Our behavior is partly a product of our perception of the world, and aggressive individuals have been found to have ‘hostility biases’ in their perception and interpretation of social information. Four types of hostility biases can be distinguished: the hostile attribution, interpretation, expectation, and perception bias. Such low-level biases are believed to have a profound influence on decision-making, and possibly also increase the likelihood of engaging in aggressive acts. The current review systematically examined extant research on the four types of hostility bias, with a particular focus on the associations between each type of hostility bias and aggressive behavior. The results confirmed the robust association between hostility biases and aggressive behavior. However, it is still unknown how exactly hostility biases are acquired. This is also caused by a tendency to study hostility biases separately, as if they are non-interacting phenomena. Another issue is that current approaches cannot directly quantify the latent cognitive processes pertaining to the hostility biases, thus creating an explanatory gap. To fill this gap, we embedded the results of the systematic review in a state-of-the-art computational framework, which provides a novel mechanistic account with testable predictions.

我们的行为在一定程度上是我们对世界的理解的产物，而且好斗的人被发现在对社会信息的理解和解释中存在“敌意偏见”。可以区分四种类型的敌对偏见：敌对归因，解释，期望和感知偏见。这种低水平的偏见被认为会对决策产生深远的影响，并且可能还会增加参与侵略性行为的可能性。本篇综述系统地检查了对四种类型的敌对偏见的现有研究，特别关注每种类型的敌对偏见与攻击行为之间的关联。结果证实了敌对偏见和攻击行为之间的牢固关联。但是，仍然不清楚如何精确地获得敌对偏见。这也是由于倾向于单独研究敌对偏见而引起的，就好像它们是非相互作用的现象一样。另一个问题是，当前的方法无法直接量化与敌对倾向有关的潜在认知过程，从而造成解释上的差距。为了填补这一空白，我们将系统评价的结果嵌入了最新的计算框架中，该框架提供了具有可测预测的新颖机制。