• Replayability
• Discretionary time on task
• Intrinsic interest

One issue that comes up -- especially when considering customized game content generation -- is that if an experience is personalized based on a player model and thus unique, then between-player experiences cannot be easily compared.