社会与工程心理学研究室知识库

自杀是一个严重的社会问题和公共卫生问题，亟待我们深入研究并展开实际工作。已经有学者开始关注通过社交网络平台收集用户的心理健康信息，其中也包括对自杀风险的识别，但是目前仍没有切实可以投入实践的自杀风险识别和干预系统。通过本研究，我们希望探索一条利用微博进行自杀风险识别和干预的新路径，发现更多潜在的有自杀风险的个体，并主动为他们提供帮助，以降低自杀风险。
本课题开展了三个部分的研究：
第一部分，微博用户自杀意念影响因素的结构方程构建，比较自杀意念高低组在微博语言使用特征上的差异。为自杀识别和干预工作提供理论支持。
第二部分，基于微博文本的自杀风险识别，分成两个实验：实验 1，首先设计两种文本识别方式，即关键词筛选和利用机器学习建模筛选。实验 2，考察基于微博文本的自杀风险识别效果。
第三部分，基于微博私信功能的自杀风险干预研究。首先通过焦点小组访谈，探索通过私信干预的可行性。尝试发送私信，进行初步的干预，并通过回访考察私信的干预效果。
主要结果如下：
（1）在微博语言使用特征上，自杀意念高分组和低分组在“焦虑词”和“死亡词”的使用频率有显著差异（P<0.05）。通过数据的拟合，支持了假设模型 2。
（2）相比于关键词筛选的击中率 0.47，虚报率 0.66，使用计算机训练建模的效果更好，击中率为 0.58，虚报率0.25。通过微博文本识别到的有自杀风险的个体经人工检验确实属于自杀高危人群。
（3）通过焦点小组访谈，我们微博私信功能可以作为初步干预的手段联系用户以提供帮助。通过发送私信，获得了用户较为积极的反馈，私信回访的结果也表明，尽管对于大多数用户，我们并没有能直接改变其现状。但是也有用户因为接受了我们的私信而拨打热线电话，并且对于自助干预系统，用户表达了较为积极的态度。
以上的研究结果表明，现代日益蓬勃发展的社交媒体平台，以微博为代表，为我们的自杀识别和干预工作提供了新的契机，通过对微博数据的分析，可以实现对于高危人群的筛查，并且通过微博私信功能主动接触那些有风险的个体，主动为他们提供帮助，可以取得一定的效果，通过微博的预警和干预机制可以有效的提升自杀干预工作的广度和效率。

Suicide is a serious social problem and public health problem, it is urgent for us to study and carry out practical work. Some scholars have begun to collect information about users' psychological health through social network platform, including the recognition of suicide risk, but there is still no real suicide risk identification and intervention system to put into practice. Through this study, we hope to explore a new way to use Weibo to conduct suicidal risk identification and intervention, find more individuals who are in suicide risks, and actively help them to reduce their suicide risk.
This topic mainly includes three studies:
The first study constructs structural equation of influence factors of suicidal ideation for the Weibo users and compares the difference of upper-and-lower groups of suicidal ideation in using characteristics of Weibo language. This study will provide theoretical support for suicide recognition and intervention work.
The second study is divided into two experiments based on risk identification of Weibo text: Experiment 1, the accuracy of two kinds text reorganization ways in identification of single Weibo text is compared, namely keywords selection and model selection through machine learning. Experiment 2, to investigate the effect of suicide risk identification based on micro-blog text.
The third study explores suicide risk intervention based on private messages of Weibo. Firstly, to explore the feasibility of intervention by focus-group interviews. Secondly, to send private messages to help the individuals who are in suicide risk, and estimate the effect of sending private messages.
Main results are as follows:
(1) In terms of using characteristics of Weibo language, there are apparent differences（P<0.05）in the use frequency of upper and lower groups in the word- "anxiety" and word of "death". By fitting of data and presuppose model 2, the fitting is good and the structural equation model is obtained by further path optimization.
(2) Compared with hit rate of keyword filtering-0.47, false rate- 0.66, effect that adopts the computer training model is better, hit rate is 0.58, the false rate is 0.25. And individuals identified by micro-blog's text of suicide risk are indeed high risk groups for suicide.
(3) Through interviews of focus group, private messages can be used as a means of preliminary intervention to contact users to provide help. By sending private messages, we get positive feedback from the users. The results show that, although for the majority of users, we cannot directly change the status quo. But there are still some users called the hotline for help because of the private message. And for self-help intervention system, the majority of users expressed a more positive attitude.
The above researches show that the modern booming social media platforms, represented by Weibo, have provided new opportunity for suicide identification and intervention work. Through the analysis of the Weibo data, the regional restriction can be broken through for screening of high-risk groups, and risky individual can be voluntarily contacted and get help through the private messages, thus obtaining certain effect. Through early warning and intervention mechanism of Weibo, the scope and efficiency of suicide prevention work can be effectively improved.