A rich body of research points to racial biases in so-called police officer dilemma tasks: participants are generally faster and less error-prone to “shoot” (vs. not “shoot”) Black (vs. White) targets. In three experimental (and two supplemental) studies (total N = 914), we aimed at examining the cognitive processes underlying these findings under fully standardized conditions. To be able to dissect a-priori decision bias, biased information processing and motor preparation, we rendered video sequences of virtual avatars that differed in nothing but the tone of their skin. Modeling the data via drift diffusion models revealed that the threat of a social group can be explicitly learned and mapped accordingly on an a-priori response bias within the model (Study 1). Studies 2 and 3 replicated the racial shooter bias as apparent in faster reaction times in stereotype-consistent trials. This, however, appears to result from stereotype-consistent motoric preparations and execution readiness, but not from pre-judicial threat biases. The results have implications especially for automatic stereotypes in the public.

大量研究指出了所谓的警务人员困境任务中的种族偏见：参与者通常更快，更不容易“射击”（而不是“射击”）黑人（与白人）目标。在三项实验（和两项补充）研究中（总共N = 914)，我们的目标是在完全标准化的条件下检查这些发现背后的认知过程。为了能够剖析先验决策偏差、有偏差的信息处理和运动准备，我们渲染了虚拟化身的视频序列，这些虚拟化身除了皮肤色调外别无其他。通过漂移扩散模型对数据进行建模表明，可以明确地学习社会群体的威胁，并相应地映射到模型中的先验响应偏差（研究 1）。研究 2 和 3 复制了种族射手偏见，这在刻板印象一致的试验中反应时间更快。然而，这似乎是由刻板印象一致的运动准备和执行准备造成的，而不是来自司法前威胁偏见。结果对公众的自动刻板印象尤其具有影响。