人才测评的实施过程是评分者、受测者以及测评情境交互作用的动态过程。由于所有的选拔决策、研究结果都基于评分者的评分结果，所以评分者的有效性是非常重要的。评分过程需要评分者从受测者所表现出的诸多行为中辨别、筛选、记录，并通过回忆、概括和判断做出最终评价，是一个高度认知负荷的过程。而对于完整的心理活动而言，认知与人格是一个有机整体，该整体包含两者的共同作用和交互影响。本研究包括两个部分：通过实施录像观察、模拟评分，研究一对33名新手评分者的评分特性进行了分析，并在此基础上开发适合于新手评分者的培训方法；研究二对44名新手评分者实施培训，在保证评分有效的基础上，对他们进行了工作记忆和人格的测量，旨在探讨工作记忆和人格分别对评分者有效性的影响。主要结论如下：研究一发现：对于新手评分者而言，（1）评分有效性比较低，表现在较低的评分者一致性，对评分维度、被评价者较低的区分准确度上；（2）评分在各维度上拉开的差异较大，且较为严格，没有表现出晕轮效应或宽容效应；（3）表现出了专业和经历方面的差异，心理学专业的以及经历较多的新手评分者，相对而言评分有效性更高；（4）记录质量也会影响到评分有效性，记录质量越高，评分有效性越高。在针对以上特性的基础上，（5）设计将行为观察培训和参照系培训相结合的培训方法。研究二发现：首先，通过培训，（1）新手评分者的评分者一致性和评分准确度得以提升；（2）评分仍然较为严格，但在各维度上拉开的差异降低。同时，（3）注意转换和抑制能力越强，沉稳性越高，有助于提升对各测评维度的区分；（4）决断力越强，区分受测者的准确度越高；（5）利他性越强，越容易发生评分者误差，表现在容易发生晕轮效应，以及评分倾向于宽松；而坚韧性越强，越不容易发生晕轮效应。

Personnel assessment is a dynamic processe which reflects the interaction of raters, ratee and rating conditions. Since all of the selection decisions and the research results are based on the scores, rating accuracy is very important. For raters, the appraisal process is of high cognitive loading. They need to observe, identify, take notes of the subjects’ acts, and through the memories, to make judgments of a final evaluation. As for the integrity of the psychological activities, cognition and personality is an organic entirety, which includes both the respective role and interaction. Two studies were implemented: Study I firstly examined the appraisal features of 33 novice raters, and then dePned specifically training methods for them. Study II in ensuring the effectiveness of the scores, by testing the working memory and personality of 44 novice raters, discussed the influences on the rating accuracy. The main conclusions are as follows: Study I found, for the novice raters, (1) rating effectiveness was low, included low rater consistency, low accuracy of distinguishing dimensions and candidates; (2) there were neither halo effect nor tolerance effect; (3) psychological and experienced subjects had higher rating effectiveness; (4) the higher quality of the records, the more effectiveness of the rating. In view of the above characteristics, (5) depended the training method in a combination of behavioral observations and frame-of-reference training.Study II found that: firstly, through training, (1) rater consistency and rater accuracy improved; (2) rating was also stringent, but the disparity across the dimensions minished. In addition, (3)the stronger ability of attention conversion and restrain, the more steady nature would enhance distinction accuracy of the dimensions; (4) the stronger of the determination, the higher accuracy of the distinction among subjects; (5) the more altruism, the more prone to rater error, reflected in the high halo effect, as well as loose tends; but the more resilient, the less prone to halo effect.