Hiding in Plain Sight: A Turing Test on Fake Persona Spotting

隐藏在众目睽睽之下：针对假冒角色识别的图灵测试

• 批准号: 2590138
• 金额: --
• 依托单位: Lancaster University
• 依托单位国家: 英国
• 项目类别: Studentship
• 财政年份: 2019
• 资助国家: 英国
• 起止时间: 2019 至 无数据
• 项目状态: 已结题
• 来源: https://gtr.ukri.org/projects?ref=studentship-2590138
• 关键词: Hiding; Plain; Sight; Turing; Test

My PhD research focus is on fake identities in the online space, specifically social media, and how these fake identities present in the form of fake social media profiles. The main overall aim of the research is to investigate whether humans can accurately detect fake social media profiles and the processes they use to detect said profiles. The bulk of the current literature on fake social media profiles and how they can be detected is derived from computer science researchers, who have developed a series of algorithms or data bots (Efthimion et al., 2018; Van der Walt & Eloff, 2018) to detect and remove said profiles from social media platforms. However, there is a lack of research in this area from a psychological perspective, specifically in relation to what is actually analysed on the profile to conclude that it is fake (i.e. the profile characteristics, and who the people are behind the creation of such profiles). To investigate these issues, we seek to extend traditional research on social perception and 'realistic accuracy judgements' (Funder, 1999) to the domain of real and fake persona by applying a 'Turing test' style methodology.A total of three studies have been completed so far, all of which involved the same methodology: participants were required to view a series of fake and real Facebook profiles (screenshot images of the profile) and asked to judge the authenticity of the profile. The fake profiles were created specifically for this research using google images and photoshop. The information regarding the characteristics of the profiles that were manipulated were gathered from statistical reports regarding Facebook use, and the limited literature available on the topic. The real profiles were obtained from people known to the researcher to allow for them to be officially authenticated as real. Once participants had made their authenticity judgement of the profiles, they were asked to click on the areas of the profile they used when making their judgement. The purpose for this being that it would allow us to see which characteristics of the profile are used the most and thus may provide us answers as to which characteristics to look out for when judging a profile ourselves. Results from these three studies indicate that humans overall are not very accurate at judging the authenticity of social media profiles, a result that was predicted due to the larger amount of literature on human deception detection accuracy being around the level of chance (DePaulo, 2004). Additionally, consistently across each study, participants clicked the most on the photos on the profile, which came under the umbrella characteristic of Photo-Type, more than any other characteristic, suggesting an overreliance on the visual imagery presented on the profile and less so on the language used and items posted on the 'wall'. The fourth study that is currently ongoing is using the same methodology as the previous three, with the introduction of a different culture. The study will be comparing profiles from Western culture to Eastern cultures, specifically India. India is the country with the highest Facebook use around the world, with over 40 million users, and so is an area of great interest to this research. The main aim of this study is to analyse whether humans from one culture are more accurate at detecting fake profiles from another culture or from their own culture. Further, the aim is to also understand what constitutes a fake social media profile in another culture, i.e., what areas of the profile are they looking at/using when making their authenticity judgement. Do these areas differ from the results found previously from Western participants? To test this, the same methodology will be used alongside a series of new Indian fake profiles that have been created and real Indian profiles that are being sourced from International students at Lancaster University.

我的博士研究重点是在线空间（特别是社交媒体）中的虚假身份，以及这些虚假身份如何以虚假社交媒体资料的形式呈现。该研究的主要总体目标是调查人类是否能够准确检测虚假社交媒体资料以及他们用于检测所述资料的过程。当前关于虚假社交媒体资料及其检测方法的大部分文献都来自计算机科学研究人员，他们开发了一系列算法或数据机器人（Efthimion 等人，2018 年；Van der Walt & Eloff，2018 年）从社交媒体平台检测并删除所述个人资料。然而，从心理学角度来看，该领域缺乏研究，特别是在对个人资料进行实际分析以得出其是假的结论方面（即个人资料特征以及创建此类个人资料的幕后人员是谁） ）。为了研究这些问题，我们寻求通过应用“图灵测试”风格的方法论，将社会感知和“现实准确性判断”（Funder，1999）的传统研究扩展到真假角色领域。总共进行了三项研究到目前为止，所有这些都涉及相同的方法：参与者被要求查看一系列虚假和真实的 Facebook 个人资料（个人资料的屏幕截图），并要求判断个人资料的真实性。这些虚假的个人资料是专门使用谷歌图像和 Photoshop 为这项研究创建的。有关被操纵的个人资料特征的信息是从有关 Facebook 使用情况的统计报告以及有关该主题的有限文献中收集的。真实的个人资料是从研究人员认识的人那里获得的，以便他们能够被官方认证为真实的。一旦参与者对个人资料的真实性做出判断，他们就被要求点击他们在做出判断时使用的个人资料的区域。这样做的目的是让我们能够看到配置文件的哪些特征被最常使用，从而可以为我们提供在自己判断配置文件时要注意哪些特征的答案。这三项研究的结果表明，人类总体上在判断社交媒体资料的真实性方面并不是很准确，这一结果是由于大量关于人类欺骗检测准确性的文献都处于偶然水平而预测的（DePaulo，2004） 。此外，在每项研究中，参与者对个人资料上照片的点击次数一致，这些照片属于照片类型的总括特征，比任何其他特征都多，这表明他们过度依赖个人资料上呈现的视觉图像，而不是其他特征。使用的语言和张贴在“墙上”的物品。目前正在进行的第四项研究使用与前三项相同的方法，并引入了不同的文化。该研究将比较西方文化和东方文化的概况，特别是印度。印度是全球 Facebook 使用率最高的国家，拥有超过 4000 万用户，因此也是本研究非常感兴趣的领域。这项研究的主要目的是分析来自一种文化的人类在检测来自另一种文化或来自自己文化的虚假个人资料方面是否更准确。此外，目的还在于了解在另一种文化中什么构成了虚假社交媒体资料，即他们在做出真实性判断时查看/使用资料的哪些部分。这些领域与西方参与者之前发现的结果有何不同？为了测试这一点，将使用相同的方法以及一系列新创建的印度假档案和来自兰卡斯特大学国际学生的真实印度档案。