Artificial Intelligence in the Staffing Process: Performance Comparisons of (Un)supervised Learning for the Screening of Job Applications

Artificial intelligence has the potential to change not only technology related jobs but also administrative jobs. While work on artificial intelligence has a relatively long tradition in research areas such as computer science or statistics, human resource management is an area in which the topic has only a short history. There are many proposals for the use of artificial intelligence in human resource management, but there is a lack of empirical examples. This paper attempts to exemplify a concrete use case through a pilot study with three small datasets. Using simple cluster analysis from unsupervised learning and a neural network from the supervised learning methods, application letters are analyzed. Depending on the quality and size of the dataset, it can be shown that cluster analysis can significantly replicate human judgment in terms of application quality groups via document-word matrices and furthermore determine the optimal number of groups for judgment. Thus, recruiters would only need to look at a few applications for each determined group in order to determine the quality of the specific group. With the use of the neural network, an increase in performance could be achieved both in the replication of human judgments and in the case of the optimal number groups for judgment depending on the number of hidden neurons used. However, this is limited to the training data set. In this respect, it can be concluded that the use of both methods represents a promising approach for the introduction of artificial intelligence in recruiting.