中国科学院行为科学重点实验室知识库

心理指标识别建模是基于海量数据结合计算机机器学习算法识别心理特征的一种新兴方式。由于传统纸笔测量方式所存在的诸多限制,本文对基于社交媒体数据的心理建模方法及应用于心理测量的可行性进行综述,介绍了特征及提取方法、常用机器学习算法以及应用场景,并对心理指标识别建模的优势和不足进行了总结与展望。该测量方法基于社交媒体数据,相比自我报告法具有时效性高、可回溯测量、生态效度好等独特优势。然而,基于社交媒体的心理指标识别建模方法也在学习成本、硬件成本等方面存在局限性。未来研究人员需要进一步探索社会媒体信息与用户心理变量间的关联机制,并将心理指标识别模型同传统心理学研究方法结合进行更多的探索和应用。心理指标识别建模结合心理测量基本原理和计算机领域机器学习的技术,将为心理学研究打开一扇新的大门。

Modeling psychological indexes (i.e., psych-modeling) is an emerging method that uses machine learning algorithms to identify psychological indexes based on big data. This paper reviews the feasibility of psych-modeling methods based on social media data in the field of psychometrics. Frequently used data types and machine learning algorithms are introduced. Then, we summarize psych-modeling’s application to various scenarios together with its strengths and weaknesses. Compared with traditional self-reporting methods, psych-modeling has some advantages, including better performance in retrospective studies, greater ecological validity, and greater time-efficiency. However, psych-modeling has several limitations. For example, researchers need to spend extra time and effort to learn this new method and bear the inevitable cost of hardware. In future studies, researchers could investigate further how user’s behavior on social media relates to psychological indexes. We also expect psych-modeling will be used in future psychological studies. By combining psychometrics and machine learning, we believe psych-modeling could make great contributions to psychology research and practice in the future.

国家社科基金重点项目(17AZD041);; 国家自然科学基金项目(31700984);; 中国科学院青年创新促进会资助