网络游戏成瘾，是一种行为类型的成瘾。在过去的研究中，很多研究者发现网络游戏成瘾和药物成瘾，以及赌博成瘾在行为和神经影像学上表现上具有很多共同特征，如决策能力的受损，在风险决策中作出一系列不利决策，自我控制和修正自身行为的能力更差，行为更冲动，以及负责风险感知，冲突监控，价值加工的脑区的激活异常等等。但是，很少有研究者从生理角度研究成瘾人群的风险决策特征，并且对于强化学习的重要依据，与不确定性正相关的期望误差在网络游戏成瘾人群中是如何被利用并表现在行为上，我们还不太清楚。本研究使用行为实验与眼动仪，通过在线填写问卷进行筛选的方式，34名大学生和研究生参与了实验，为了更好地评估网络游戏成瘾人群的决策与学习特点，我们针对性地修改了仿真气球模拟任务(BART)，使实验范式可以适用于我们的研究目的。结果发现，网络游戏成瘾人群和健康人群相比，其瞳孔大小更少的受任务之间风险变化的调节，网络游戏成瘾人群的瞳孔对期望误差的敏感度，在爆炸任务中，而不是在不爆炸任务中，要显著的低于健康人群。并且，这种任务之间风险差异导致的瞳孔对期望误差敏感度的差异，对网络游戏成瘾人群带来的调节效应要远远比健康人群低，表现出一种在任务和被试类型之间的交互作用。另外，网络游戏成瘾人群和健康人群的瞳孔对期望误差的敏感度差异在高风险的爆炸条件下和低风险的不爆炸条件下，有选择性的影响了爆炸任务，并且有选择性的影响了在低于期望误差的期望时的行为表现，表现出一种在高风险情境下，瞳孔的期望误差敏感度对行为改变的独特调节效应。最后，我们发现在爆炸任务的第一次反馈阶段，网络游戏成瘾量表的分数，即网络游戏成瘾的严重程度，表现出一种和瞳孔对期望误差敏感度的负相关，但这种负相关仅仅出现在网络游戏成瘾人群中，而不是健康人群中，这表明网络游戏成瘾量表反映的成瘾程度暗示了大脑水平的异常变化，并直接地影响了期望误差对瞳孔的调节。综上，本研究发现(1)在任务间的风险变化，有差异地影响了瞳孔对期望误差的敏感度，网络游戏成瘾人群和健康人群相比，更少地受风险变化的影响(2)在网络游戏成瘾人群和健康人群中表现出的瞳孔对期望误差的敏感度差异，有选择性地影响了爆炸任务中的行为改变(3)网络游戏成瘾人群的成瘾严重程度，直接地影响了瞳孔对期望误差的敏感度，网络游戏成瘾程度越高，瞳孔对期望误差的敏感度越低。

Internet game addiction is a type of behavior addictions. In past, many researchers found that Internet game addiction, drug addiction and gambling addiction have many common behavior and neural characteristics. For example, all have decision deficits in risk decision making, and the people with internet game addiction also show abnormal activities in brain regions involved in self-control, behavior adjustment, risk perception, conflict monitoring and value processing.

However, few studies have investigated the characteristics of internet game addiction in risk decision making by using physiological methods. Moreover, in the field of reinforcement learning, it is not clear how prediction error is used to guide behavior in people with internet game addiction. In order to evaluate the characteristics of risk decision making and learning in people with internet game addiction, we adapted the balloon analogue risk task. Here, the results showed that the people with internet game addiction’s pupil sizes were less regulated by the risk change between tasks. Only in the blast task, the pupil sensitivity to prediction error in people with internet game addiction was significantly lower than the healthy people, and the difference in the pupil sensitivity to prediction error caused by the risk between tasks shows stronger moderating effect in healthy people, showing an interaction between task and subject type. We also found that the difference of pupil sensitivity selectively influenced the behavior performance when the expectation is lower than the prediction error in blast task, showing a unique regulatory effect of pupil sensitivity to prediction error on the high-risk condition. Finally, in the first feedback stage of blast task, the score of Internet game addiction scale, that is, the severity of Internet game addiction, showed a negative correlation with the average response intensity of pupil to prediction error. This correlation only appeared in the people with Internet game addiction, but not in the healthy people. 

In general, our study found that (1) The risk change between tasks had different effects on the sensitivity of pupils to prediction errors, and the people with internet game addiction are less affected than the healthy people. (2) The sensitivity of pupils to prediction errors was differential between the people with internet game addiction and the healthy people in the blast task. (3) The severity in internet game addiction correlated with the sensitivity of pupil to the prediction error. The higher the degree of addiction in online game player, the lower the sensitivity of pupil to the prediction error.