说谎行为的自发性眼动特征研究
文献类型：学位论文
| 作者 | 李俊鸿
|
| 答辩日期 | 2024-06 |
| 文献子类 | 硕士 |
| 授予单位 | 中国科学院大学 |
| 授予地点 | 中国科学院心理研究所 |
| 其他责任者 | 高文斌 |
| 关键词 | 测谎 自发性眼动特征 神经网络模型 滞后序列分析 |
| 学位名称 | 应用心理硕士 |
| 学位专业 | 应用心理 |
| 其他题名 | A study of spontaneous eye movement characteristics of lying behavior |
| 中文摘要 | Deception and lies are common phenomena in human society ， and the identification of lies has always been an important research topic in the field of science. People are eager to identify lies in order to understand the truth and make correct decisions ， so research on lie detection technology is constantly evolving. Among them，non-contact eye tracking technology has attracted attention due to its concealment，and the emergence of this technology enables people to more effectively identify lies，which is incomparable to traditional contact lie detection technology. With the advent of the big data era ， using datasets to train neural networks has become an effective method，which has become one of the hottest research directions currently. This study combines eye tracking technology with neural network models， aiming to explore new effective indicators in the field of eye tracking lie detection based on traditional eye tracking indicators. This study consists of two sub studies: Study 1 used a laboratory experiment method，with 57 college students as the research subjects. Before the interview began， participants were asked to fill out a social attitude survey questionnaire ， which included 20 social attitude questions. Participants were asked to score 7 points based on their true thoughts in order to collect their extreme beliefs. During the interview， the subject was asked about their extreme beliefs，and their eye movement behavior during the question answering task was used as a lie detection indicator to explore the eye movement trajectory characteristics under different task conditions (planned lying， planned truth ， spontaneous lying ， spontaneous truth). After the interview ， the participants completed a post interview questionnaire on the difficulty and honesty of their responses. Study 2 collected eye movement data from Study 1，preprocessed and cleaned the data，divided the dataset into a test set and a training set，and used a feedforward neural network to construct a spontaneous eye movement feature prediction model for lying behavior，exploring the predictive ability of eye movement feature indicators to reveal lying behavior. As a result，it was found that: (1) Individuals who engage in temporary lying during the interview process exhibit a gaze shift pattern in the human body area that is different from planned lying， planned truth，and temporary truth. Specifically，the gaze shifts more significantly in the left area of the human body，including the face，chest，and right shoulder，as well as facial self transfer. Compared to the other three experimental conditions ， the temporary lying condition showed a more pronounced tendency for gaze transfer in the face，chest，and right shoulder regions，with only one difference in self transfer compared to the other conditions. (2) Traditional eye movement indicators such as total fixation duration，average fixation duration，fixation frequency，and eye skip frequency did not show significant differences under different lying conditions，but there were significant differences between different regions of interest. (3) Compared with traditional eye tracking indicators，a post hoc comparative analysis of horizontal and vertical eye tracking acceleration，as well as eye tracking acceleration under four different experimental conditions ， found that there was a significant difference in horizontal eye tracking acceleration between temporary lying and temporary truth，a significant difference in vertical eye tracking acceleration，and a significant difference in eye tracking acceleration. Temporary truth is faster than temporary lying in terms of horizontal eye movement acceleration ， vertical eye movement acceleration，and eye movement acceleration. (4) The classification of planned lies and temporary lies，as well as planned lies and temporary truth，using neural network models exceeded the random level. The combination of eye tracking technology and neural network models has shown great potential in the field of lie detection. By analyzing eye tracking data and training neural network models，we can more accurately identify the lying behavior of participants. This discovery is not only of great significance in the fields of psychology and sociology，but may also play an important role in legal and judicial practice. |
| 英文摘要 | 欺骗与谎言是人类社会普遍存在的现象，对于谎言的识别，一直都是科学领域研究的重要课题。人们渴望识别谎言，从而了解事实真相，作出正确决策，因此对测谎技术的研究也在不断更新换代。其中，非接触式眼动追踪技术因其隐蔽性而引人瞩目，这一技术的出现使得人们可以更有效地识别谎言，而这是传统接 触式测谎技术所无法比拟的。随着大数据时代的到来，利用数据集对神经网络进行训练已成为一种有效的方法，这已成为当前最热门的研究方向之一。本研究将眼动追踪技术与神经网络模型相结合，旨在基于传统眼动指标探索眼动测谎领域 的新型有效指标。 本研究包括两个子研究：研究一采用实验室实验的方法，以 57 名高校学生 为研究对象，在访谈开始前，要求被试填写社会态度调查问卷，问卷包含 20 道 社会态度问题，要求被试根据真实想法进行 7 点评分，从而收集被试极端信念。 在访谈过程中，主试对被试的极端信念进行提问，通过其在回答问题任务下的眼动行为作为测谎指标，探索不同任务条件下（计划说谎、计划真话、临时说谎、 临时真话）的眼动轨迹特征。访谈结束后，被试完成有关回答难度和诚实度的访后问卷。研究二采集研究一的眼动数据，对数据进行预处理和清洗，将数据集分为测试集和训练集，采用前馈神经网络构建说谎行为的自发性眼动特征预测模型，探索眼动特征指标对揭示说谎行为的预测能力。结果发现： （1）在访谈过程中进行临时说谎的个体，其在人体区域表现出区别于计划 说谎、计划真话和临时真话的注视转移模式，具体表现为注视在包含脸部、胸部和右肩部在内的人体左侧区域进行更为显著转移，以及脸部的“自转移”。同时临时说谎条件相较于其他三种实验条件，在脸部、胸部以及右肩三个区域存在更加明显的注视转移倾向，同时仅有一个区别与其他条件的自转移。 （2）总注视时长、平均注视时长、注视次数、眼跳次数等传统眼动指标在不同说谎条件下未出现显著差异，但在不同兴趣区之间有显著差别。 （3）对相较于传统眼动指标，在对四种不同实验条件下的水平眼跳加速度、 垂直眼跳加速度以及眼跳加速度进行事后对比分析发现，在临时说谎与临时真话之间，水平眼跳加速度存在边缘显著差异、垂直眼跳加速度存在显著差异、眼跳加速度存在显著差。临时真话无论是在水平眼跳加速度、垂直加速度以及眼跳加速度要快于临时说谎。 （4）通过神经网络模型对计划说谎和临时说谎、计划说谎和临时真话进行分类，分类结果超过了随机水平。 眼动追踪技术与神经网络模型相结合，在测谎领域呈现出了巨大的潜力。通过对眼动数据的分析和神经网络模型的训练，我们能够更准确地识别被试的谎言行为。这一发现不仅对心理学和社会学领域具有重要意义，还可能在法律和司法实践中发挥重要作用。 |
| 语种 | 中文 |
| 源URL | [http://ir.psych.ac.cn/handle/311026/48138]  |
| 专题 | 心理研究所_健康与遗传心理学研究室 |
| 推荐引用方式 GB/T 7714 | 李俊鸿. 说谎行为的自发性眼动特征研究[D]. 中国科学院心理研究所. 中国科学院大学. 2024. |
入库方式： OAI收割
来源：心理研究所
浏览0
下载0
收藏0
其他版本
除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。