CAREER: Human Behavior Assessment from Internet Usage: Foundations, Applications and Algorithms

人类与计算机的互动，尤其是与互联网的互动，正在不断增加。因此，研究使用电脑的人的思维、行为和态度的网络心理学领域变得越来越重要。迄今为止，网络心理学的研究仅通过自我报告的调查来收集互联网使用数据，这种方法存在许多缺点，包括人为错误，由于用户对社会可取性的关注而产生的偏见，以及所获得数据的数量和维度的限制。PI在这个项目中的愿景是基于真实的互联网使用数据推进人类行为评估，也就是说，在谨慎维护用户隐私期望的同时，在没有人工干预的情况下，连续、被动、不引人注目地收集互联网使用数据。为此，PI将开发实践基础，创建两个支持互联网的应用程序（一个与在线心理保健有关，另一个与在线社交有关），并利用行为心理学的结果设计分类算法，以证明能够获得重要见解，并根据互联网使用情况检测行为相似性（与抑郁症状和社交偏好有关）。该项目建立在PI的初步工作基础上，该工作利用了从思科NetFlow记录中收集的真实互联网使用数据，以确定与抑郁症症状相关的细粒度互联网使用特征，并将以大学生为主要研究对象。PI将与心理学和社会科学、临床精神病学以及联邦调查局网络犯罪部门的专家合作开展当前的项目。结果将包括对受误差和偏差影响最小的互联网使用情况的准确描述，以及用于评估人类行为的大量高粒度互联网使用特征，这些结果将产生易于适用于实际用途的见解和结论。更广泛的影响：这项研究将提高我们基于互联网使用数据评估人类行为的能力。在线精神卫生保健应用程序将提供互联网支持的主动、早期和成本效益高的精神卫生保健，以补充现有临床环境中的护理，而在线社交应用程序将通过利用相互的社交偏好来改善在线社交体验。该研究在减少网络欺诈和检测网络欺凌方面的应用是立即的。PI将与HBCU和RUI机构以及K-12学校合作，进一步加强项目外展。

Human interaction with computers, and especially with the Internet, is ever increasing. As a consequence the field of cyber-psychology, which studies the thinking, behavior and attitudes of the person using the computer, is of growing importance. To date, studies in cyber-psychology have collected Internet usage data by means of self-reported surveys only, an approach which suffers from a number of drawbacks including human error, bias due to user concerns relating to social desirability, and limits on the volume and dimensionality of the data obtained. The PI's vision in this project is to advance human behavior assessment based on real Internet usage data, that is to say Internet usage data collected continuously, passively and unobtrusively without manual intervention while carefully maintaining user privacy expectations. To these ends, the PI will develop practical foundations, create two Internet enabled applications (one relating to online mental healthcare and the other to online socializing), and leverage results in behavioral psychology to design classification algorithms that demonstrate the ability to achieve significant insights and to detect behavioral similarities (relating to symptoms of depression and preferences in social contacts) based on Internet usage. The project builds on the PI's preliminary work that exploited real Internet usage data collected from Cisco NetFlow records to identify fine-grained Internet usage features associated with symptoms of depression, and will be executed with college students as primary subjects. The PI will collaborate in the current project with experts from the psychological and social sciences, clinical psychiatry, and the FBI's cyber crimes division. Outcomes will include an accurate characterization of Internet usage minimally affected by error and bias, and a large number of Internet usage features of high granularity for assessing human behavior, which together will yield insights and conclusions readily adaptable for practical utility.Broader Impacts: This research will advance our ability to assess human behavior based on Internet usage data. The online mental healthcare application will yield Internet enabled proactive, early and cost -effective mental health care to complement care in existing clinical settings, while the online socializing application will lead to improved online social experiences by leveraging mutual social preferences. Applications of the research to mitigating Internet fraud and detecting cyber bullying are immediate. The PI will collaborate with HBCU and RUI institutions as well as K-12 schools to further enhance project outreach.