GENERATING BIODATA FROM SCRATCH: CREATING AND VALIDATING A COMPREHENSIVE COLLECTION OF BIODATA ITEMS AND SCALES

有效的入职前测试有助于聘用工作表现更好的合格员工。生物数据清单是一种招聘前评估，它使用有关生活史和过去经历的问题来有效地评估各种结构的求职者。研究一直支持他们在招聘方面的有效性，但生物数据的实施受到生物数据问题的专有性质和可获得性限制的阻碍，以及工作环境中关于个别项目和量表属性的数据不足。为了克服这些挑战，该项目利用自然语言处理(NLP)来建立最全面、可公开访问的生物数据存储库。与链接的工作分析相结合，该存储库存储了数十个结构中的数千件物品，包括它们的心理测量属性。因此，该项目有助于使生物数据民主化，使用户能够在组织内创建量身定做的生物数据量表，并对不同类型的生物数据清单的效力进行严格的测试。此外，这个项目将探索生物数据在减少不利影响和测试偏差方面的潜力，解决人员选择中的一个问题。这个项目从制作原型项目开始，以评估数十个与工作相关的结构。这个初始题库正在扩大，方法是通过NLP使用自动项目生成，并与研究人员精炼和内容有效性判断相结合。正在收集生物数据内容的数据，以及与工作相关的相关变量。正在开发经验的、合理的和混合的评分关键，以及一个工作分析工具，以帮助用户识别最佳的生物数据内容，并根据工作分析估计有效性。最后，检查所有内容的潜在组别差异，以确定对不利影响或统计测试偏差的易感性。总之，这项研究建立了一个全面的生物数据存储库，并与生物数据量表和项目的可靠性、有效性和不良影响信息的经验数据库相结合，从而改进了招聘方法，扩大了生物数据知识和专业知识。该奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

Valid pre-employment tests facilitate hiring qualified employees who perform better at work. Biodata inventories are a type of pre-hire assessment that uses questions about life history and past experiences to effectively evaluate job applicants across a wide range of constructs. Research consistently supports their efficacy in hiring, and yet biodata implementation is hindered by the proprietary nature and restricted availability of biodata questions, alongside insufficient data on individual item and scale properties across work settings. To overcome these challenges, this project leverages natural language processing (NLP) to build the most comprehensive, publicly accessible biodata repository. Paired with a linked work analysis, the repository houses thousands of items across dozens of constructs, including their psychometric properties. This project, therefore, helps democratize biodata and empower users to create tailored biodata scales within organizations, as well as provide rigorous tests of the efficacy of different types of biodata inventories. Moreover, this project will explore biodata’s potential for reducing adverse impact and test bias, addressing a concern in personnel selection.This project begins by crafting prototypical items to assess dozens of work-related constructs. This initial item pool is being expanded by using automated item generation via NLP, paired with researcher refinement and content validity judgments. Data are being gathered for the biodata content, along with work-related dependent variables. Empirical, rational, and hybrid scoring keys are being developed, and a job analysis tool as well to help users identify optimal biodata content and to estimate validity based on the job analysis. Finally, all content are examined for potential group differences to determine susceptibility to adverse impact or statistical test bias. In sum, this research establishes a comprehensive biodata repository paired with an empirical database of reliability, validity, and adverse impact information for biodata scales and items, thus leading to improved hiring methods and expanding biodata knowledge and expertise.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.