近几年来，心理测验在人事与组织领域中发展十分迅速，大部分企业已经意识到人格的重要性，逐步将人格测验运用到人才选拔当中。边缘型人格作为一种负向人格，无论对个人还是组织都会带来一定的影响，所以组织有必要将此类人群筛选出来，作为人才选拔中的重要参考因素。目前为止，关于边缘型人格测验工具的使用，主要包含诊断量表与筛查量表，而这两种类型的测验工具，一是对主试要求高，二是受社会赞许性的影响，应聘者的作假动机更强，所以均不适用于在组织中进行大规模的团体施测。条件推理测验是一种测量内隐人格的新型工具，由于采取间接测量的方式，具有较好的掩蔽性，让受测者很难猜出测验的真实目的。但该测验选项的设置限制了受测者的思维，同时选项也会给受测者起到提示的作用，使结果具有局限性与偏差。有研究指出，通过分析个体所写的文本，可以挖掘出该个体的人格特质。所以，本研究希望开发建构反应式边缘型人格条件推理测验，对被试的作答文本进行分析，建立边缘型人格的预测模型，用于组织的招聘选拔中。本研究包括三部分：

研究一开发测验工具，即建构反应式边缘型人格条件推理测验的开发。根据施测结果修改题目、探讨测验的施测形式，并对其进行信效度的检验。研究二根据对大学生群体在题目上作答的文本进行人工编码，形成题目的标准答案，基于word2vec和标准答案的相似度方法，建立边缘型人格的预测模型。研究三根据量表结果将数据进行分类，基于word2vec建立机器学习模型的分类方法，探究边缘型人格在二分类下，预测模型的建立效果。

结果可知，该自编测验具有良好的信度与效度。对被试作答的文本进行分析，利用word2vec建立机器学习模型，使用朴素贝叶斯的算法，建立的边缘型人格的预测模型效果最好，模型的Precision_-为0.7500、Recall₊为0.4516、macro-F1为0.5624、G-mean为0.5541，在测试集中，筛选出的边缘型人格倾向的人群占比为13.21%。

In recent years, psychological test has developed rapidly in personnel and organization fields. Most enterprises have realized the importance of personality, and gradually applied personality test to talent selection. Borderline personality, as a negative personality, has certain influences on both individuals and organizations. Therefore, it is necessary for organizations to screen out this kind of people as a reference factor for recruitment and selection. Diagnostic scale and screening scale are mostly used for borderline personality test which are not suitable for large group testing in organizations since high requirments for the interviewers, and applicants have stronger motivation to cheat because of social approval. A new tool, conditional reasoning test, measures implicit personality based on its indirect measurement which is difficult for the subjects to speculate the actual purpose of this test. However, the setting for the test options limits the subject thinking, while these options give the subject a reminder. In consequence the results have limitations and biases. Some studies have pointed out that through analyzing the text written by an individual, the personality traits of that individual can be exposed. Therefore, this study intends to develop the constructed response conditional reasoning test for borderline personality, analyze the answers of the subjects, and establish a predictive model of the borderline personality, which can be used in the recruitment and selection in organizations. This study consists of three parts:

Study 1 will develop a test tool, which the constructed response conditional reasoning test for borderline personality. According to the result of the test, Revising the title, discussing the test form, and testing the reliability and validity of the test. Study 2 will accord to the text the college students answer on the question, the standard answer of the question is encoded by hand, and the model of margin personality is established based on the similarity method of word2vec and the standard answer. Study 3 will classified the data according to the results of the scale. The classification method of machine learning model based on word2vec is applied to explore the effect of prediction model of borderline personality under binary classification.

The results showed that the self-designed test had good reliability and validity. Analyzing the text of the answers, using word2vec to build the machine learning model, using the naive Bayes Algorithm, the prediction model of the borderline personality is the best. The Precision_- was 0.7500, Recall₊ was 0.4516, macro-F1 was 0.5624, G-mean was 0.5541, and the proportion of borderline personality was 13.21%.