Abstract:

Within educational games an essential problem is how to gather user information without destroying the game play. Even though, there exist several possibilities of tracking the user behaviour, most of the objective behavioural data are ambiguous and lack subjective meaning. However, the assessment of subjective data by, e.g., questionnaires either destroys the game play or could be only made when the game has finished. The proposed methodological framework illustrates how the combination of data tracking by logfiles and Natural Language Processing (NLP) could lead to a more sophisticated interpretation of user behaviour which in turn enables an optimization of adaptive user support in the context of serious games.
Data tracking by logfiles as well as NLP can be considered as non-reactive methods without demand effects. They are not dependent of the compliance of the player and in most cases the player is not even aware of the data-assessment.
Logfile-data delivers quantitative objective data regarding the activity of the user (e.g., assessed by click-rate) or his/her direction of behaviour (e.g., interaction with a specific NPC). Information gathered by NLP could enrich the quantitative logfile-data by qualitative information, e.g., if the interaction/dialogue with another player is friendly or hostile. Thus, logfile-data and NLP-analyses deliver divers information about the player. The combination of both methods provides not only an extended view of the players but could also deliver a surplus meaning. An example is the analysis of the interaction between two players: The pure activity level and interaction rate indicates if and how extensive the players are in contact. The NLP information could give evidence if the two players like or dislike each other. The combination of both information provides information if they have an intensive useful collaborative or a destructive quarrel. Accordingly, the combination of logfiles and NLP could also lead to different interventions of the system. If there are only a few negative comments there might be no need for an intervention (of the system). However in the case of a quarrel the system should lead the players to another scenario which might help to resolve the conflict. Contrariwise, in the case of positive comments the system should rather react if the interaction rate is to low.
This example illustrates also, that the rules for combining the information of logfiles and NLP as well as the accordingly adaptive functionalities of the system has to be carefully defined. The rules could be based on psychological and pedagogical theories and/or on existing research findings on games and game-based learning.

Analogous to the example described, the combination of logfiles and NLP could also lead to a more precise and sophisticated assessment of player’s motivational state, the player’s interest for a specific topic or tool, the (para)social interaction of a player or the shared interests of a subgroup of users within a multiplayer scenario.

The described methodological framework can be used for optimizing user-support and adaptive gaming not only in the context of serious games but also in other e-learning environments or pure entertaining video games. The described indices and cases are only exemplarily and are open for further extensions.

Keywords:

Educational games, data-tracking, non-invasive interventions, logfiles, Natural Language Processing.