Increasing eyewitness identification accuracy in lineups using 3D interactive virtual reality

准确的证人识别(ID)是警方调查和国家安全调查的基石。然而，在过去的一个世纪里，用于显示阵容的技术并没有根本改变。在世界范围内，警方向目击者展示了2D摄影阵容。目击者在身份测试中经常出错--对现实世界警察身份识别游行的分析表明，目击者往往会将已知的无辜嫌疑人指认为罪犯。目击者的错误可能会造成可怕的后果，大约70%的错判都与目击者的错误有关。该项目实施了3D图像技术和虚拟现实(VR)的最新进展，以提高证人身份识别的准确性。我们将开发一种新的阵容程序，使目击者能够旋转3D人脸，并在VR中查看它们移动到不同角度。我们的项目汇集了来自英国(伯明翰大学、斯特林大学)、加拿大(维多利亚大学)和德国(马克斯·普朗克人类认知与脑科学研究所和柏林洪堡大学)的国际学者和从业人员组成的团队。研究人员在各自领域中名列前茅，拥有开发交互式3D VR阵容系统所需的独特专业知识。我们的跨国方法将允许广泛的最终用户进行可靠的测试和外部验证。我们的目标如下：1)开发新的方法来创建人脸的3D表示，允许编辑和调整外观属性(例如，表情、光照)以及在VR中进行真实感(神经)渲染的方法；2)进行实验室实验以确定交互式3D VR阵容相对于2D阵容的优势，以及主动相对于被动人脸探索的优势；3)测试不同类型的检索线索(例如，垂直移动、面部表情、光照)对准确性和可靠性的影响；4)确定与阵容表现相关的个人差异特征(例如，人脸处理能力、对交互式3D VR阵容的参与度)；5)探索在人脸处理能力与阵容表现(例如，辨别准确性)和2D和3D任务中模型人脸处理之间的关系；以及6)向警察从业者展示结果，以制定在一系列刑事司法背景下实施交互式3D VR阵容的政策和实践建议。总而言之，我们的项目结合了各自领域领先专家的理论和最先进的计算技能，大大改进了应用环境中的阵容程序。

Accurate witness identification (ID) is a cornerstone of police inquiries and national security investigations. Yet, the technology used to display lineups has not fundamentally changed over the past century. Worldwide, police present witnesses with 2D photographic lineups. Eyewitnesses often err on ID tests-analyses of real-world police identification parades reveal that witnesses quite often identify known-innocent suspects as culprits. Eyewitness mistakes can have dire consequences and been implicated in about 70% of wrongful convictions. This project implements recent advances in 3D image technology and virtual reality (VR) to increase witness ID accuracy. We will develop a new lineup procedure that enables witnesses to rotate 3D faces and view them moving into different angles in VR. Our project brings together an international team of academics and practitioners from the United Kingdom (University of Birmingham, University of Stirling), Canada (University of Victoria), and Germany (the Max Planck Institute for Human Cognitive and Brain Sciences, and Humboldt-Universität zu Berlin). The researchers are among the best in their fields and have the necessary and unique expertise to develop the interactive 3D VR lineup system. Our cross-national approach will allow for robust testing and external validation by a wide range of end users. Our objectives are as follows: 1) develop novel methods for the creation of 3D representations of human faces that allow for editing and adaptation of appearance properties (e.g., expressions, lighting) as well as methods for photorealistic (neural) rendering in VR; 2) conduct laboratory experiments to establish the advantage of interactive 3D VR lineups over 2D lineups, as well as active over passive face exploration; 3) test the effect of different types of retrieval cues (e.g., vertical movement, facial expressions, lighting) on accuracy and reliability; 4) identify the individual difference characteristics (e.g., face processing ability, engagement with interactive 3D VR lineups) associated with performance in lineups; 5) explore the relationships between individual differences in face processing ability and lineup performance (e.g., discrimination accuracy) and model face processing in 2D and 3D tasks; and 6) present results to police practitioners to develop policy and practice recommendations for the implementation of interactive 3D VR lineups in a range of criminal justice contexts. In sum, our project combines theory and state-of-the art computing skills of the leading experts in their respective fields to substantially improve lineup procedures in applied settings.