认知与发展心理学研究室知识库

欺骗是人类社会交互中的常见现象，在心理学研究中被定义为有意地误导他人的行为。人类的欺骗识别准确率仅仅略微高于随机水平，大约为54%。提高识别准确率是当前研究和应用领域面临的一个重要挑战。以往研究表明多线索有助于提高欺骗识别的准确率，达到67.86%；反馈有助于提高识别准确率。本研究使用经典的欺骗识别测试探讨欺骗识别中反馈和多通道线索的作用及认知机制。
研究一（实验1）考察反馈和多通道线索在欺骗识别中的作用。实验将被试随机分配到反馈组和无反馈组，完成欺骗识别测试。被试任务是观看录像并判断录像中模特的行为是否在误导他人。结果发现，反馈和多通道线索都提高了被试的识别准确率。
研究二（实验2和实验3）探讨欺骗识别中反馈作用的认知机制。实验2中被试完成欺骗识别测试，被试任务是观看录像并判断录像中模特的行为是否在误导他人，被试可以选择是、否、不确定。结果发现，以不确定反应次数大于0、5、10为标准，无论视觉线索还是视听线索条件下，反馈组不确定反应的人数均显著少于无反馈组；剔除不确定反应后，发现反馈组与无反馈组的识别准确率没有显著差异。比较实验2与实验1的结果表明，反馈能够降低被试在欺骗识别中的不确定性。实验3通过眼动指标考察欺骗识别中反馈对非言语视觉线索的注意加工的影响。结果发现，被试对脸部的注视时间最长、注视点个数最多，手部次之，腿和脚部最少，反馈组与无反馈组对非言语视觉线索的注意模式没有差异；反馈组的眼跳时长显著大于无反馈组。这表明脸部线索是识别者最主要的信息来源，而反馈对识别者在非言语视觉线索上的注意模式没有影响；眼跳指标表明反馈可能影响识别者对线索的认知加工深度。
研究三从视觉通道和听觉通道角度分别考察欺骗识别中多通道线索识别优势的认知机制。实验4通过眼动指标考察视听多通道线索条件下被试对非言语视觉线索的注意加工。结果发现，视听多通道线索条件下被试对非言语视觉线索的注视点个数更少；被试在腿和脚部线索的注视时长更短。这表明在视听多线索条件下识别者可能对非言语视觉线索的注意资源分配减少。另外，本研究编码建立标注副言语指标的欺骗语音数据库，并通过相关分析探讨多通道识别优势的相关线索和指标。结果发现，视听多通道线索条件下的被试识别准确率与副言语指标存在显著正相关，项目识别准确率与副言语指标——重复次数存在显著负相关。
本研究使用欺骗识别测试探讨欺骗识别中反馈和多通道线索的作用及认知机制。本研究的发现有助于丰富和发展欺骗识别研究的理论，有助于对欺骗识别训练提供理论参考。另外，编码建立标注副言语指标的欺骗语音数据库有助于为计算机领域的谎言自动识别研究提供训练数据支持。

Deception is a common aspect of human social interaction. In psychological research, deception is defined as a deliberate attempt to mislead others. Previous studies have shown that people achieved only a slightly better than chance accuracy, about 54%. Improving deception detection accuracy is now a challenge in research and application. Previous studies had shown that deception detection from multiple cues could achieve 67.86% accuracy. Studies had also revealed that feedback training observers are better at deception detection performance than control observers. However, how do feedback and multiple cues play a role in deception detection remains to be investigated. In the current study, we used the typical lie detection test to investigate the role of feedback and multiple cues in deception detection, and the cognitive mechanisms.
In study 1, we conducted a lie detection test. During the test, participants were randomly assigned to the feedback group or the control group. They were instructed to watch video fragments of 6 models and judge whether the model’s behavior was misleading or not. Results showed that feedback observers had a better performance than control observers. The accuracy of audiovisual cues condition was better than visual cues condition. There was no interaction between feedback and cue modalities.
In study 2, two experiments were conducted to investigate the mechanism of feedback in deception detection. In experiment 2, the participants were also asked to complete a lie detection test. Besides the truth-lie option, they were given an additional choice of indicating their uncertainty. Chi-squared tests were conducted to analyze the frequencies of participants who made uncertain responses more than zero, five and ten in each condition. The results showed that there were significantly less participants who made uncertain responses in the feedback group than in the control group. What’s more, after eliminating the uncertain responses, there was no main effect of feedback in experiment 2. Compared with experiment 1, the results suggested that feedback could reduce the observer’s uncertainty in deception detection. In experiment 3, we used eye-tracking technology to investigate whether feedback could affect participant’s attentional processing of nonverbal visual cues in deception detection. Analysis of fixation counts and fixation duration showed that both feedback observers and control observers allocated much more attentional resources to face than to hand, leg and foot. But the feedback observers had a longer saccade duration than the control observers. These results suggested that the participants relied on the face value mostly in deception detection, and feedback did not affect the observer’s attentional modes. However, feedback might deepen cognitive processing of the nonverbal visual cues.
In study 3, we investigated the mechanism of multimodal cue’s advantage in deception detection from the visual modality and auditory modality separately. In experiment 4, we used eye-tracking technology to investigate the attentional processing of nonverbal visual cues in multimodal cues based deception detection. Results showed that observers had less fixation counts on nonverbal visual cues, and shorter duration on the leg and foot cues in the multimodal cues condition than in the visual cues condition. The results suggested that less attentional resources were allocated to the nonverbal visual cues in the multimodal cues condition. Furthermore, we coded eight paraverbal indicators and constructed a speech database for deception detection research. Then, we did correlation analysis on the observer’s accuracy and the Likert 7 rating of cues used in the lie detection tests, also on the item’s accuracy and the eight paraverbal indicators. Results showed that there was a significant positive relationship between the observer’s accuracy and the paraverbal cues, and a significant negative relationship between the item’s accuracy and repetition.
Overall, the study confirmed the positive roles of feedback and multimodal cues in deception detection and theoretically explored the correlated cognitive mechanisms. The results would contribute to deception detection studies in both theory and practice. Besides, the database could serve the purpose of automatic deception detection studies in computer science.