See through them: a framework for inferring the cognitive states of puzzle game players via eye gaze

In this paper, we investigate the potential of eye gaze modality in understanding the continuous cognitive states of the players when they are involved in an episode-based puzzle game and how this phenomenon can contribute insights to the game designers to achieve a better QoE of the game. We selected a puzzle game called "Machinarium" as the experimental interface to experiment. We collected the gaze data of the players and inferred their cognitive states in each intersection of decision-making from a game level. The inferred cognitive states were compared to the ground-truth experiences from the players via questionnaire and the official visual guidance extracted from the walkthrough of the game level. The results showed that the implemented framework could infer the cognitive states of the players in a guaranteed accuracy. Besides, the similarities and differences between the players' actual performance and the game level's visual guidance could be the feedback to impact the further optimization of the game design.