Theoretical Background

The concept of this approach is based on reinforcement learning. Reinforcement learning originates from behavioral science and it has progressed as a well-established learning algorithm in machine learning and in computational neuroscience. In reinforcement learning the primary goal of a subject is to achieve reward; however, this expected reward can be different from the received one.

Reward

The purpose of the research is to help educational games to acquire an improved performance and to provide a higher level of engagement. Maintaining attention is one of our primary objectives. The framework achieves this by applying different kinds of feedback during the ongoing gameplay. We distinguished between three types of feedback: difficulty, reward type and reward value. In an initialization step the game is responsible to register its rewards to the system.

The difficulty offers how complex and challenging the next section of gameplay should be to maintain attention. It mostly depends on the previous section and the efficiency of the solution. The experts decide different difficulty levels in the game according to the educational methods they use. The reward value represents the quantity of reward received after the actual gameplay. Although, reward experience is increasing by receiving greater reward, it can be extended by introducing the reward type, because motivation can vary among age groups and gender groups.

Sensors

EEG

EEG can identify the reward prediction error. The feedback-related negativity (FRN) as a frontocentral negativity appears after negative feedback. The FRN tracks the difference between the values of actual and expected outcomes, or reward prediction error.

Facial Gestures

The recent times, tablets are equipped with built-in front-facing cameras, so that we can capture facial expression of subjects. During a game session these facial gestures reflects the difficulty of the game, it can be obtained via machine vision algorithms. After the subject received the reward, it facial expression represents the effect of the obtained reward.

Response Time

Response time provides an information how fast the subject could solve the task and how fast it could complete the game session.

Supervisor

The framework enables the supervisor not only to monitor the measurement process, but to review processes by recommending the next difficulty level.

Monitoring Heart Rate Variability

Heart rate variability (HRV) power spectrum is a technique for measuring current mental effort as a function of time.

Eye tracking

The changing diameter of pupil characterizes the mental attempt to solve a certain task.

Current research focus + results