Voluntary employee turnover incurs significant direct and indirect financial costs to organizations of all sizes. A large proportion of voluntary turnover includes people who frequently move from job to job, known as job-hopping. The ability to discover an applicant's likelihood towards jobhopping can help organizations make informed hiring decisions benefiting both parties. In this work, we show that the language one uses when responding to interview questions related to situational judgment and past behaviour is predictive of their likelihood to job hop. We used responses from over 45,000 job applicants who completed an online chat interview and also self-rated themselves on a job-hopping motive scale to analyse the correlation between the two. We evaluated five different methods of text representation, namely four open-vocabulary approaches (TF-IDF, LDA, Glove word embeddings and Doc2Vec document embeddings) and one closed-vocabulary approach (LIWC). The Glove embeddings provided the best results with a positive correlation of r=0.35 between sequences of words used and the job-hopping likelihood. With further analysis, we also found that there is a positive correlation of r=0.25 between job-hopping likelihood and the HEXACO personality trait Openness to experience. In other words, the more open a candidate is to new experiences, the more likely they are to job hop. The ability to objectively infer a candidate's likelihood towards job hopping presents significant opportunities, especially when assessing candidates with no prior work history. On the other hand, experienced candidates who come across as job hoppers, based purely on their resume, get an opportunity to indicate otherwise.

No abstract

In this work, we demonstrate how textual content from answers to interview questions related to past behavior and situational judgement can be used to infer personality traits. We analyzed responses from over 58,000 job applicants who completed an online text-based interview that also included a personality questionnaire based on the HEXACO personality model to self-rate their personality. The inference model training utilizes a fine-tuned version of InterviewBERT, a pre-trained Bidirectional Encoder Representations from Transformers (BERT) model extended with a large interview answer corpus of over 3 million answers (over 330 million words). InterviewBERT is able to better contextualize interview responses based on the interview specific knowledge learnt from the answer corpus in addition to the general language knowledge already encoded in the initial pre-trained BERT. Further, the “Attention-based” learning approaches in InterviewBERT enable the development of explainable personality inference models that can address concerns of model explainability, a frequently raised issue when using machine learning models. We obtained an average correlation of r = 0.37 (p < 0.001) across the six HEXACO dimensions between the self-rated and the language-inferred trait scores with the highest correlation of r = 0.45 for Openness and the lowest of r = 0.28 for Agreeableness. We also show that the mean differences in inferred trait scores between male and female groups are similar to that reported by others using standard self-rated item inventories. Our results show the potential of using InterviewBERT to infer personality in an explainable manner using only the textual content of interview responses, making personality assessments more accessible and removing the subjective biases involved in human interviewer judgement of candidate personality.